Final Report Summary - INDICSER (Indicators for evaluating international performance in service sectors)
Executive Summary:
The main objective of the INDICSER project was to provide information regarding the performance of service sectors, filling a gap in the existing knowledge base. The project had three strands – measurement and conceptual issues, developing indicators for evaluating service sector performance and analytical research on various aspects of that performance. The main highlights of the project are as follows.
Measurement issues tackled in the project included extending EU KLEMS so industries were classified according to NACE rev 2 and producing internationally consistent series on labour force composition. The project devoted effort to estimating investments and capital stocks for intangible assets, starting with novel estimates for firm specific human capital (training) and firms’ own account organisational capital using harmonised databases, such as the EU LFS. It developed composite indicators for innovation and ICT use and new measures for labour market regulation and the market environment. In non-market sectors the project applied a number of approaches to measuring performance in the health sector and demonstrated the practical difficulties that researchers encounter in trying to carry out international comparisons in this area. Educational attainments were evaluated against the Europe 2020 targets while a novel study compared the research performance of the University sector in the EU. The conceptual research identified areas where current practice was problematic and suggested alternative approaches, e.g. for financial services, the impact of health interventions and new measures for collective services.
The project produced a number of data series, both cross sector and for specific industries, which are available for public use on the project website – www.indicser.com. Labour composition variables include employment and wage bill shares distinguishing gender, age and skill for 1-digit industries, 24 EU countries and covering the period 2002-2009. The website contains indicators by industry of innovation, ICT use, intangible assets, competition, market structure, and product and labour market regulations. Sector specific indictors include a comprehensive database on the EU financial services sector including, for example, measures of bank output, efficiency, and international activity. Indicators for the education sector include enrolments, quality adjusted outputs and labour productivity growth.
The analytical research on the market services sectors suggested much larger increases in labour hoarding in the period covered by the financial crisis relative to previous downturns. A study of employment protection legislation indicated that this form of labour market regulation resulted in lower productivity growth but also lower wages and appeared not to have an adverse impact on employment. Research for the financial services sector highlighted potential negative impacts of financial regulation on growth should the latter lead to less international integration. Impacts of financial services on investment, however, depend on countries’ degrees of uncertainty aversion, bank market power and the degree to which firms depend on external finance. Other research on market services includes papers on growth accounting impacts of training and other intangible assets, the links between sectoral risk, intangible assets and rates of return and the impacts of ICT on output volatility at the sector level. Research papers for nonmarket services include an examination of equality of opportunity in both health and education and measures of education inequalities that also take account of outcome inequalities. A study for the UK examined the impact of early years (pre-school) education on pupils’ subsequent performance. Finally the project highlighted some important areas where significant work is needed to ensure performance can be measured across EU countries – these include services producer price indices, bank output, health sector performance and output of collective services.
Project Context and Objectives:
The primary motivation for this project was to produce indicators that would inform the academic and policy communities on service sectors as an engine of growth. At the time the proposal was drafted the productivity literature suggested an important role for service sectors in explaining Europe’s productivity gap with the US (Timmer et al. 2010). This literature pointed to growth driven by fast adoption of and intelligent use of information and communications technology (ICT) which was concentrated in service sectors (Van Ark, O’Mahony and Timmer, 2008). In addition the proposal was written near the start of the financial crisis and so suggested the need for increased attention on financial markets, how they operated and measurement of the output of this important sector.
While much of this literature was focused on market services a parallel literature was emerging which highlighted the role of non-market services in aggregate growth. These services, comprising education, health, collective services and public administration, represented a large part of economic activity and so efficient production in these sectors had a significant impact on growth. This became more important after the financial crisis as austerity measures instigated by many governments could be mitigated to some extent by enhanced productivity in these services financed largely by government. However, evaluating the performance of non-market services was severely constrained by the fact that their output was poorly measured.
While the above provided the main motivation for the project, the way it progressed was altered to some extent by the changing economic and policy environment. The Europe 2020 agenda became more defined after the start of the project and it became apparent that performance in production industries, especially manufacturing, is very much part of this agenda. This meant that the measurement exercise, although initially focused on services, was broadened in many respects to ensure all sectors of the economy were covered.
The main objectives of the INDICSER project as set out in the programme of work were to develop indicators which are able to provide information on the determinants of growth in service sectors in the European Union. To this end, the proposed work programme suggested the following criteria for the development of indicators:
- The indicators need to cover measures of industry-level output, input and productivity growth
- The indicators need to cover determinants of productivity, such as innovation, technology and market environments
- The indicators need to be feasible and internationally comparable
- The indicators need to be relevant and useful for economic and social policy analysis
The project divided services into two broad groups, market services and non-market services, as the measurement issues and state of development of the data required varied across the two.
In the case of market services the aim was to develop indicators of the determinants of output, such as skills, ICT and intangible capital, and the market environment in which firms operated with the financial sector receiving particular attention. Alongside the construction of indicators was a drive to understand the main measurement constraints faced by the national and international statistical bodies in measuring performance in service sectors. This involved both analysis of conceptual problems and data constraints. In addition some analysis of the impact of these indicators in explaining the drivers of productivity growth was incorporated in the work programme.
For non market services a first objective was to consider what indicators should we be measuring and how these relate to measures in use for market services. This was to be followed by analyses of data issues and practical importance. For non-market services such as health and education, the main emphasis was on measuring output and applying novel concepts of output measurement to a range of countries.
The project was divided into seven main work areas corresponding to the following workpackages and objectives:
WP1: Input, output and productivity trends in service sectors
The objective in this work package was to provide the basic data on service industry outputs, inputs and productivity, by updating and extending the EU KLEMS database, dealing with changes to the industrial classification, updating to add some years following the financial crisis and a re-examination of the labour composition series.
WP2: Technology, Innovation and Intangibles Indicators
This aimed to develop new indicators of innovation, technology and intangible investment in market services and to gauge their impacts on growth.
WP3: Market Environment Indicators
The objective of this work package was to develop indicators for the market environment interpreted in a general sense to cover the markets in which service sector firms deliver their outputs and purchase inputs and so includes measures of competition, regulation and international markets and evaluate the usefulness of these indicators in explaining growth and employment.
WP4: Financial Services
Work Package 4 is the first of the sector specific areas of research and focused on the financial services sector. With the onset of the financial crisis it became more important than ever to understand developments within the industry as well as the industry’s impact on economic growth. This involved both a critical evaluation of measurement methods and the compilation of a range of financial system performance indicators, including bank efficiency estimates and indicators of the linkages between the financial sector and non-financial industries.
WP5: Health services
The aim of this research was to review current practice in measuring both output and productivity performance of the healthcare production sector and to produce estimates for a few selected countries. Indicators of value for money were also planned, again for a few countries. It also aimed to review the literature on innovation in health care and develop methodologies for taking account of this for particular treatments.
WP6: Educational services
The aim of the research on education was to develop indicators of educational attainment, education inequalities, volume and quality adjusted output measures and indicators of productivity performance across a range of countries. Experimental quality indicators were also to be reviewed for the UK.
WP7: Experimental output indicators
This last part of the main work programme was designed to devise possible measures of output for a number of difficult to measure sectors in both market and non market sectors. These included insurance, distributive trades, collective services and university research.
In addition to the above work programme, the project aimed to provide a summary of the data constructed, an overview of the main methodological and data issues encountered throughout the project and recommendations for improving the evidence base – this was main objective of WP8: Synthesis and Recommendations.
Project Results:
Each workpackage of the INDICER project produced three types of outputs:
* Papers on conceptual and measurement issues
* Development of Indicators
* Analytical papers on drivers of productivity and economic performance across EU countries.
This report considers each of these three areas in turn
Conceptual and measurement issues
Market services
An output for Workpackage 1 was a conceptual paper on output measurement in market services, titled ‘A Quarter of a Century Progress Report on the Services Sector Productivity Statistics: A Europe-United States Perspective’ by Tarek Harchaoui. This paper was motivated by the observation that the deterioration from the mid 1990s in Europe’s productivity performance relative to the US coincided with the ‘renaissance’ of the US statistical system, which has been upgraded in many important respects. With these efforts, there is now a consensus in the economics profession that the US statistical system has set a new frontier in official statistics. This paper raises the natural question whether measurement differences could account for much of the observed productivity gap, i.e. was the European statistical system ‘left at the station’ while its US counterpart ‘departed’. This paper’s retrospective examination of the development of the services sector productivity statistics in both Europe and the US suggests the presence of some circumstantial evidence in support of measurement differences. The evidence based on a ‘structured guess’ suggests that the upgrade in the US services sector statistics translated into two kinds of enhancements in the post-1995 period—a considerable reduction in the contribution of industries that traditionally dampened the aggregate productivity trend combined with a higher contribution of those that generally lifted it. Those industries that dampened productivity include most of business services, hotels & restaurants and personal services whereas the dynamic productivity sectors included retail, wholesale and transport. This contrasts markedly with Europe where the contribution of these two sources remained unchanged. Although much was achieved in Europe following the publication of Eurostat’s ‘Handbook on Price and Volume Measures in National Accounts’ this retrospective examination concludes that much more progress is required if Europe is to catch up with the US in measurement of economic statistics. In particular there are important gaps in terms of scope of the service producer price index program and the timing of its implementation.
A number of conceptual and measurement outputs arose from the research carried out for workpackage 2. An early contribution, which has since been published in the Review of Income and Wealth was ‘Human Capital Formation and Continuous Training: Evidence for the EU Countries’ by Mary O’Mahony. This put forward a method for estimating investment in intangible training capital, a large component of firm specific intangible capital identified in the pioneering paper on measuring intangible capital by Corrado, Hulten and Sichel (2005, 2009). The paper suggested using information on training propensities and duration of training from the EU LFS combined with earnings data from EU KLEMS and so enabled measurement of intangible training capital for most EU25 countries (apart from Malta) and by 1 digit industry group. The raw data suggested that the most highly skilled were most likely to receive training, more women were trained than men and training was also concentrated in younger age groups. The estimates of training investment showed some interesting variations across countries and industries. Training capital as a percent of GDP was higher in market services than in production industries, relatively high in all industries in the UK and some Scandinavian countries and low across the board in countries with poor relative productivity performance such as Italy and Greece. Training capital intensity was significantly higher in the EU15 group of countries than in the new member states.
Following on from this first paper on intangible capital, the research teams at BHAM and ZEW then investigated the feasibility of measuring all types of intangible assets by industry. The feasibility of constructing intangible capital estimates for all categories used in previous projects such as INNODRIVE was uncertain at the beginning of the INDICSER project and became possible after the release of the INTAN-Invest data in 2012 (www.intan-invest.net) described in Corrado et al. (2012). The methodology employed is described in the paper ‘Intangible Investment at the Industry Level: Growth Accounting’ by Thomas Niebel, Mary O’Mahony and Marianne Saam.
Much of the work on indicators of innovation and ICT usage involved putting together data series from a range of sources as described below. Christian Rammer from ZEW authored a methodological note, ‘A Summary Index on Innovation in Services’ on constructing a composite indicator of innovation by combining various series from the Community Innovation Surveys (CIS). A set of 20 innovation indicators representing both inputs to and outcomes of innovative activities of service firms were used and weighted by a value that represents the significance of the indicator for total performance. The data were taken from the three most recent waves of the Community Innovation Surveys (CIS) and relate to the years 2004, 2006 and 2008. The Summary Index on Innovation in Services (SIIS) methodology is in line with that used for the Innovation Union Scoreboard of the European Commission (European Commission, 2012). The results show considerable differences in innovation performance of service industries. The ranking of service industries by their degree of innovativeness largely confirms prior work but this paper produced more detailed results on the industry level, including summary indexes for service industries based on NACE rev. 2.
The results show that R&D services (division 72 in NACE rev. 2 and 73 in NACE rev. 1.2) report by far the highest SIIS value which is more than 50% higher than the SIIS value of the industry that follows, which is computer programming and consultancy (62 / 72). The high value for R&D services is straightforward since producing new knowledge and innovations is the main purpose of the economic activity of firms in this sector. High SIIS scores are also reported for telecommunications (61 / 64, in 2004/2006 including postal services), broadcasting (NACE rev. 2 division 60) which was not part of the sector coverage of CIS in 2004 and 2006 and information services (NACE rev. 2 division 63, which was part of divisions 72 and 92 in NACE rev. 1.2). Other service industries with a high SIIS are financial intermediation (divisions 64 to 66 in 2008 and 65 to 67 in 2004/06) and architectural/engineering services (71 in 2008, 74.2/74.3 in 2004/06). Services industries that show a low innovativeness in all years are trade and repair of motor vehicles (50 / 45), retail trade (52 / 47), real estate (70 / 68) and rental and leasing (71 / 77). In 2008, another group of services industries not covered in earlier CIS also report a low SIIS, including legal advice and accounting (69), employment services (78), security services (80), and cleaning and facility management (81). The results were found to be robust over time and to changes in the measurement assumptions.
The SIIS also varied considerably by country and the results differ quite considerably compared to country rankings for other innovation indicators such as R&D expenditure per GDP, patent applications per population or the summary index of the Innovation Union Scoreboard (European Commission, 2012). Estonia reports the highest SIIS score, followed by Cyprus and Portugal. From the group of countries that are considered as ‘innovation leaders’ in the Innovation Union Scoreboard, only Denmark and Germany are found among the top performers in services innovation whereas Sweden only ranks tenth and Finland twelfth in the SIIS. The UK and the Netherlands, which are strongly specialised on service sectors and are both often regarded as leading new trends in services, report low SIIS scores which place them on rank 17 and 19, respectively, of the EU-27 countries.
The research team at ZEW also constructed a composite index for ICT, reported in ‘An Internet-Based Index on ICT usage (IICT)’ by Daniel Cerquera, Miruna Sarbu and Patrick Schulte. This was based on data from The Community Survey on ICT usage in E-Commerce (EUROSTAT ICT survey). The index covers the period between the years 2002 and 2009, for 13 countries with the information disaggregated for 18 industries. This composite indicator was based on five series all of which had good coverage in the surveys. The results show considerable diversity across services sectors and across countries but generally an upward trend in ICT usage. Sweden, Finland and the Netherlands had the highest average ICT usage. In all 13 countries the communications sector (NACE rev 1 64), computing services (72) had ICT usage significantly above average. Other sectors that had high usage were wholesale trade (51) and film and broadcasting (921-922) and electricity, gas and water supply (40-41 for all countries apart from Finland. Within manufacturing the chemicals, fuel, rubber and plastics group had the highest ICT usage but this was only marginally above the average across all industry groups. Construction, restaurants and personal services had generally well below average ICT usage. These results confirm the importance of ICT usage in service sectors.
Workpackage 3 was concerned with developing indicators that described the market environment in which firms operated. In terms of competition and market structure the project concentrated on constructing indicators based on company accounts data as a supplement to those readily available elsewhere, e.g. in the Eurostat structural business statistics. Given the well known problem of lack of coverage of small firms in company accounts the indicators chosen concentrated on those that captured diversity across firms, although some standard indicators such as the Herfindahl and price cost margins were also included (see discussion below). Two variables that captured the dynamics of industry performance were productivity dispersions (measured by differences between the top and bottom cohorts of firms) and a variable which captured the extent to which sales were concentrated in younger firms. These variables are best constructed at the 3 digit industry level. An econometric analysis was undertaken at the company level to construct Boone indicators of the elasticity of profits with respect to variable costs.
The indicators vary considerably across countries and industries. Nevertheless some general conclusions emerge. Productivity dispersions tend to be much larger and more frequent in services industries than in production, especially at the 3 digit level. Although the average age of firms is generally lower in services than in production industries, there were proportionally more industries located in manufacturing and other production sectors that were dominated by younger firms than in services. This may seem surprising at first sight but many of the sectors that produce new products are in fact in manufacturing – these include manufacturing of computers, electronic goods and instruments. Services industries such as retail trade and other business services appear to be more dominated by older firms in the past decade. Nevertheless some services such as communications and R&D services in general are dominated by younger firms. The Boone indicator suggests that profits in services industries tend to be less responsive to changes in variable costs than in production, especially manufacturing. Together these findings paint a picture of services being less competitive and less dynamic than manufacturing.
A number of conceptual innovations arose from workpackage 4 on financial services. In a paper published in the Review of Income and Wealth, ‘Bank Output Measurement in the Euro Area: a Modified Approach’ by Antonio Colangelo and Robert Inklaar, the authors argue for a new approach to measuring bank output. Banks do not charge explicit fees for many of the services they provide, bundling the service payment with the offered interest rates. This output therefore has to be imputed using estimates of the opportunity cost of funds. They argue that rather than using the single short-term, low-risk interest rate as in current official statistics, reference rates should match the risk characteristics of loans and deposits. The authors estimate that this would lower euro area imputed bank output by, on average, 28–54 percent compared with the current methodology, implying that euro area GDP (at current prices) is overstated by 0.11–0.18 percent. They also show that this adjustment leads to more plausible shares in value added of income from fixed capital in the banking industry.
There are also conceptual issues in measuring bank output in constant prices. The INDICSER review paper ‘Output Measurement in Market Services’ by Robert Inklaar and Marcel Timmer contrasts the approaches used by statistical offices in the US and Europe. While the US bank output series in the National Accounts is based on a ‘constant services per loan/transaction’ basis, in European countries it is usually based on a ‘constant services per euro’ basis. The issue can be illustrated using the example of residential mortgages. This is a straightforward loan category since such a loan is used for buying a house. Consider the typical case in many countries of a few years ago when house prices were rising rapidly, also relative to the general price level. If the average house price increases, the average mortgage loan tends to increase by a similar amount. So if the average mortgage increases by, say, 10 percent, does the bank have to expend 10 percent more effort in screening and monitoring (constant services per euro)? Or is a more plausible assumption that every mortgage requires a similar amount of screening and monitoring over time (constant services per loan)? A similar argument can be made on the deposit side. A deposit account is mostly useful because it provides access to many transaction services in addition to safekeeping. So if account holders value these transactions, it makes sense to count the number of transactions to arrive at a measure of depositor services at constant prices. So again, ‘constant services per transaction’ seems preferable to ‘constant services per euro’. This paper examines the relative magnitudes of these differences in measurement practices on the two sides of the Atlantic. They suggest that since 1995, European bank mortgage output has been systematically overstated compared to the US, but on the deposit side that European growth would have been higher if the same methods as in the US had been in use. Deposit accounts and residential mortgages account for only part of bank output and their relative importance in total bank output varies considerably over time. Still, what can be safely concluded is that the differences in measurement practices between European countries and the US make it very hazardous to compare the resulting output series from the National Accounts.
Workpackage 7 produced two conceptual papers relating to the Market services sectors. The paper ‘Measurement of Insurance Output: Application of the Gross Output Approach’ by Martin Weale discusses the measurement of the gross output of the insurance sector, arguing that this should be done with reference to this risk carried. The implication of this is that an increase in insurance premia as a result of an increase in risk is treated as a volume increase and not a price increase. The consequences of this approach for the measurement of the gross output associated with life insurance and annuity provision as well as general insurance is discussed. It is shown that changes in annuity rates resulting from changes to mortality rates should be treated as volume changes. The approach is illustrated with reference to data on property insurance drawn from the United Kingdom’s Living Conditions and Food Survey. The illustrative data suggest the proposed measure can differ from alternatives based on deflated premia or numbers of policies but the author notes that the quality of the data is poor and so these calculations should only be seen as illustrative. The choice between a measure based on the value of insured property and a measure based on premia deflated by building costs has to depend on expert information as to why premia have changed over time. If the change is largely because the running costs of insurance companies have changed or because more property is being insured, then the measure based on estimated insured risk is appropriate. If, however, a driving influence has been changes in risks of pay-out, then the deflated premia measure is more appropriate. If the increase in premia has been driven by an increase in the value of property insured, then the insured risk measure and the deflated premia measure should coincide. Comparing this approach to the current one of either measuring numbers of policies (e.g. vehicle years insured) or deflating a nominal measure of revenue using the GDP deflator, it seems likely that, it might result in rather more volume growth relative to price growth, at least in the UK. Car insurance premia have risen recently because of perceived increased risk, observed as increased claims. This should be seen as a volume increase. To the extent that a volume measure of the cars people insure has increased, because of improvements in the quality of cars, this should be treated as a volume increase while reliance on vehicle years, which is current UK practice, means it is counted as a price increase.
The other market services sector which was covered in workpackage 7 was Retail Trade. The paper ‘The Europe-U.S. Retail Trade Productivity Gap in a Rear-view Mirror’ by Tarek Harchaoui asks the question is the Europe-US retail trade productivity gap a genuine phenomenon or the result of a variety of measurement issues. The paper provides a review of methodologies currently in use to measure the output of distributive trade services. This highlighted the main shortcomings of current methodologies and suggested improvements. As the US has changed its output measurement methodology, while Europe has not, the paper provides a new US-EU comparison based on a harmonised methodology to provide a more realistic comparison. With a harmonized measure of real output, the post-1995 period now suggests a 0.5 percentage point productivity gap in favour of the US, down from the ‘official’ 1.5 percentage points. The productivity gap in favour of the US retail trade sector still holds albeit with a more modest order of magnitude. This new gap is further reduced to one-quarter of a percentage point as a result of a counterfactual experiment that asks what would productivity performance look like had differences in the economic structures (scale economies) between the two economies been accounted for.
Non-market services
The research carried out for Workpackage 5 on health services is an illustration of ambition being severely constrained by poor data. The conceptual framework for measuring the output and productivity of the health sector was set out in ‘Measuring the productivity of the Healthcare sector: Theory and Implementation’ by Antonia Huttl, Matilde Mas, Agnes Nagy, Guldem Okem, Mary O’Mahony, Erica Schulz and Lucy Stokes. This advocated measuring output as a cost weighted activity index adjusted for quality and measuring inputs by cost weighted aggregates of labour, capital and intermediates. This paper also set out the data requirements for estimating health sector output and productivity.
In practice it proved difficult to match outputs across time and countries. The research began with an attempt to compare hospital inpatient outputs across four countries, Germany, Hungary, Spain and the UK. Hospital procedures are classified according to Diagnosis Related Groups (DRGs). Although these frequently have similar titles they cannot easily be matched across countries – this was one finding from the FP7 funded project Euro DRG (www.eurodrg.eu). However such matching across countries is not required if the interest is in aggregate growth or by broad disease category, since we can estimate a cost weighted index for each country separately and compare these aggregates. The greatest difficulty in matching hospital procedures across time occurred in Germany as described in detail in the paper ‘Approaches to calculate CWOI for DRG hospitals in Germany’ by Erika Schulz. In all countries there were changes in DRG groups over time, mostly splitting groups into finer ones to more accurately describe treatments. This type of change was relatively easy to deal with through various aggregation procedures in the two adjacent years when the change occurred. However in Germany as well as changes in the number of DRG groups patients migrated between DRG groups. Therefore the German researchers had to resort to using confidential patient level data to match across time and these data were only available for a limited number of years. The consequence was that estimates were available for a shorter time period for Germany than for the other countries. Countries also vary in the way they estimate unit costs by treatment - the estimation of unit costs are discussed for Germany and Hungary in the two papers mentioned above, for Spain in ‘Diagnosis-related groups (DRG). Estimation of average weights and costs’ by Matilde Mas and Juan Perez Ballater and for the UK in ‘Unit cost data for England: A note on Reference Costs data and methodology’ by Lucy Stokes.
As well as volume of activities, it is necessary to try to adjust these for quality change to arrive at a cost weighted output index. In all four countries there were data on in-hospital death rates at the DRG level so it was possible to use survival rates as a quality adjuster. However this is far from the ideal adjustment for quality since the majority of treatments are for non life threatening conditions and for these we should use information on health status before and after treatment. Such information is not routinely available for any of the countries studied. The UK recently produced experimental measures for a few conditions – these are reviewed in the paper ‘Review of indicators of outcomes and effectiveness of treatment for some specific diseases’ by Mary O’Mahony and Lucy Stokes – but results are not available for the time period covered by the INDICSER project. The project team experimented with using the self reported health status measures in the EU Survey of Income and Living Conditions (EU SILC) but these showed such large variation across country and time, even adjusting for demographic differences, that they were deemed unusable.
While the attempt to measure output for hospital inpatients was reasonably successful, matching other outputs were not feasible – these outputs account for about 50% of healthcare expenditures excluding primary care. Such activities include outpatient treatments, rehabilitation and long term care, outpatient mental health treatments, ambulance services, diagnostic tests and other procedures such as renal dialysis. The need to take account of these activities was highlighted in the research on Hungary, described in the paper ‘Volume and productivity of the Hungarian inpatient health care – a case study’ by Antónia Hüttl and Ágnes Nagy. The Hungarian healthcare system went through a major restructuring in the mid 2000s with many patients being transferred from hospitals to other parts of the system, including rehabilitation and long term care. Just examining hospital inpatients alone gives a poor impression of the performance of healthcare in Hungary and so it is essential to include these other areas of treatment. The UK has the most complete data on non-hospital outputs of the four countries included, with detailed data on volumes and unit costs for a large and growing number of activities. In Spain there were data on volumes but no unit costs whereas in Germany there was some information on rehabilitation and mental health but not other activities. For Hungary the researchers had access to volumes and unit costs for both rehabilitation and mental health. Primary and community care were not included in the analysis given the poor quality of data in all countries.
Nevertheless the project did produce an international comparative analysis of the performance of the healthcare sector which is not available in the literature to date. In ‘Output and productivity growth in the healthcare sector: a study of four European countries’ by Antónia Hüttl, Matilde Mas, Ágnes Nagy, Mary O’Mahony, Erika Schulz and Lucy Stokes, estimates of hospital output, other output, labour input and labour productivity growth are compared across the four countries over the period 2003-2009 (2004-08 for Germany). In all countries cost weighted activity growth is greater than the growth in the sum of activities so that on average the number of patients treated for the most costly procedures are growing faster than for lower cost procedures. The paper argues that this is likely to be due to the increased share of older and elderly patients in each country. The cost weighted activity indexes show the highest growth in the UK by about two percentage points per annum higher than the next country, Germany, Spain’s activity growth is a little lower than Germany’s and in Hungary growth was significantly negative, due to the restructuring mentioned above. Adjusting for survival rates makes a small contribution to output growth with the greatest impact in the UK. In general nonhospital outputs grow at slightly higher rates than inpatient treatments, with the greatest difference in UK outpatients which were growing at a rate of about 8% per annum over the period 2003-2009 – outpatients represented about 20% of UK expenditures. Overall adding the additional outputs raised UK growth over the period 2003 to 2009 to 5.6% per annum as against 5% for inpatient treatments. In Hungary, taking account of rehabilitation and long term care turned around the overall negative growth in cost weighted activity to small positive growth. In Spain and Germany the available evidence suggests also an upward adjustment comparable to that in the UK. Examination of output growth by broad disease category also shows some striking differences. Growth in Germany is the most similar across disease categories while the expansion in output in the UK is fairly evenly spread but somewhat more variable than foe Germany. In Spain and Hungary in contrast growth tends to be more concentrated in a few conditions.
The ideal measure of productivity is multi-factor productivity which takes account of labour, capital and intermediate expenditures such as drugs. It was not possible to obtain data on capital and intermediates in a form that was comparable across the countries. Therefore attention focused solely on labour productivity growth. Here we matched labour input to the output measure available so that labour productivity growth referred only to hospital treatments for Germany, all outputs other than primary and community care for the UK and intermediate measures for Spain and Hungary. In all countries the growth in employment of relatively high paid doctors was much more rapid than for nurses so labour input was quality adjusted by weighting each type of labour by its relative earnings. The results show both the UK and Germany experiencing labour productivity growth rates close to 3% per annum, but with very different profiles. Germany’s labour input was almost flat whereas in the UK this was growing at about 2% per annum. This is likely to reflect catch up of the UK to Germany as the latter spent a greater share of GDP on health over many decades. In Spain labour input grew faster than output so there was a small negative growth in labour productivity. In contrast labour input fell by more than 2% per annum in Hungary so its labour productivity growth was on a par with that achieved in the UK and Germany.
Overall these results suggest significant differences in the performance of the healthcare sectors across the four countries. With extensions to more countries and time periods it might be feasible to investigate how much of this is due to demographic and disease incidence differences relative to variations in systems of provision. However such an extension would require access to better data than appears to be currently available.
Although productivity is a useful measure of how much physical output can be produced per unit of physical input it does not take account of the welfare value of treatments to patients or the costs of these treatments in terms of payments to inputs, especially labour. Countries are likely to vary significantly in the extent to which they pay medical professionals which in turn may reflect restrictions on entry. In addition the suppliers of medical equipment and drugs may charge different rates to different healthcare suppliers, depending on the system of provision. The second main task of this workpackage was to set out a framework for estimating value for money indicators for healthcare. The theoretical framework is set out in ‘Health Capital: Methods and Data Requirements’ by Francisco J. Goerlich Gisbert. This paper presents the basic methodology for constructing Health Capital series following the ideas of Cutler and Richardson (1997, 1998, 1999). Health capital is defined as the discounted value of the utility from health of a person over time. Measuring this requires three pieces of information: numbers of years of life by age group from life tables, quality of life which can be derived from measures of self reported health status and the value of a year in perfect health (VSLY). The paper stresses the strong points and weakness of this particular methodology, as well as discussing linkages with other relevant and related literature, data requirements and possible solutions for a practical application to some European countries on the principle of comparability.
The conclusions from this paper can be summarised as follows. First, the value of a year in perfect health (VSLY) is of particular importance in these calculations. The effect of changing the VSLY is multiplicative in that it acts as a scale factor, and changing this can result in significant effects. The literature review suggests weak evidence regarding underlying estimates of the VSL, and this is a matter of concern for the purposes of the exercises proposed in the paper. Second, there is much discussion in the health literature about weights and how these have changed over time. There is currently no consensus among professionals of health care evaluation programs on how to elicit these weights. Third, there is also some debate about changes in disease prevalence. In particular, regarding the fact that prevalence changes according to medical innovations, treatment and diagnostic tools on the part of the medical profession, and not only because of true changes in population health. Fourth, discounting also has a significant effect on results, especially when comparing cohorts. With no discounting, infants have the largest improvements. On the other hand, with higher rates of discounting the emphasis is put on the elderly. Fifth, value for money indicators are of little use without detailed data on costs provided by health services. When we compare cohorts, or people of different ages, the lack of medical expenditure broken down by age is a severe limitation when analyzing the efficiency of the health care sector by means of value for money indicators. This is because health declines as age increases and also costs rise with age.
In ‘Health Capital: Practical implementation with an application for Spain’, by Francisco J. Goerlich Gisbert and Juan Perez Ballater, the authors first tried to implement the model with data for Spain which highlights the sensitivity of results to choices on the VSLY, discount rates and quality adjustments. The paper ‘Health Capital: an application for Germany, Hungary, Spain and the UK’ by Francisco Goerlich, Juan Perez Ballaster, Antonia Huttl, Mary O'Mahony, Erika Schulz and Lucy Stokes attempts to implement the model for the four countries included in the health sector productivity research. The paper illustrates the sensitivity of the estimates to the use of different assumptions in estimating health capital. In particular the results are sensitive to the use of a fixed VSLY across countries relative to one that assumes a unitary elasticity with respect to GDP. It compares the resulting measures of health capital aggregated across age groups, both quantity and quality adjusted, with information on healthcare expenditure to provide a broad indication of value for money of the different health systems. This employs the assumption of a fixed VSLY and 3% discount rate. In all four countries, value for money appears to decrease over the period 2005 to 2009, With Spain showing the greatest, and Hungary the least, decline. Health capital per capita showed small annual increases over this period, ranging from about 0.4% in Hungary and Germany to 0.2% in the UK and 0.04% in Spain. Therefore this declining value for money largely reflects the changes in health expenditure per capita. In terms of levels, while there are some differences in the magnitude of the numbers according to the approach taken, in all cases, Hungary demonstrates the greatest value for money on this basis, and Germany the least. However these estimates come with a serious health warning given the lack of data on health expenditures by age. The authors conclude that while the development of such data would be valuable for this purpose, given demographic changes and an aging society, such information is clearly also of much wider importance.
The third conceptual paper in this workpackage was concerned with measuring welfare benefits from specific health treatments. The paper ‘A Cost-benefit Analysis of Cataract Surgery based on the English Longitudinal Survey of Ageing’ by Martin Weale, explores the self- reported effect of cataract operations on eye-sight. A non-parametric analysis shows clearly that cataract patients report improved eye-sight after surgery and a parametric analysis using ordered probit analysis provides further information on the nature of this, showing that the beneficial effect is larger the worse was self-reported eye-sight ahead of surgery. For the 5.6% of patients with excellent eye-sight ahead of surgery, however, it is found that surgery is not associated with improved eye-sight in the short term. Nevertheless, the long-run effect is suggested to be beneficial. Calibrating the results to an existing study of the effect of imperfect eyesight on quality of life, the impact of cataract operations on quality-adjusted life-years is found to be similar to that established in specific studies and well above the costs of cataract operations in almost all circumstances. Thus the magnitude of the average expected life-time gain in welfare relative to the cost of surgery suggests that, overall, the widespread provision of cataract surgery is easily justified. The key contribution of this paper is to show that data on self-reported health- in this case eye-sight- in a general-purpose panel survey where patients are tracked through time can be used to examine the benefits of medical interventions, so as to work out the expected gain in welfare resulting from such interventions. Obviously surveys of this type can be used only for procedures which are widespread, since the sample has to contain enough cases for meaningful statistical analysis to be possible.
Similarly to the health sector research, workpackage 6 was concerned largely with measuring performance of the education sector. Here the volume measure of output, numbers of pupils enrolled at different levels of education, is much simpler than for health so the focus was much more on measuring quality of output. This also meant that a much larger number of countries could be compared. Describing progress in the volume of output in itself produced useful results, especially in relation to the Europe 2020 targets – these are given in the paper ‘Recent Developments in Selected Education Indicators and the Relation to Europe 2020 Targets’ by Mikkel Barslund, published in the National Institute Economic Review. The paper first looked at recent developments in educational attainment – measured by the number of years a 6-year old can expect to stay in education until his/her 30th birthday across EU countries, and compared the levels and changes in the EU27 and the US. Second, within this framework, the development in educational attainment differences for each gender was also investigated. Finally, the paper took a critical look at the Europe 2020 targets for tertiary education and made an assessment of which countries are likely to meet their national targets. The study builds on detailed educational enrolment data from Eurostat and, for the US, the Bureau of Labor Statistics.
The average expected number of years spent in schooling across the EU27 countries was 15.7 years in 2009, up significantly from 14.6 years in 1998. There are clear differences among countries. In Finland a 6-year old can expect to spend 17.4 years in the educational system before his/her 30th birthday, whereas in Bulgaria the figure is less than 14.5 years. All member states have seen an increase in years of school attendance during the period from 1998 to 2009. Increases are also unevenly distributed among member states, but there has been a tendency for laggards to catch up, so that inequality in educational duration has decreased within the EU27. By 2009, the EU27 countries had closed the educational attainment gap with the US. This contrasts with the situation in 2000, where children living in the US could expect to stay in education for more than half a year longer than the average child living in the EU. Most of this increase – three-quarters of one year – comes from increased tertiary enrolment. From 2000 to 2010, the share of persons aged 25 to 34 years with a tertiary education increased from 23% to 33%. The paper points to a large gender gap in favour of females spending more time in education than males, mostly in the 20-24 age group. The paper concludes that most EU countries will meet their 2020 targets but it is unlikely that France, Portugal, Austria, Spain and Ireland will meet their national targets. Ireland, France, Portugal – and to some extent Austria – had ambitious targets, given their starting points. The problem for Spain is that it has slid backwards in terms of enrolment and attainment. The paper also concludes that it would be preferable to supplement the Europe 2020 tertiary attainment target for 30-34 year-olds with an enrolment target for 20-24 year-olds.
Number of pupils enrolled is a crude measure of output and should be supplemented by adjustments for the quality of education attained. The project considered a number of different possibilities to adjust output measures for quality. In ‘Quality Adjusted Education Output in the EU’ by Yasheng Maimaiti and Mary O’Mahony estimates are presented for quality adjustments based on test scores for secondary education and cost weighted enrolments by field of study for tertiary education. The paper concludes that the PISA reading scores are reasonable quality indicators for secondary education. Separating field of study for tertiary education is potentially useful but there is very little information on unit costs by field.
An alternative approach is to employ earnings outcomes and information on the probability of being employed to distinguish different types of education. This is the approach adopted for post compulsory education in ‘Output growth in the post-compulsory education sector: the European experience’ by Mary O’Mahony, José Manuel Pastor, Fei Peng, Lorenzo Serrano and Laura Hernández. This paper produced comprehensive measures of output based on enrolment, completion rates, expected wages, employability and labour market participation issues. The paper presents a number of variants depending on how to value graduates versus non graduates at each level and whether education impacts on the probability of being employed.This approach was applied to estimate the rates of growth of the output of the post-compulsory education sectors of 27 European countries over the period 2005-2009. The results show the importance of complementing raw educational data with labour outcome information when measuring output in this sector. The average results for the European countries and the whole period indicate that the output per student would have grown at an annual rate of 0.33%. This means a rate of growth for output as high as twice the one corresponding to enrolment. However the heterogeneity in terms of the growth of the output per student across Europe is as high as found for enrolment growth. The more pronounced reductions are those of Latvia, Slovenia, Hungary and Poland. Other countries also show a decrease in that ratio, namely Ireland, Spain and the UK. In the opposite situation we find the remaining countries. For some of them the increase is quite substantial, in particular in Cyprus, Czech Republic, Netherlands, Belgium, Denmark and Austria. In all those countries the output per student in the post-compulsory education sector present annual average growth rates above 1.5% during the period 2005-2009. Hence quality adjustments based on labour market outcomes are potentially useful for post compulsory education.
While it was not possible to find a suitable adjustment for primary education, in the spirit of the experimental nature of much of the project, researchers at NIESR examined how useful school level data for the UK might be used as a potential quality adjuster. This is reported in the paper ‘Quality and Outcomes in English Schools: the relationship between school inspection and pupil attainment’, by Anitha George, Lucy Stokes and David Wilkinson. This work attempted to develop a new education quality index for England by exploiting data on school inspections collected by Ofsted (Office for Standards in Education, Children's Services and Skills) linked to data on pupil attainment. In the UK National Accounts, education output is quality adjusted using an index of pupil attainment at age 16. One obvious problem with this is that if quality of primary education improves then it could take up to ten years for this to appear in the quality adjustment. Therefore this paper tried to develop a more immediate quality indicator. It uses econometric models to assess the relationship between school quality in England and pupil attainment as measured by national assessments at age 7 (Key Stage 1, KS1), 11 (KS2) and 16 (KS4). The estimated coefficients from these models are then used to produce a quality index based on the share of schools that were rated ‘outstanding’ for each Key Stage of the English education system. The results suggested lower quality adjustments for primary education than secondary although the time period over which these calculations were undertaken was very short.
Growth in an overall quality adjusted output index and labour productivity growth rates were reported in the paper ‘Output and Productivity in the Education Sector’ by Mikkel Barslund and Mary O’Mahony. This paper combines the results from the quality adjustments to secondary and post compulsory education. The test score approach has its main advantages in dealing with compulsory education. Here it is difficult to link education to earnings as students up to the regulation school leaving age have no comparator group to measure outcomes without this level of education. In the aggregate estimates unadjusted enrolments were used for primary school and enrolments adjusted by test scores for lower secondary. Test scores do not work so well for post compulsory education. First the available international tests are administered to students at the end of compulsory education with little information available for those aged between 16 and 18. There are no international tests available for further and tertiary education. However at these levels it is possible to compare earnings of those with and without this education so arguably the earnings outcomes is a better measure. In this paper we use their baseline measure from the O’Mahony et al. paper above that counts non-graduates at each level as earning 50% of graduates and takes account of the impact of education on the probability of being employed. All three levels, primary, lower secondary and post compulsory are then combined using cost weights, with the latter based on expenditure per pupil from Eurostat. On average across countries and time, the aggregate quality adjusted growth adds about 20% to growth in cost weighted enrolments, but there is large variation across countries.
Combining with labour input from EU KLEMS labour productivity growth was estimated over the period 1999 to 2007 and then contrasted with the period 2007 to 2010 following the financial crisis. In many countries in the period from 1999 to 2007 hours worked growth was greater than output growth leading to negative labour productivity growth. In particular Greece, Ireland and Spain showed very significant growth in hours worked over this period and so strongly negative labour productivity growth. In addition labour productivity growth was negative in some of the larger countries such as Germany, France and the UK. A number of countries nevertheless showed positive labour productivity growth. These include the Czech Republic, Denmark, Finland, Italy, Hungary and Portugal. With the exception of Denmark, these countries had very low or even negative growth in hours worked. A few of the smaller countries, Latvia, Malta and Slovenia experienced strong reductions in labour input and consequently high labour productivity growth. In the period following the financial crisis many of the countries that showed poor performance up to 2007, e.g. Greece, Ireland and Spain, improved their relative position after the crisis. This was mostly due to reductions in hours worked rather than increases in output. Overall across the whole time period there appears to be some evidence of convergence of labour productivity growth over time; this was confirmed by a significant negative correlation between levels in 1999 and growth thereafter. However, there remains a large divergence in performance across countries, with some of the larger countries showing strong negative productivity growth rates. This might reflect failure to adjust completely for quality change but it is difficult to argue that these differences are purely a measurement mirage. The magnitude of the quality adjustments required to turn the growth rates positive go well beyond those seen in national accounts adjustments, e.g. through quality adjustments to deflators. Therefore the results suggest there is scope for productivity gains in the education sector in EU countries.
Research on non-market services carried out for workpackage 7 led to the paper ‘Towards Outcome-related Measures of Collective Services: an Assessment of the Protection provided by Policing in the United Kingdom’ by Martin Weale. This suggests one way of applying the principle of quality adjustment to policing output, is by making the assumption that the protection offered by the police (the outcome) depends on what there is to be protected as well as on the amount of policing offered. This approach suggested in the paper is that policing output should be measured with reference to i) the amount of human and physical capital protected and ii) the quality of that protection. This is in keeping with the principle that, while in current prices, outputs should be measured by inputs, nevertheless, volume measures should reflect changes in the quality of those output. The output of the protective services is protection, so that an indicator of the volume of what is protected offers an obvious measure of what is protected while a measure of the quality of that protection is provided by an indicator of the extent to which effective protection from undesirable outcomes is available. Illustrative calculations are performed for the United Kingdom and suggest that over the period 1999 to 2003 the national accounts grossly overstated the output of policing.
The final paper for workpackage 7 relating to non-market services on University research output, went beyond the remit and produced results for all EU countries, rather than just focusing on a few. This paper ‘The Research Output of European Universities, 1996‐2010’ by José Manuel Pastor, Lorenzo Serrano and Irene Zaera, analyses the difficulties related to the measurement of the research output of universities and proposes a simple overall indicator which incorporates quantitative (scientific publications) and qualitative aspects (citations) to permit the decomposition of the influence of the two factors. On the basis of this indicator homogeneous comparisons are made of the relative research output of the countries of the European Union and its evolution during the period 1996-2010. Those results show that the UK, Germany, France, Italy and the Netherlands are the European countries with the greatest research output in the period analysed. However, these countries show in general the most moderate growth, never exceeding 7.0% annually. Other countries, which started from lower levels, have registered more intensive growth, of approximately 14%, as in the case of Cyprus, Portugal, Romania, Lithuania or Ireland. Whatever the case, growth in the EU (6.5% annually) has been more intensive than the average worldwide (5.9%) and, above all than that of the US (2.7% annually). The greatest part of that progress in the EU is due to quantitative aspects, but relative quality is significant and contributes more than a quarter of the overall gain. All EU countries except Cyprus have experienced improvements relative to the worldwide average. By contrast, in the US a slight decline in terms of relative quality has occurred. The analysis indicates that Denmark, the Netherlands, Sweden and the UK are the leaders in research output quality, occupying positions between 30% and 42% above the worldwide average. In contrast, the countries of Eastern and Southern Europe generally perform below the average, although some of these countries are the ones that have progressed furthest during the period. These results are robust to adjustments to take account of difference in the fields of research across countries. The results indicate that that some countries are getting more value for the money allocated to University R&D than others.
The INDICSER database
The first main output of workpackage 1 was the extending the EU KLEMS database to incorporate revisions to the standard industrial classification from NACE rev 1 to NACE rev2 and include additional years so that analysis could be carried out for the period following the financial crisis. Originally, the idea was to update the EU KLEMS database in the old original NAC rev 1 and in the first half of the INDICSER project, we provided a detailed industry breakdown of the EUKLEMS from 31 to 72 industries (the March 2011 update). The methodology and data sources are described in the paper ‘EU KLEMS Growth and Productivity Accounts March 2011 update of the November 2009 Release: Description of Methodology and Country Notes’ by Reitze Gouma, Klaas de Vries and Astrid van der Veen-Moojij.
In the second half of the project, we originally thought to update the data from 2007 to 2009 using the same classification. However during the project lifetime it appeared that many national statistical institutes started to produce new time series classified using NACE rev 2, raising the feasibility of updating in the old classification. Based on research done on the National Accounts revision (as part of task one), developed in cooperation with the OECD, a conversion table and started to collect new data in NACE rev 2, in what is called the 2012 EU KLEMS release.
The methodology employed is described in a paper titled ‘National Accounts Developments and extending the EU KLEMS database’ by Reitze Gouma and Marcel Timmer. It describes a major change in the national accounts of Europe, namely the introduction of a new industrial classification which results in breaks in long-term time series of output and productivity of countries. The paper outlines how the project dealt with these changes in a new release of the EU KLEMS database in November 2012. The document consists of a general overview of the major changes and challenges in the implementation of ISIC rev 4 and NACE rev 2 in productivity databases. It then provides a description of the applied methodology for the 2012 EU KLEMS rolling updates, i.e. the general approach to updating EU KLEMS in the NACE rev 2 industry classification. Country specific notes which cover the country specific sources and methods applied in the update EU KLEMS files of November 2012 and detailed Appendix Tables are also included. The development of the database was overseen by RUG with contributions by BHAM, IVIE and DIW. All partners provided feedback on the new EU KLEMS files in order to improve quality. The RUG harmonised all received data, collected data for additional countries and produced the new EU KLEMS country files.
This new release is similar in concepts and methodologies to calculate the various growth and productivity variables as its predecessors, but it also has a number of new features;
• It provides updates to 2010 and revisions of longer time-series where national statistical institutes (NSIs) provided these.
• The data on output, value added and employment in EU KLEMS is now fully consistent with the series in the OECD Structural Analysis Database (STAN) at the corresponding industry levels.
• For labour composition use has been made of the micro-data underlying the European Labour Force Survey (LFS) for recent years (see discussion below).
• New investment data up to 2009 have been provided by the EU KLEMS consortium partners
• Most importantly, a new industrial classification is used.
Updated series are now available on the EU KLEMS web site (www.euklems.net) for the following countries (Austria, Belgium, France, Germany, Italy, the Netherlands, Spain and the United Kingdom). In this way, the EU KLEMS database has been transformed in a forward looking way, rather than backward looking and is now capable of incorporating the data in NACE rev 2 revisions that will become available in the near future, thus ensuring continuing relevance of the EU KLEMS database.
RUG, in cooperation with the other consortium partners, initiated an effort to increase the geographic coverage of productivity databases by initiating and supporting research efforts outside the EU. This has led to efforts in various countries that complement the work being done in INDICSER, namely China, India, Korea, Japan, Taiwan, Australia, Canada, US, Argentina, Brazil, Mexico and Chile. This increases the usefulness of the data for European research and policy analysis by providing international comparisons and benchmarking.
The second main data deliverable from WP1 was detailed labour accounts for EU countries, dividing employment and relative wages by gender, age group (aged <29, 30-49 and 50 plus) and skill (high, medium, low). The original EU KLEMS data on labour composition allowed each partner country to divide skill groups into categories that were best given their education system. While there are advantages to this approach, over time it became apparent that the demand by users was for data harmonised to the International Standard Classification of Education (ISCED). The research teams at BHAM and NIESR devoted some effort to investigating the ISCED levels that best capture significant differences in skill levels while maintaining a division into just three categories – the latter is dictated by sample sizes in the survey data which precludes any additional disaggregation. The examination of the types of qualifications and relative wages included in the ISCED groups suggested that ideally ISCED level 5, Tertiary education (first stage), should distinguish 5A which are long-cycle academic qualifications from 5B which are more vocational and shorter cycle programs. However the data required to divide ISCED level 5 were not available for all countries or years. Therefore the skill levels distinguished in the database were: High: University graduates or equivalents, postgraduates (ISCED 5 and 6); Medium: Upper secondary and post-secondary non-tertiary education (ISCED 3 and 4) and Low: Primary and Lower secondary (ISCED 1 and 2).
The original EU KLEMS database only had data on labour composition for a subset of countries. Using the microdata underlying the harmonised EU LFS, EU SES and EU SILC surveys, the project team produced common data by 1-digit industry group for 24 EU countries (excluding Malta, Bulgaria and Romania). These data present shares of employment and wage bills by type of worker from 2002-2008 classified according to NACE rev 1. In addition it presents employment shares for 1 digit industries for 2008 and 2009 classified according to NACE rev 2.
Workpackage 2 produced a number of data series on innovation, ICT and intangible assets. Details of methodology and sources are given in the paper ‘Indicators of innovation, intangibles and ICT use – Annotations’ by Marianne Saam, Daniel Cerquera, Mary O’Mahony, Fei Peng, Christian Rammer, Felix Roth and Miruna Sarbu. The industry coverage varies by indicator depending on data availability. In addition to the Innovation index referred to above, the website contains data on the following indicators from CIS: (2004, 2006, 16 countries).
• Share of enterprises with technological innovations (percent)
• Share of enterprises with non-technological innovations (percent)
• Share of enterprises with continuous R&D activities (percent)
• Share of innovation expenditure in turnover (percent) (percent)
• Share of sales with new products in total turnover (percent)
In addition to the aggregate ICT index described above, Indicators from the Eurostat ICT survey (16 countries) included in the database are:
A. General Indicators (Annual 2004-2009, Approx 1 digit NACE rev 1)
• Use of computers (% of enterprises)
• Use of internet (% of enterprises)
B. Use of software (Annual 2007-2009, Approx 1 digit NACE rev 1)
• Use of ERP software to share information on sales/purchases with other internal areas (% of enterprises)
• Use of CRM software to manage clients information (% of enterprises)
C. Use of e-Commerce (Annual 2004-2006, Approx 1 digit NACE rev 1)
• Total internet sales over the last calendar year, excluding VAT (% of turnover)
• Total internet sales over the last calendar year, B2B (% of turnover)
• Total internet sales over the last calendar year, B2C (% of turnover)
• Total internet sales over the last calendar year to other EU countries (% of turnover)
• Total internet sales over the last calendar year to own country (% of turnover)
• Total internet sales over the last calendar year to rest of the world (% of turnover)
The Intangibles database contains annual series by asset type from 1995 to 2007 for 14 countries (mainly larger EU15 plus the Czech Republic and Hungary) at the 1 digit NACE rev1 level for the following variables:
• Investment (Gross Fixed Capital Formation, Current Prices, Millions of National Currency)
• Capital stocks (Real Fixed Capital Stock, 1995 Prices, Millions of National Currency)
• Adjusted gross value added, nominal and volume
• Price Deflators
The paper ‘‘Intangible Investment at the Industry Level: Growth Accounting’ by Thomas Niebel, Mary O’Mahony and Marianne Saam details the methodology employed to construct these series. The asset types distinguished include:
• Scientific R&D
• Firm-Specific Human Capital
• New Product Development Costs in the Financial Industry
• New Architectural and Engineering Designs
• Market Research
• Advertising Expenditure
• Own Account Development of Organisational Structures
• Purchased Organisational Structures
These, plus software available from EU KLEMS, follow the asset types identified by Corrado, Hulten and Sichel (2005). Aggregate market economy nominal investments reported in the INTAN-Invest database are used as control totals. In addition some other variables are included as useful supplementary information. The proportion of the workforce who receive training underlies the training capital estimates and might be useful to academic and other users who want to focus on workforce training alone. Variables on fear and trust from the European working conditions surveys are also included.
Workpackage 3 produced indicators of Market Structure and Regulation. In terms of Market structure, two groups of data have been posted on the website, Competition indicators and productivity dispersions. These cover the market economy, excluding division O, ‘other community, social and personal service’ and for most countries the indicators are available at both 2 digit and 3 digit NACE rev 1 industries. These indicators are aggregates using the Amadeus company accounts database and in most cases are annual series from 2002 to 2009. Data are available for 19 countries (some small countries excluded due to small sample sizes).
The Competition Indicators comprise:
• AGE (Average age of firm)
• AGE* – age weighted by sales relative to unweighted ( a measure of the extent to which sales are concentrated in younger firms
• Adjusted Herfindahl Index (Sum of squared shares of sales, adjusted for sample size)
• Price cost margin
• Boone Indicator (profit elasticity with respect to variable costs)
The productivity dispersion indicators comprise:
• Value added per employee percentile (P) differences: P90/P10
• Value added per employee percentile (P) differences: P75/P25
The regulation Indicators are divided into product market and labour market. These comprise:
• Product markets (OECD indicators),
– Indicators for specific service sectors (energy, transport and communications).
– Regulation impact (knock-on) indicators, approx 2-digit NACE rev 1, 1975-2007 based on indicators for specific service sectors and input-output linkages.
• Labour markets
– The OECD Indicators of Employment Protection Legislation (Indicators of stringency of regulation for employees on regular contracts and employees on temporary contracts).
– Percentage of employees on temporary contracts.
– Measure of extent to which economy-wide EPL (for regular contracts) is binding at industry level (EPL interacted with measures of industry labour market structures)
• EPL1PT (EPL1*% of employees part-time)
• EPL1HW (EPL1*% of employees that work from home)
• EPL1SS (EPL1*% of employees that work on a Saturday or Sunday)
• EPL1UN (EPL1*%of employees that were unemployed one year earlier).
(proxy for labour turnover)).
All excel files contain information on definitions and construction of the indicators. Additional information is contained in the review papers ‘Review of Market Environment Indicators’ by Yasheng Maimaiti, Mary O'Mahony, Ana Rincon-Aznar, Stanley Siebert and , Justin Van de Ven and in ‘Notes on Labour Regulation – Its Measurement, Causes and Effects’ by Stanley Siebert.
Workpackage 4 contains a large number of Indicators related to the performance of the financial services sector. Details are contained in the review paper ‘Review of Financial Services Indicators’ by Robert Inklaar, Joaquin Maudos, Juan Fernandez de Guevara and Aljar Meesters. Many indicators cover all EU27 countries but some just refer to the larger countries, given data availability. The time period varies but many series run from 1990 to 2010/11. The construction of the indicators are based on data from a wide range of sources including the ECB, OECD, IMF and BIS, information from national statistical offices, and aggregations of data from Bankscope and Amadeus.
The indicators covered in this workpackage fall under seven general headings:
• Bank output
• Dependence on external finance
• Bank efficiency
• Bank competition
• Integration
• Internationalisation and openness
• Bank soundness
Bank Output contains indicators of bank output at current prices (euro area) and quantity indices of residential mortgages and deposit transactions (selected EU countries). Dependence on external finance is measured by the share of debt in total assets (EU27). Bank efficiency is measured both by cost efficiency, profit efficiency and cost to income ratios (all EU27). Bank competition includes a range of measures including the Lerner index, interest margin over reference rates and market concentration (CR5, Herfindahl) – all these indicators are available for EU27. Integration indicators comprise a range of measures covering money markets, debt markets, equity markets and banking markets with the country coverage varying by indicator. The database contains one indicator on internationalisation and one on openness based on market shares of the external sector – these are available for the euro area. Finally bank soundness is measured by a range of indicators including compensation for risk bearing, returns on assets and equity, non-performing loans, financial stability and risk exposure.
The data underlying the research in workpackage 6 are also included in the database. The indicators cover the period 1999-2010 and most EU countries and comprise:
• Enrolments
– From Eurostat (but with adjustments), EU27 – 1999-2010
• Output Indices
– Lower secondary, PISA adjusted
– Non-compulsory, earnings adjusted
• Labour Input
• Labour productivity
Indicators of health (workpackage 5) are not included in the database as many of the underlying data series employed were confidential and available only to the individual researchers in the countries concerned. The relevant measures are included in the papers referred to in the previous section.
Analytical Research
The project also produced a number of analytical research papers. In workpackage 1 the new measures of labour composition were employed in a paper ‘New Measures of workforce skills in the EU’ by Lili Kang, Mary O’Mahony and Fei Peng, published in the National Institute Economic Review. This analysis was designed to yield a more complete picture of the growth in average skill levels embedded in the EU workforce, comparing with competitor countries such as the US and China. The harmonised data from the EU surveys discussed in the database section were employed to extend coverage in existing databases to more countries, to cover the period of the financial crisis, and to skills acquired through informal workforce training. The results indicate growth in labour quality in the EU15 marginally below the US. There was evidence of convergence of the group of new member states to the EU15 but no sign of convergence of China to more developed regions. Extending to the period following the financial crisis demonstrated a pronounced rise in labour quality in most countries after 2007, consistent with theories of labour hoarding, and this appeared to be much stronger than in previous recessions. There were however some notable exceptions, namely Greece, Italy and Poland. The results are suggestive of greater hoarding of high skill labour in services than in production industries but a reverse finding for countries more focused on production such as Germany, the Czech Republic, Hungary and Italy. This paper also discusses the expansion of conventional measures of labour quality to include informal training (using the data arising from workpackage 2) and that this leads to small but significant increases in the growth of human capital in some EU15 member states. As training investments are considerable greater for high than for low skilled workers, firms’ prior investments in training raise the relative cost of firing the most highly skilled workers. As training has been growing rapidly in recent years, this opens the possibility of link between training and the extent of labour hoarding in the recent downturn. The descriptive evidence supports this - some of the countries that showed a deceleration in the growth in labour quality from 2007, notably Italy, Greece, Poland and Cyprus, have relatively small training investment to GDP ratios, consistent with the idea that the costs of firing workers with higher skills may be lower in these countries. However more work is needed before we can be confident that such a link exists.
In workpackage 2 both the papers on training, Human Capital Formation and Continuous Training: Evidence for the EU Countries’ by Mary O’Mahony and intangible investments, ‘Intangible Investment at the Industry Level: Growth Accounting’ by Thomas Niebel, Mary O’Mahony and Marianne Saam, produced growth accounting results that inform the literature on drivers of productivity growth. The findings from the training paper were discussed in previous sections but it is also important to highlight the finding that expenditures on training and general schooling appear to be complementary, not substitutes. This has the implication that higher spending by firms on training does not absolve governments from spending on general education. The growth accounting estimates by industry using the data on all types of intangible assets suggest that since intangible capital assets by type varies across sectors with R&D most important in manufacturing whereas organisational capital dominates in many service sectors. In terms of contributions to labour productivity growth, however, in general there appear to be common sectoral patterns across countries, with high investment in all sectors in some countries (the UK and the Netherlands) and low investment in others (Italy and Spain). This in turn suggests that constraints on intangible investments might be coming from the market environments and regulation rather than due to variations in industry structures across countries.
In ‘Human Capital Spillovers: The Importance of Training’ by Mary O’Mahony and Rebecca Riley, the authors suggest a role for intangibles in enhancing knowledge transfer and spillovers to wages. In particular employer provided training may increase the relevance of knowledge exchange to the production process and this is likely to be linked with the use of ICT. This paper analyses the extent of knowledge spillovers from tertiary education within broad sectors and explores the importance of training in determining the extent of these knowledge spillovers. The paper makes use of the European Community Household Panel (ECHP), where individuals can be tracked across time, linked to training stocks developed for INDICSER and ICT capital from EU KLEMS. The longitudinal nature of the ECHP allows the analysis to address a number of econometric issues that are often ignored in this literature. The analysis is carried out for four EU countries, France, Germany, Spain and the UK where the relevant data are available. The results are consistent with the presence of significant spillovers from tertiary education at the sector level. In particular they suggest that a one percentage point increase in the sector share of tertiary educated workers/hours raises individuals’ wages by approximately 0.8%. This also implies that individuals do not internalise the full benefits of their human capital investments. The question of what might generate such spillovers from education is very much under-explored in the literature. Using data on employers' investments in training the paper finds that these are positively associated with the extent of spillovers from tertiary education; in some specifications ICT is also positively associated with knowledge spillovers. The difference in the magnitude of sectoral spillovers from education between sectors/countries with the highest training intensity compared to those with the lowest training intensity is in the order of 1 percentage point. Education spillovers of this magnitude are potentially very significant in terms of explaining observed wage (and by association productivity) differences across sectors and countries.
The impact of intangible assets is also investigated in the paper ‘Intangible Capital, Risk and Interindustry Differences in Rates of Return – Evidence from Germany’ by Bernd Görzig, Martin Gornig and Axel Werwatz which analyses the persistence and the determinants of rate of return differences among sectors of the German economy. The analysis proceeds in two stages. In the first stage, it investigates the persistence of the inter-sectoral rates of return differences during a period of nearly 40 years (1970-2007). Using data for Germany from the EU-KLEMS database, the analysis of both β-convergence and σ-convergence shows that there is some convergence but that relative positions are nonetheless fairly stable and persistent. The variation of return rates across sectors thus has some transitory and permanent components. The second stage of the analysis attempts to explain this variation by using a rich data set based on establishment-level information from the EUKLEED project, available for the 1999 to 2003 period. This compares the explanatory power of both more established factors of intersectoral rate of return differences such as the sectoral intensities of tangible and human capital, openness and scale but also more novel explanatory variables such as sectoral risk (based on variations in profits) and the intensity with which sectors employ intangible capital. The paper finds that the more traditional explanatory variables have less explanatory power than their more recent counterparts. In particular, sectoral risk is found to have a strong and robust positive effect on the rate of return to capital of a sector. These results are driven by the service industries who tend to have both very high values of the risk indicators as well as above-average measured returns. An effect of differences across sectors regarding the intangible capital intensity only comes to the fore if all other factors are properly accounted for, particularly risk. This suggests that differences in intangible capital are most useful for explaining return differences among different service sectors or among different manufacturing sectors, respectively, rather than between the two broad sectors.
The impact of ICT on economic performance is addressed in the paper ‘The Impact of ICT on the Volatility of Economic Performance: A Sectoral Analysis for the EU’ by Daniel Cerquera. This paper analyzes whether existing inter-sectoral relationships (described by input-output flows) explains the impact of ICT on volatility in economic output at the sectoral level. The paper presents a theoretical motivation based on the aggregate impact of misallocation in the production of intermediate inputs, showing that there is a channel through which ICT use at the sectoral level can affect the volatility of aggregate economic output. The analysis covers a subset of European countries between 1995 and 2006. The results of the paper show that the impact of ICT use on output volatility at the sectoral level in Europe is positive and statistically significant for sectors with stronger (commercial) relationships with other sectors. This effect appears to be concentrated in the service sectors. If the impact of ICT on output volatility is transmitted through the inter-sectoral linkages, then this impact should be expected to be stronger for economic sectors that are highly (commercially) connected with other sectors. This is precisely the case for the service sectors where, as shown in the descriptive part of the paper, these sectors demonstrate higher multipliers because all sectors in an economy require their output as intermediate inputs. This result does not suggest the role of inter-sectoral linkages is not present in the manufacturing sectors. But given the data available and the methodology adopted the leading mechanism is stronger for the service sectors.
As part of the research in workpackage 3 the paper Employment protection, productivity, wages and jobs in Europe by Ana Rincon-Aznar and Stanley Siebert use industry data for 11 European countries and the US, and 19 industries over the period 1985-2007 to test for productivity, wages and employment effects of employment protection legislation (EPL). While recent empirical work highlight the negative productivity effects of EPL, one of the main contributions from our research is that EPL’s productivity, wage and employment effects can be interpreted within a conventional labour supply and demand framework. The project therefore provides the basis for an evaluation of this important policy. In accordance with conventional competitive theory, the authors find that EPL both lowers productivity and produces a compensating wage differential that lowers wages. The fall in TFP growth is considerable, about 0.5% annually, which can be compared to overall annual average growth of 1.5%. At the same time, we also find that EPL is associated with reduced real wage growth (about 0.25% annually, relative to a sample average annual increase of 2.4%), which we interpret as reflecting the compensating wage differential associated with job security. Whether the reduction in wage growth is sufficient to compensate for reduced productivity can be simply judged in our framework from employment growth, which is not reduced significantly in industries affected by EPL. Thus, the benefit of job security provided by EPL appears broadly to be paid for by the workers themselves. Manufacturing appears more constrained by EPL than services as EPL reduces TFP more in manufacturing than services. This result is according to expectations since manufacturing has naturally lower labour turnover rates than services, and hence cannot rely as much on “natural wastage” to accommodate EPL. To some extent then, EPL may represent a problem for manufacturing. However, wages also fall more in manufacturing than services, leaving employment again not much affected, so the challenge to manufacturing might be more apparent than real. The downside seems to be an increase in hours worked per worker, which implies that EPL concentrates employment on insiders at the expenses of outsiders such as young workers.
A country case study arising from workpackage 3 is ‘Competition on business services markets in Hungary’ by Éva Palócz and Peter Vakhal. This finds that foreign owned companies play a significant role on the Hungarian business service market. In a growing number of cases, foreign companies have settled in Hungary in order to make use of their firm specific knowledge combined with lower wage costs on the host market, in order to carry out exports from the host country, similarly to manufacturing firms. Thus, unlike the traditional theory of services FDI, they have been created not only for serving the host markets. The export intensity of foreign business firms in Hungary is rather high in all subsectors: in some (NACE two digit) branches the share of exports in total sales reached 70-90 percent. Beside large subsidiaries, small foreign owned service units are also strongly export-oriented. Mixed ownership brings also significant benefits for both partners: foreign partners contribute usually with technological investment, with marketing and sales of services produced by the domestic partner. Market concentration in business services, measured by net turnover, is typically not high (on average it is similar to those in manufacturing) but rather volatile. The appearance or disappearance of a large (typically, multinational) firm can basically change the concentration index. The concentration, measured by exports, is significantly higher. Since the concentration index is largely influenced by exporting foreign companies, their activity doesn’t disturb the domestic market. Analyses of survival functions showed that the expected lifetime in the manufacturing industry is significantly higher than in the business services sector. More than 30% of manufacturing companies founded in 2000 have been operating in 2010; this rate is only 23% in case of business services. Statistical calculations show that the first three years after foundation are the most risky in both sectors. In the case of business services the first year is crucial: the hazard rate of closing down the business is almost 50%. After the third year the risks are increasing only moderately and stabilize after about seven years. By analyzing the survival risks with the help of possible influencing factors the authors found that the in case of companies belonging to the manufacturing industry the average hazard rate is 16 percentile less than in case of business service companies. Other influencing factors were the company size (larger companies have lower risks) and export revenues.
A number of analytical papers resulted from the research carried out for workpackage 4 on financial services. In ‘The Impact of the Financial Crisis on Financial Integration, Growth and Investment’ by Robert Inklaar, Juan Fernández de Guevara and Joaquín Maudos, published in the National Institute Economic Review, the authors analyse the impact of financial development and European financial integration on growth. First, they quantify the effect of financial development on growth in Europe, isolating the effect of European financial integration based on a counterfactual analysis. Results show that both financial development and financial integration have been important in driving the growth in European economies. From the introduction of the euro and the Financial Services Action Plan in 1999 to 2007, increases in financial development have contributed 0.18 percentage points to economic growth of 2.1 per cent. Financial integration in turn was responsible for 0.03 percentage points of this total effect. Since the beginning of the crisis in 2007, the contribution of financial development and integration has gone down, with an annual contribution of 0.06 percentage points for overall financial development, of which 0.02 percentage points is due to financial integration. These results highlight the costs involved if financial integration were to be reversed. The financial crisis saw calls for increased lending to domestic companies at the expense of lending abroad. In addition there were arguments that greater financial stability could be a consequence of less financial integration, as losses in one country would lead to less contagion in others. Regardless of the validity of these arguments, it is important to realise that European economies would pay a price in terms of lower economic growth for policies that would reverse financial integration.
The second part of the paper analysed how investment responds to a financial crisis. The authors find that, following a financial crisis, investment declines more in countries with a greater degree of uncertainty aversion, which can be informative for evaluating post-crisis economic performance. While advocating caution in drawing inferences from a broad cross-country analysis for individual countries, the authors suggests that these results could help explain why Ireland, a country with a low degree of uncertainty aversion, has recently been showing more promising signs than Portugal or Greece, both countries with a very high degree of uncertainty aversion. More important from a policy perspective is that the degree of uncertainty aversion, like other aspects of a country’s culture, changes slowly, with measures based on surveys across different decades showing a broadly similar ranking of countries. In other words, the degree of uncertainty aversion should be taken as a ‘fact of life’ and policy adjusted accordingly. It might be better to provide a stable perspective on rescue funds rather than many short-term benchmarks if the goal is to revive autonomous growth in countries with high degrees of risk aversion.
The paper ‘Who’s afraid of big bad banks? Bank competition, SME, and industry growth’ by Robert Inklaar, M. Koetter and F. Noth tests how bank market power influences technical change and resource allocation of informationally opaque small and medium sized firms. This uses a dataset with approximately 700,000 firm-year observations of German SMEs. It identifies the effect of bank market power using the dependence on external finance by industry and the regional demarcation of the German banking market. The authors combine comprehensive SME data with prudential regulatory bank data on market power. The novel SME sample allows estimation of three different growth components: input growth, technical change, and a term that captures gains from the reallocation of production factors from unproductive to more productive SMEs. The results indicate that bank market power generally spurs aggregate SME growth, which indicates that banks need to realize sufficient margins to generate useful private information. A 1%-increase of average bank Lerner indices per region increases aggregate SME output growth by 0.12% at the median level of industry dependence on external finance. Aggregate output growth is primarily due to technical change, but the reallocation of resources from low-productivity SME to high-productivity SME is also of economic significance. However market power reduces SME growth in industries that depend heavily on external finance. The results could thus be indicative of nonlinearity in the market power-growth relationship: if market power is too high, rent extraction and lock-in effects have their greatest impact. But if market power is too low, not enough useful information is developed to identify the most productive firms and investment projects. Even small banks may extract rents from locked-in firms that depend heavily on external finance, which may entail negative growth. Hence, regional market conditions should matter for antitrust policies rather than considerations of bank size alone.
The paper ‘Does banking competition affect Europe's growth’ by Rients Galema, Michael Koetter, Aljar Meesters and Robert Inklaar also investigates the implications of European banking market competition on the reallocation of productive factors across firm and thereby aggregate growth. Here the authors use a sample of large firms from the Amadeus company accounts database to estimate production functions and associated growth components and bank-specific data from Bankscope, to estimate average market power in EU banking markets with Lerner indices. While the empirical set-up is chosen to be as comparable as possible to the paper by Inklaar et al. from above, the analysis shows no significant effect of bank market power on growth of (non-financial) firms. This could be related to data quality. For instance, the Bankscope data do not allow for a complete coverage of national banking markets, unlike the German study and the Amadeus data cover many larger firms that are typically less reliant on bank finance than the SMEs in the German study. Alternatively, bank market power dynamics could be too modest, while cross-country institutional differences could be too large to find a significant effect.
Taken together, these studies thus do not present a fully consistent set of results on the broader economic impact of bank market power. But what can be concluded is that there is no evidence that bank market power (at the levels observed across Europe) has hampered growth in the rest of the economy, and there may have been some benefits for European SMEs. This suggests that if efforts to bail-out and prop-up European banks since 2007 have led to increased market power, this is not a cause of great concern for the growth performance of the broader economy. The project also led to two methodological papers on the method of frontier analysis by Aljar Meesters, ‘A Note on the Assumed Distributions in Stochastic Frontier Models’ and ‘A Non-Linear Stochastic Frontier Model’ with an application to the banking sector.
The paper ‘Indicator based analysis of the Hungarian Banking Sector’ by Katalin Mérő employs the indicators constructed for the project to review performance in a country that has gone through major restructuring of its financial services sector. This considers the size of the banking sector and depth of financial intermediation, firms’ dependence on external finance, bank profitability and efficiency, bank competition, integration, internationalization and openness of the banking market and soundness of the banking system. The main conclusions are as follows. The first set of indicators of financial depth and the size of the banking sector reflects underdevelopment of the Hungarian banking system compared to countries with similar levels of per capita GDP. Accordingly, significant financial deepening is necessary but is attainable at present with global financial markets are characterized by general deleveraging and when the sovereign risk of Hungary is amongst the highest in international comparisons. The low levels of loan-to-GDP ratios are coupled with a high level of household indebtedness implying little scope to increase lending to households. The Hungarian firms dependence on external finance is higher in the case of SMSs and less for large corporations. Until the mid 2000s the Hungarian banking sector had a superior profitability compared to the Euro zone’s banks as a consequence of high interest margins which attracted foreign funds and led to an extremely high loan-to-deposit and foreign funding ratios. The profitability shift started in the mid 2000’s, when the Euro zone banks’ profit ratios increased to exceed that of the Hungarian banks and this situation worsened following the financial crisis so that by 2011 the banking sector as a whole was making a loss. At the same time the cost efficiency of the banking sector improved significantly. The Hungarian banking market has a medium level of concentration but differs by sector with concentration very high in the retail market, and competition in the retail market was restricted by the lack of proper credit information systems. The Hungarian banking market is an open market - direct foreign ownership exceeds 80 %, the proportion of foreign liabilities is almost 30 % while the proportion of foreign assets is over 10 %. The Hungarian national champion, the OTP has an East-European international subsidiary network. Since the openness is coupled by a low level of competition on the retail market, the Law of One Price does not prevail. During the crisis the Hungarian banking system became highly exposed to portfolio deterioration and increased provisioning costs. Nevertheless the Hungarian banking sector is stable, as banks have a strong capital position, and they have capital buffers well above the regulatory requirements to cover any potential losses. Finally due to the high loan-to-deposit ratio, and consequently, the system’s large exposure to foreign finance, liquidity is especially important as regards the soundness of the banking system.
The health workpackage produced one review and one analytical paper. In ‘Innovation strategy in the health sector and some recommendations for the European Union’ by Guldem Okem, the author discusses recommendations for the realisation and enhancement of innovation environment in the health sector in the EU in view of the specific features of their innovation structures and processes. It reviews both the current state of innovation activities and drivers of this, including R&D, patents, funding and government incentives, public-private partnerships and market access. This paper defines the healthcare sector as comprising health service providers, pharmaceuticals, medical devices and technology, and medical biotechnology and these sectors are interrelated, knowledge-based and technology-intensive. Innovations have to take place in a market that is riddled with market imperfections, uncertain demand and asymmetry of information between the providers (e.g. doctors) and users (patients). Public funding of R&D in healthcare is lower in general in Europe than in the US, as are clinical trials per inhabitant, number of R&D researchers, and health patents. Europe performs relatively well in pharmaceuticals and medical devices but is now facing increasing competition from emerging economies including Brazil, China and India where economic and research activities are migrating outside of Europe. The innovation stages are spreading on a global scale and mostly performed by young companies but there is absence of well-developed venture capital in Europe in addition to low number of private investors. Price controls and reimbursement arrangements, licensing arrangements and other administrative burdens can hamper innovation. The paper puts forwards recommendations on how to improve innovation in the healthcare sector in Europe including more effective coordination of EU policy objectives across the various levels of EU institutions and more effective coordination between member states. There is also a need for an institutional structure to coordinate between R&D activities and funds. Correcting market imperfections should also be a priority. Knowledge about markets, access to funding opportunities and skilled labour should be easier. The public sector regulates the market and acts as the principal buyer of health services on behalf of citizens. The pricing and reimbursement processes and licensing arrangements can be utilised to support new medicines and health technologies with priority granted to innovative products. The conservative reimbursement criteria induced by the funding shortages should not effectively constrain the utilisation of innovative products and processes, and hence suppress innovation. Specific skills and manpower that is needed for health sector innovation should be planned and supported in meeting the strategic needs of the market including removing barriers to mobility of skilled workforce.
The paper ‘Relative Equality of Opportunity: A Statistical Approach’ by Carmen Herrero, Ildefonso Mendez and Antonio Villar cuts across the research for both workpackages 5 and 6. This paper analyses Equality of Opportunity when agents’ achievements are expressed in terms of ordered categorical data. The paper provides a cardinal evaluation function that exploits the relative frequencies of the achievements of the different groups. The proposed strategy of analysis is based on the following principles. First, agents are classified into groups that gather people with similar external circumstances. Second, the outcome distribution of the agents of the different groups is regarded as an expression of their opportunities. Third the relative advantage of the different groups in terms of their corresponding outcome distributions is compared. This method is applied in this paper to the specific context of equality of opportunity and illustrated by means of two different applications in European countries. The first one refers to equality of opportunity in education and the second one focuses on equality of opportunity in health. The first application links equality of opportunity to the independence of the students achievements on their parents education. Data on the education level of parents and their children is taken from the Survey of Health, Ageing and Retirement in Europe (SHARE), a cross-national survey with a panel design representing the population of individuals aged over 50 years in some European countries. The result suggest that the difference in the relative education opportunities of children coming from differently educated parents is highest in Southern countries like Greece, Italy and Spain, and lowest in Northern countries like Denmark and Sweden. Spain is the only country in which the relative disadvantage in education opportunities of the children coming from the less educated fathers significantly improved when moving from older to younger cohorts of parents. The paper also finds evidence of convergence across European countries in the relative opportunities of children coming from less educated parents. The health application employs data from EU SILC and finds that the distribution of health opportunities across European countries has remained quite stable across cohorts. It also finds that Southern Europeans enjoy worse relative health opportunities than most of their European counterparts.
The paper ‘Indicators of educational inequalities’ by Márton Medgyesi also addresses issues of equality of opportunity in education. This paper reviewed three different perspectives on educational inequality and the main indicators that were proposed by the social sciences literature to measure these. The first perspective was inequality of educational outcomes. This literature proposes to calculate inequality indices like the Gini coefficient either on aggregate data on educational attainment of the population or individual level data on educational attainment. The second perspective was that of inequality in educational opportunities, that is, differences in educational outcomes related to the parental background. The third perspective is related to the distribution of school resources between children from different social background. The aim of these studies is to say whether poor/minority children typically attend schools with inferior quantity or quality of school inputs. In each case methodological properties of frequently used indicators were first discussed. Then the paper presented comparative evidence on educational inequality indicators for EU countries, based on aggregate data from the Barro-Lee dataset and survey data from European Social Survey and Adult Education Survey. The indicators of inequalities in educational outcomes and inequality of opportunity suggested the pattern emerging among groups of EU countries. The Southern European countries are characterized by high degree of inequality in educational attainment and also a high degree of inequality of opportunity in education. The Nordic countries have middle level inequality in educational attainment but seem to be more mobile than other European countries. The Western European countries occupy middle of the country ranking both in terms of inequality of educational attainment and also in terms of inequality of opportunity. The EU 12 countries are characterised by a low level of inequality in educational attainment. In terms of mobility there are differences within this group: the Baltic states are more mobile, while Central Eastern European countries exhibit a higher degree of inequality of opportunity in education.
Finally in workpackage 6 the paper ‘Does Early Education Influence Key Stage 1 Attainment? Evidence for England from the Millennium Cohort study’ by Anitha George, Lucy Stokes and David Wilkinson examines whether, in an era of near universal provision in the UK, early education is still associated with detectable improvements in outcomes for children. Earlier research suggested that early education improves cognitive and social development for children while they are still attending whereas the longer-term impacts depend on the quality of early education, but these results referred to a time when early years education was not so prevalent. The analysis focuses on attainment assessments when children were aged seven and finds that the overall impact of early education on attainment is modest, but that the impact is generally greater for those children who experienced poverty when they entered early education. These findings for children who experienced early poverty may be a reflection of the policy over the period under consideration of targeting high quality early education for children from disadvantaged backgrounds. A tentative conclusion from this research is that universal provision, with a focus on high quality services for disadvantaged children, can improve attainment levels for these children. Although the authors do not find large impacts for the population as a whole they point out that early education can also have wider benefits, for example, allowing parents able to find work to do so whilst their child is at an early education setting. Early education may also lead to longer-term benefits, so a full assessment of the impact of early education needs to consider a much wider array of outcomes than considered in this paper.
Potential Impact:
Impact
The main impacts of this type of infrastructure projects are the following:
- Use of the database by academics and the policy community
- Insights into measurement and conceptual issues and data deficiencies that inform national and international statistical agencies, and academics and the policy community who use the data
- Analytical results that inform policy.
Our expectation based on previous similar undertakings is that the take-up of the data will be significant, especially as many of the variables can be linked to the EU KLEMS database and the WIOD database. The project team have already used these combined datasets in the analytical research as reported above. In addition the sector specific data (financial services and education) should be useful to the research and policy community working in these areas.
The research team have publicised the data in the various academic and policy conferences listed below. As regards the EC community, the main features of the database were presented at the SERVICEGAP and INDICSER final conference in Brussels in January 2013 and previously at workshops organised by DG Enterprise and Industry and DG Research. Links to the database will be posted on the EU KLEMS, World KLEMS and WIOD web-sites.
The research team were active throughout the project in highlighting features of the database at various international conferences. In particular the World KLEMS consortium organised two major conferences in Boston in 2010 and 2012 where this work has been brought together and a platform for future work and cooperation established, see www.worldklems.net. This website provides information about the conferences, papers presented and datasets created.
Significant advancements were made in measuring intangible capital. Following the presentation of her work on training capital by Mary O’Mahony at an OECD conference on intangible assets in Paris in March 2012 (it became apparent that there were a number of groups working on developing industry estimates of intangible assets. Therefore a workshop was organised by Mary O’Mahony and Marianne Saam with Mariagrazia Squiciarrini and her research unit at OECD on constructing intangible capital, in particular organizational capital, at the sectoral level. Cecilia Jona-Lasinio from the INNODRIVE and the INTAN-Invest team also participated in the meeting, as did Matilde Mas from IVIE. The project scientific officer, Marianne Paasi also attended. This meeting was followed by a second event in Valencia in November 2012. The upshot is a willingness on the part of these teams to collaborate and ensure consistency in developing industry estimates of intangible capital. This is an important step forward as users are often unhappy when faced with different numbers for what should be the same variable.
The use of the financial services database was highlighted in presentations at various meetings and academic conferences organised by the team working in this area. Furthermore, the work on measuring bank output has led to the formation of an international taskforce of statistical agencies to re-examine and improve the methods for measuring bank output, or, more specifically, Financial Intermediation Services Indirectly Measured (FISIM). The work on the performance of the health sector will be presented at an NBER organised conference in Washington in October 2013.
The research team see the database as a dynamic resource which will be extended and updated on a rolling basis beyond the life of the project. Recent data deliveries from Eurostat will allow the updating of the labour composition estimates to 2011. This will also facilitate updating of the training estimates. In addition tabulations obtained from Eurostat will allow estimation of the skill component of the labour composition estimates at the 2 digit industry level. The EU KLEMS consortium is committed to continuing updating of that database for selected countries. The market environment competition and productivity dispersion indicators which are currently on the website are classified according to NACE rev1. Given that the EU KLEMS database is now moving to NACE rev 2, these indicators have been reclassified to NACE rev 2 and will shortly appear on the website. A revised version of the product market regulation knock on indicators, using a different methodology devised by partners in the SERVICEGAP project, will also be posted on the INDICSER website.
An important by-product of these infrastructure projects is that the process of constructing indicators throws up conceptual and measurement issues that inform research in these areas. Equally important is that the attempt to match data to the theoretical measurement frameworks highlight deficiencies in the available information and can produce recommendations for future data gathering exercises.
In the summary report on ‘Output measurement in services’ prepared for workpackage 8, Marcel Timmer highlights the uneven progress being made in Europe in services producer price indices (SPPI). While the leading European countries, such as the UK, provide SPPIs for over 20 services industries in 2011, some countries such as Italy lag far behind. Another major challenge highlighted is that coverage is no longer expanding by much and the report suggests that austerity is severely hampering the improvement of services sector statistics. While it is understandable that the financial crisis has led to a reappraisal of statistical priorities, the author concludes that lack of priority in this area means that Europe will continue to lag behind the United States in the quality of measurement in services.
The work on the financial services industry highlights many issues in current practice that inhibit international comparisons of output and productivity growth and this is emphasized again in the workpackage 8 report. This suggests that the framework of the System of National Accounts (SNA) has never been a natural fit for banks and insurance companies. While the SNA is well-suited for measuring the value and price of goods and services that are provided and paid for within a single period, many banking and insurance activities are intimately connected to financial returns over time. Also since the price is an implicit one and because the services provided are hard to make explicit, it is not readily feasible to survey producer prices as in other industries. Although alternative approaches are widely used in measuring nominal and real output in financial services, there is no academic consensus on the right approach. An important finding of the project is that different approaches lead to very different outcomes. This suggests that much better data and more debate is required to move forward measurement for this important sector.
The impact of the work on non-market services is likely to inform statistics producers on the feasibility of measurement in an area where progress is very slow. One important impact of the work on health is that discussions between the Hungarian research team and the Hungarian Central Statistics office has provided an opportunity to revise the way in which the provision of health services is accounted in the Hungarian national accounts. During the project period regular consultations were held with the staff of the Hungarian Statistical Office and the experts from the NHIF administration were also involved. It was agreed that instead of the presently used fairly simplified method the volume index of inpatient care provision will be estimated as recommended in the project, using the weighted composite volume index of activities adjusted with the hospital survival rate. The revision and the new results will be published in the Hungarian Statistical Review.
The attempt to compare performance of the health sector across countries highlighted some important data deficiencies in all countries studied and illustrates problems that need to be addressed in constructing volume measures of output for that sector. The conclusion of the workpackage 8 report on non market services is that the attempts to measure and compare output and productivity growth in the health sector highlighted that the data required are intrinsically linked to the system of provision of the service. This implies that debate on measurement needs to involve health experts who understand the nature of the patient journeys through the system. In addition undertaking international comparisons requires researchers who are familiar with the systems of provision and most important have access to data. Many countries will not allow access to researchers outside their borders. One important data deficiency, which is somewhat more difficult to understand, is the lack of information on health expenditures by age in most European countries. This is crucial for estimating value for money indicators.
For collective services the project work put forward a method for measuring volume indicators for protective services. The workpackage 8 report discusses how the measure employed for policing could be extended to other protective services such as fire and the data required. These approaches can be viewed as applications of the basic principle of volume measurement- that volume indices should reflect changes in quality. Equally, as in most other areas of national accounting, measurement of quality is imprecise and this should not be seen as a reason to ignore the issue.
The results from the analytical papers can be used to inform policy and many highlight important policy relevant issues as discussed above. The review paper on innovation in healthcare services highlighted some important areas for potential improvement as detailed above. The work on educational attainment suggests that it would be preferable to supplement the Europe 2020 tertiary attainment target for 30-34 year-olds with an enrolment target for 20-24 year-olds, since the sole focus on the 30-34 year olds leave Governments with little time to pursue policies affecting the ability to reach the target. In addition it highlights that the gender gap is widening and member states could implement policies that ensure the attainment rates of men rise to match those of women. The research on training highlights that the better educated receive more training and that some groups such as older, low skilled workers appear to lose out. Policy might examine the incentives faced by firms to train these groups.
The impact of some innovative ways of comparing performance (based upon the paper by Herrero et al. on equality of opportunity) is worth mentioning. The capability of this technique of dealing with categorical data can be exploited in many services, especially in health and education, where data frequently are available as categories. The ability of dealing with these types of data without any need of cardinalisation provides robustness to the analysis performed this way. Once such comparisons are undertaken, it is straightforward to come up with policy recommendations, by simulating ways of improvement in different countries.
Dissemination
Dissemination activities undertaken in INDICER included the following
Project Website:
A project website was created at the beginning of the project. The aim of the website was to provide information on both the researchers and the research produced by the INDICSER project. The website address is www.indicser.com.
Review papers:
These were mostly produced in the early stages of the project and were designed to review available data and highlight particular measurement issues. These are posted in the ‘Review Papers’ section of the website. The paper topics included innovation, ICT and intangibles; market environment; financial services; health services and education services. Recent additions include a paper reviewing innovation in healthcare services and one on developing a composite indicator of innovation.
Discussion papers:
The website contains over 30 discussion papers covering conceptual and measurement issues and the analytical papers. Many of these papers subsequently were published in academic journals. These are posted in the ‘Discussion Papers’ section of the website.
Policy briefs:
These covered market services and the EU productivity gap with the US, workforce training in the EU, education attainment and Europe 2020 targets, the impacts of employment protection legislation and the performance of the health sector in four countries. These are posted in the ‘Policy Briefs’ section of the website.
Database:
The project developed a number of data series as outlined above. These have been posted on the web-site for public use. All data are in easily downloadable excel format, with detailed descriptions of variables and sources in both the excel files and where necessary in separate explanatory documents. The datasets and corresponding documentation are posted on the ‘Data’ section of the website.
Project meetings: The research team presented the results of the projects at four INDICSER conferences held in London (February 2010), Valencia (April 2011), Budapest (September 2012) and Brussels (January 2013 - the latter joint with the SERVICEGAP project). These conferences covered the broad range of project topics and included presentations by the project research team plus invited external speakers and members of the project advisory board. These conferences were also attended by academics and policymakers. The main features of the database were presented at the Brussels policy conference. A number of workpackage specific workshops were also organised on health (Valencia, June 2010, and London, June 2012), Innovation and intangibles (Mannheim, June 2010), Market Environment (London, November 2010), Financial Services (Valencia, November 2010) and Education (Brussels, December 2010, and London, March 2012). The results of the first stage of the project were disseminated at a workshop at CEPS in Brussels in March 2011 and at a workshop organised by DG Enterprise and Industry in Brussels in October 2011.
External dissemination: The project team also presented their research at a number of academic and policy conferences, workshops and seminars. These included presentations on the research on extending EU KLEMS and intangible assets at the biennial conferences of the International Association for Research on Income and Wealth, in Switzerland, August 2010 and Boston, August 2012, and at the two World KLEMS conferences in Boston (2010, 2012). The research on training was disseminated at a number of academic conferences including the Royal Economic Society, University of Surrey, April 2010; The Scottish Economics Association, Perth, April 2010, and the OECD conference on Intangible assets, Paris, March 2012. The findings from the research on financial services was presented at the SERVICEGAP kick-off meeting, Birmingham, May 2010, at a conference organized by the Bank of Luxembourg, the SERVICEGAP mid-term meeting, Dublin, June 2011, The Spanish Economic Association (SAEe), Madrid, December 2010 and the XVII Finance Forum organized by the Spanish Finance Association in Elche (Spain), November 2010. Research on the health sector was presented at a conference in Valencia organized by the BVAA foundation (April, 2010), at the European Conference on Health Economics, Zurich, July 2012 and at the Spanish Economic Association Vigo, December 2012. Research on education was presented at Early Childhood Environmental Ratings Scales international workshops, London, (May 2011, May 2012) and the Early Years to Early Careers Workshop, London, February 2013. Other presentations include at the ZEW ICT conference, July 2012; the North American Productivity Workshop, Houston, Texas, June, 2012; European Workshop on Efficiency and Productivity Analysis XII, Verona, Italy; LLAKES conference, London, October 2012; University of Murcia, February 2012, University of Navarra, October 2012; Workshop on Productivity in the Service Sector, Challenges of Measurement, ZEW, Mannheim, November 2012; e-Frame Conference on Measuring Well-Being and Fostering the Progress of Societies, OECD Paris, June 2012; a workshop organised by College of Business of the University of Bangor (Wales), June, 2011; the XXI Jornadas de la Asociación de Economía de la Educación (AEDE), Oporto, July 2012.
List of Websites:
www.indicser.com