VoxEU Column Industrial organisation

Firm-level data: Who said that they are too difficult to use for policy?

Policies aimed at enhancing firm productivity may greatly benefit from firm-level evidence. Unfortunately, micro-founded data, particularly of cross-country nature, remain largely unavailable. This column presents a new firm-level database built by a research network of the EU system of central banks (CompNet). This data base allows investigating how firm size and labour costs interact at different levels of productivity. This new cross-country data base, and its potential to expand, could be of great policy value.

Conceptual underpinnings

Recent policy debate has increasingly used a more over-compassing definition of competitiveness, with a special focus on its drivers at the firm level and its impact on growth and productivity (Altomonte et al. 2012). The theoretical underpinning of this broader view relates, for instance, to the framework provided by Melitz (2003), which assumes a distribution of firms’ productivity not symmetric around the mean. In turn, this supports:

  • An emphasis on resource reallocation towards most productive firms as a mean to increase aggregate productivity; as well as
  • The need to move away from merely considering average performance – which can be misleading – towards analysing rather the underlying productivity distribution.

Empirical literature for both the US (Bernard et al. 2011) and a number of EU countries (Mayer and Ottaviano 2007) has confirmed that, in general, firm-level productivity is typically distributed following a Pareto distribution. As already stated by Altomonte et al. (2011), policies aimed at enhancing competitiveness may greatly benefit from firm-level evidence; unfortunately, micro-founded data availability remains inadequate to undertake meaningful analysis, particularly of a cross-country nature. This is why CompNet has devoted the past years to the creation of a firm-level indicator database which would be of high standard and detail, yet, available for as larges a set of EU countries as possible.

The database

It comprises indicators based on firm-level information but aggregated at the two-digit industry level to preserve confidentiality, on productivity, labour costs, and employment, computed and collected using the highest standards of comparability across a huge sample of about 700,000 firms per year operating across 58 sectors, from 11 EU countries and over 15 years. The underlying paper shows that the indicator data base is – in terms of coverage and representativeness of the sample – rather superior to other existing comparable data sets (such as Amadeus).

Main results

Labour productivity levels and distribution across countries are depicted below. Apart from the differences in levels – not fully comparable since there is no PPP adjustment – is the strong heterogeneity of firm labour productivity (i.e. the width of the range of values) within and across countries, which is remarkable. In particular, of high policy relevance is the high skewness of the distribution within each country, as represented by the difference (shown to be statistically significant in the paper) between median and average of the productivity distribution. Far from being normal – with many firms centred around the average performance level – the respective country distributions indicate that there are just a few highly productive firms, and a lot of them which are low, and very low-productive. For policy, this implies, first, that when talking about competitiveness/productivity of a country, we can, and must go, much deeper than simple averages (ULC, market shares), since the true relevant distribution is far from symmetric, and, second, that by looking at averages only we are obviously missing to incorporate important empirical evidence that is now increasingly available. This calls, among others, for targeted policies along the productivity distribution derived by empirical findings on how firms in different quintiles of the productivity distribution interact with specific determinants, such as firm size (already available), and others (soon to be available), such as trade and financial constraints.

Figure 1. Labour productivity levels

Among the applications currently possible, we can look at the way the size of the firms and their respective labour costs differ for the least and most productive firms, within a different sector; and compare these across countries. Figure 2 below shows that most productive firms (P90) are up to 10 times larger than the least productive firm (P10) in the same 2-digit sector.

Figure 2. Average size per labour productivity percentile (2003-2007)

Moreover, looking in particular at Spain (Chart 3 below), unit labour costs (ULC) of firms at the bottom and the top of the productivity distribution (P10 and P90) appear to have reacted very differently to the crisis. ULC went up rather sharply before the crisis only for low productive firms, and then declined. High productive firms saw instead their ULC moving only slightly all along the period.

Figure 3. Unit labour cost in manufacturing sector (Spain 2002-2010)

The CompNet data collection method

The CompNet team has established a potentially very powerful research infrastructure, that is, a way of having research teams interacting to answer policy relevant issues, drawing also from firm-level data. Since individual firm-level data are confidential and cannot be shared outside the respective countries, the working method was to establish a small team at the ECB coordinating 13 national teams, which have run – on their computers and with their national firm-level data – programs to compute pre-agreed indicators of competitiveness/productivity (see also Bartelsman et al. 2004, 2009 ). This has implied, at first, high set-up costs, in order to agree on a number of technical features (e.g. how, and which indicator to construct, how to treat outliers, variable definitions, use of deflators, time horizon, and so on). Eventually, however, a very solid structure of expert correspondents has been created, which could rather easily be activated for answering additional issues based on very detailed data. As a matter of fact, CompNet is now extending some of the computations previously mentioned to a number of directions. For instance, the team is now analysing interactions between productivity and financial conditions (e.g. are financially constrained firms less productive? What’s the role of their size? Is credit efficiently allocated across firms, within a given sectors? Has the allocation efficiency changed during the crisis?), as well as labour markets, mark-ups, and trade. The potential policy value of such extensions can hardly be underestimated.

Conclusions

The novel data base built by CompNet represents a powerful tool to complement standard analysis on competitiveness drivers based on macro, as well as detailed sectoral analysis. Having ensured cross-country comparability and indicators availability allows firm-level analysis to make a leap frog step from detailed ‘one off-one country’ studies to a powerful tool for regular policy analysis at the EU level. The relevance that this can have for country surveillance, for instance, in the evaluation of structural reforms, is obvious. The potential of developments of the tool is just started being tapped. It is hoped that in order to allow to exploit such potential fully, statistical offices and national authorities will consider releasing some of the constraints existing in some countries regarding the use, and particularly the matching of different types of firm-level data (for instance, customs data vis-a-vis firms balance sheets and labour statistics). In this context, aiming at setting up a ‘single market for firm-level data’ in Europe would be an additional target for policymakers, that is worth pursuing. Some of the applications mentioned in this article provide, in our opinion, a strong case for it.

Disclaimer: This article draws from incredibly hard work by a CompNet task force composed by colleagues at the ECB, 13 EU national central banks and a number of statistical offices and research institutes. It includes comments by Gian Marco Ottaviano and Carlo Altomonte. Any reporting mistakes however are only mine. The opinions expressed are also my own and do not necessarily reflect those of the ECB or the EU system of central banks.

References

Altomonte C, G Barba Navaretti, F Di Mauro and G Ottaviano (2011) "Assessing competitiveness: how firm-level data can help", Bruegel Policy Contribution 2011/16, November 2011, Brussels.

Altomonte C, T Aquilante and G Ottaviano (2012) “The Trigger of Competitiveness - The EFIGE Cross Country Report", Bruegel Blueprint Series. Volume XVII, July 2012, Brussels.

Bartelsman E, J Haltiwanger and S Scarpetta (2004), "Microeconomic evidence of creative destruction in industrial and developing countries." The World Bank, Policy Research Working. Paper No.3464, December.

Bartelsman, E, J Haltiwanger and S Scarpetta (2009), “Measuring and analysing cross-country differences in firm dynamics,” in Producer dynamics: New evidence from micro data, eds (Dunne, Bradford and Roberts), University of Chicago Press.

Bernard A, J Jensen, S Redding and P Schott (2011), “The empirics of firm heterogeneneity and international trade”, Annual Review of Economics.

CompNet Task Force (2014), "Micro-based evidence of EU Competitiveness: the CompNet database". ECB Working paper no 1634, February.

Melitz, M (2003),“The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity”. Working Paper 8881, NBER, April.

Mayer, P and G Ottaviano (2007), “The Happy Few: the internationalisation of European firms”, Blueprint 3, Bruegel.

4,199 Reads