Aarhus University Seal / Aarhus Universitets segl

The Dynamics of Productivity in Danish and European Manufacturing: Comparing Results across Methods, Countries and Datasets

Project: Research

  • Department of Economics
  • University of Chicago
  • Princeton University
See relations at Aarhus University


The widening productivity gap between the US and the European Union (EU) has been a major concern for politicians for quite a long time. As a way to address this problem, the so called Lisbon Agenda was adopted in 2000 and proposed a set of measures to reverse that trend. The aim was to make the European Union "the most dynamic and competitive knowledge-based economy in the world capable of sustainable economic growth with more and better jobs and greater social cohesion, and respect for the environment by 2010". In practice, Member States were given plenty of leeway (substantial freedom) to implement these guidelines. A mid-term evaluation report in 2005[1] showed that progress was so far unconvincing. In particular, R&D investments were still lagging behind the US and Japan.
In recent studies, despite impressive economic performance and GDP growth over the last decade, Denmark has been shown to be lagging behind in terms of labor productivity and total factor productivity (TFP) growth (see e.g. the KLEMS project[2] or the 2005 OECD Economic Survey), two key measures of economic progress (see below for more). Yet, it is among the frontrunners in terms of R&D investment, together with Sweden and Finland.
One criticism against these types of ranking is that the variables considered might be mismeasured. Many of these studies typically compute total factor productivity as the part of output that can not be explained by the other factors of production (see Syverson, 2004 and Foster, Haltiwanger and Syverson, forthcoming for recent examples). However, these measures suffer from a few weaknesses, as recognized by the authors themselves: the deterministic nature of the model and the necessary assumptions on firm behavior and market structure (see also Van Biesebroeck, 2007).
Another strand of literature tries to estimate the coefficients of a production function, assuming that error term, or the unexplained part, measures TFP. The most important problem is then to deal with the following endogeneity problem: firms might take their input decisions according to what they know about their productivity levels (unobserved to the researcher). A standard methodology in this literature uses control functions to deal with this (Olley and Pakes, 1996, henceforth OP). OP show that the more productive firm will invest more and therefore, to eliminate the endogeneity problem, the investment function can be inverted, so that productivity can be written as a nonparametric function of investment and capital to proxy for the contributions of TFP. Their approach has been criticized subsequently by Levinsohn and Petrin (2003, henceforth LP) and Ackerberg, Caves and Frazier (2007, henceforth ACF), who have offered alternative non parametric approaches to deal with the endogeneity problem.
We plan to contribute to the existing debate in a few dimensions:
- Productivity and Pricing Decisions
In the literature on the estimation of TFP, one really hard problem to solve is to deal with the fact that firms charge different prices and have different markups. A similar difficulty arises if exporting firms charge different prices abroad. In contrast, when deflating sales with the Producer Price Index (PPI), as is commonly done in empirical work, we explicitly consider only the average price in the industry and do not consider pricing heterogeneity within the industry. This introduces a bias in the estimation of TFP. To solve this problem, one would need very precise information about the various products produced by firms, imported and exported, as well as precise information about prices and quantities of these products, so that we could define the evolution of prices at the firm level. Recent papers have suggested a similar approach (e.g. De Loecker, 2007; Foster, Haltiwanger and Syverson, 2007; Jaumandreu and Mairesse, 2007). As we discuss below, recent Danish datasets provide that information. The project would contribute to the literature thanks to the richness of the data.
- Productivity and Methodology
A concern when structurally estimating productivity, as in the Olley and Pakes model, is that the estimate of the structural productivity term, which affects input choice in OP's model, may not correctly approximate for a firm's real productivity. Fox and Smeets (2007a) show that the OP estimator contributes very little to an increase in the statistical fit of productivity regressions. Including a nonparametric function of investment and capital to proxy for the contributions of non-age TFP does not reduce the empirical puzzle that many firms in the same industry produce many more outputs for the same set of inputs. Preliminary results show that the ACF methodology suffers much less from that criticism as their methodology does a much better job at explaining productivity differences between firms. The analysis so far only covered a limited period of time (2001-2003). We will use new version of Danish datasets covering much longer periods (1988-2005, see below). We will also test the robustness of our results thanks to a newly developed international dataset (see next point as well).
- Productivity: An International Comparison
This project attempts to provide a first truly international comparison of TFP and TFP growth using comparable firm-level data available from the dataset ORBIS. The dataset provides accounting information from companies across the globe (US, Canada, Mexico, Japan, Thailand, South Korea, Argentina, Brazil, as well as all EU countries). This would offer a benchmark of the Danish situation and at the same time would allow us to better understand productivity between sectors and between countries. We will focus on the differences in sector composition.
- Productivity and Human Resources Policies
When considering what productivity really means, new approaches try to explain why some firms are more productive than others, therefore trying to better understand this residual of the production function. Few studies have taken into account the fact that input quality differences could explain TFP differences between firms. Economists since at least Griliches (1957) have argued that productivity dispersion reflects the quality of inputs across firms. Economists working with US manufacturing plant data typically measure inputs as the dollar value of physical capital and the number of workers at a firm. Sometimes, employees are separated into production and nonproduction workers. Not surprisingly, labor and capital vary in much greater detail. Two types of machines may have different uses and may not be perfect substitutes, and two types of workers may not have the same contributions to firm output. Fox and Smeets (2007b) have started exploring this issue by using matched employer-employee panel data from Denmark to measure whether the characteristics of workers at a firm impacts its productivity. They find that input quality measures do not explain most productivity dispersion, despite statistically precise coefficient estimates.  This suggests that other forces drive the differences in TFP between firms. One explanation is the role of management, better managed firms being more productive. Preliminary evidence shows that firms with highly paid managers are more productive. Future research should be directed at better understanding the role of the labor input (whether it is the workforce characteristics, the human resources policy or the quality of the management) in explaining productivity differences. The richness of the various Danish datasets would facilitate this task.
- Productivity and International Trade
We would also like to study how productivity is related to firms' import and export decisions. We would then contribute to the debate on whether exporters are more productive than non-exporters because they learn new skills by exporting (learning by exporting) or because they take the decision to enter export markets because they are more efficient (selection). Statistics Denmark provides information about imports and exports by firm and by product as well as their origin or destination.
Effective start/end date01/09/200831/08/2011


Research outputs

ID: 128878429