How Informative Are Firms’ Statements About Competitiveness? Assessing the Quality of Information about Competitiveness in the French Quarterly Business Survey in Goods-Producing

INSTITUT NATIONAL DE LA STATISTIQUE ET DES ÉTUDES ÉCONOMIQUES

INSEE - DEEE AND CREST

JULY 2004

We are indebted to Didier Blanchet, Hélène Erkel-Rousse, Stéphane Gregoir, Laurent Ménard and Sébastien Roux for most valuable suggestions.

We are also grateful to Benoît Heitz and Jean-François Loué who provided us with macroeconomic data on competitiveness indicators.

_____________________________________________

Département des Études Économiques d'Ensemble - Timbre G201 - 15, bd Gabriel Péri - BP 100 - 92244 MALAKOFF CEDEX - France - Tél. : 33 (1) 41 17 60 68 - Fax : 33 (1) 41 17 60 45 - CEDEX - E-mail : [email protected] - Site Web INSEE : http://www.insee.fr

This paper is the English version of DESE Working Paper G2004/01, « La compétitivité exprimée dans les enquêtes trimestrielles

sur la situation et les perspectives dans l’industrie », downloadable on the INSEE website.

Working papers do not reflect the position of INSEE but only their author's views.

How Informative Are Firms’ Statements About Competitiveness?

Assessing the Quality of Information about Competitiveness

in the French Quarterly Business Survey in Goods-Producing Industries

Patrick AUBERT and Marie LECLAIR

2

Abstract

Since 1997 Quarterly Business Surveys in EU countries explicitly ask firms to state the way their « competitiveness » has been evolving over the last three months previous to the survey. In this study we inquire the possibility of using this subjective, self-stated information as a direct measure of competitiveness in micro- and macroeconomic studies. We therefore assess the quality and relevance of firms’ statements about their own competitiveness in the French Quarterly Business Survey by comparing it to standard, indirect measures at the individual and aggregated level.

First, we merge the Quarterly Business Survey with data that provide extensive information on possible determinants for competitiveness (accounting data, structure of the workforce, etc). A higher competitiveness at the individual level is significantly correlated with higher sales, higher production, lower costs of materials and lower labor costs.

However this correlation concerns raw values, not values relative to other firms in the sector. This result should warn us against what firms truly mean while stating the way their competitiveness evolves: they might indicate variations of the economic situation or business conditions, rather than a true evolution of their market situation relative to competitors.

Second, we sum up firm-level statements about competitiveness into an aggregate competitiveness indicator. This indicator is constructed as usual confidence indicators i.e. as a balance between positive and negative statements about changes in competitiveness. It seems fairly well correlated with other common « competitiveness » indices, such as aggregate labor productivity, exchange rates or aggregate costs relative to foreign competitors.

Quarterly Business Surveys hence seem to provide a relevant firm-level measure of competitiveness that can be used in microeconomic studies, for instance in the field of public policy evaluation. However we believe it provides additional information, rather than replaces, more traditional measures of competitiveness such as productivity or costs. Quarterly Business Surveys can also be used to construct a direct aggregate competitiveness indicator that seems fairly relevant, though the analysis might lack robustness due to the short observation time period.

Keywords: competitiveness, Business Survey

Classification JEL : L10, C80

3

How Informative Are Firms’ Statements About Competitiveness? Assessing the Quality of Information about Competitiveness

in the French Quarterly Business Survey in Goods-Producing Industries

Introduction

Following Eurostat recommendation, Quarterly Business Surveys (QBS) in EU countries explicitly ask firms to state the way their « competitiveness » has been evolving over the last three months previous to the survey, on the domestic, European and non-European markets. In France, Quarterly Business Surveys have been including this question since 1997. However, it has never been used yet to construct a measure of competitiveness, neither at the micro- nor at the macro-level, although it should provide fruitful information along several lines.

First, it provides a direct measure of competitiveness, i.e. competitiveness as managers themselves perceive it. It may therefore be more relevant than other usual proxies for competitiveness, such as labor productivity or costs. This is all the more interesting as most policies explicitly aim at enhancing « competitiveness » of firms in general, rather than components of it.

However, one major drawback is that « competitiveness » is a highly subjective concept. QBS therefore allow us to know what firms mean while stating the way their « competitiveness » evolves, in terms of standard variables like costs or productivity. In other words, it enables us to confront firms’ statements about their own competitiveness in Quarterly Business Surveys to more traditional measures from other statistical sources.

An empirical enquiry into the fundamentals behind competitiveness is all the more important as there is no such thing as an explicit definition of competitiveness in Economics. This is the paradox about it: although competitiveness is frequently quoted among major economic stakes by managers and policy makers, it still lacks a clear and consensual definition among economists.

A second paradox is that competitiveness is often quoted in a macroeconomic context, i.e. in the context of competition between countries, whereas it is a priori more relevant at the micro-level.

Because of those « paradoxes », competitiveness is a much debated and criticized notion. Critics target both its relevance and the way it is measured. In particular, country ratings along some « competitiveness index » (e.g. in the World Economic Forum annual report) have been sharply criticized, due to the fact that they relied on highly arbitrary measures1 for competitiveness (Gregoir and Maurel, 2002). The relevance of competitiveness at the macro-level has been sharply criticized by some economists (Krugman, 1994).

In this study, we inquire how informative firms’ statements about their own competitiveness are in Quarterly Business Surveys. More precisely, we aim at

1 According to Gregoir and Maurel (2002), the World Economic Forum annual report defines

competitiveness as “what is measured by the two indices proposed by the World Economic Forum”.

4

assessing the quality and relevance of this information. We check whether it is correlated or not to usual components of competitiveness, e.g. productivity and costs.

More specifically, we wonder whether self-statements about competitiveness in QBS could be useful in public policy evaluation. If this information proves relevant, it might be very useful in such policy evaluation studies, both at the micro and macro level. First, it is synthetic; therefore it should be more informative than any particular component of competitiveness. Second, it is available faster than any other « measures » of competitiveness.

We do not aim at giving an explicit definition of competitiveness. Neither do we intend to discuss the relevance of the notion on theoretical grounds. In what follows, we therefore use the word « competitiveness » in a very general meaning, including any definition that can be given in the economic literature.

The remaining of the paper is as follows. In section 1, we assess the quality of firms’ statements about competitiveness at the micro level. We compare them to other usual measures for competitiveness. To do so, we merge the QBS data with other data sources, namely the DADS file and the BRN file (see Annex 2). In section 2, we aggregate statements about competitiveness into a macroeconomic index of competitiveness. We compare it to other macroeconomic indexes, such as aggregate labor costs or productivity.

Box 1: The Data on Competitiveness

Quarterly Business Surveys (QBS) in manufacturing industries (« enquêtes trimestrielles sur la situation et les perspectives dans l’industrie ») aim at knowing managers’ opinion about their firm’s activity, results and prospects. Firms are interrogated on a quarterly basis, in January, April, July and October.

Since October 1997, the following question has been included in the survey:

Tendency during the last 3 months of your competitive position … § on the domestic market § on foreign markets within the EU § on foreign markets in non-EU countries

For each geographic market, firms can answer by: « increasing », « stable » or « decreasing ».

Our dataset contains all observations between 1997:Q3 and 2002:Q4. 4,283 firms answered the question about competitiveness, some of them for more than one good. On the whole, we have information on competitiveness for 6,380 goods (see Annex 1).

Firms are not present in the dataset every year. We therefore have an unbalanced panel over the period 1997:Q3 to 2002:Q4.

These quarterly data can be aggregated into yearly data (see Annex 2). We do so in order to merge them with other datasets, available on a yearly basis. More precisely, we merge our QBS data with the BRN file, or « Bénéfice Réel Normal », which provide information on sales, production, exports, materials, labor costs, and employment. We also merge the QBS data with the DADS (or Déclaration Annuelle de Données Sociales). The former dataset includes information on the structure of the workforce, i.e. the shares of age, gender and occupations, as well as the share of full-time workers.

5

I - Competitiveness and Microeconomic Indicators

In this section, we are interested in statements about competitiveness at the firm level. As « competitiveness » is not defined in the survey, answers actually combine true information on the competitive position of firms with subjective understanding of the notion of competitiveness by respondents. Therefore, there might be a large heterogeneity in answers, since they include differences in definitions as well as true differences in competitiveness.

In particular, respondents to the survey may assess competitiveness referring to the economic results of their firm, while others may rather refer to some more "technical" indicators (see Box 2). For instance, some companies may evaluate their competitiveness referring to their potential, whereas others would refer to their results. In the first case, a firm adopting some high performance workplace practice might consider it as an increase in competitiveness, if it expects that this will imply an increased productivity in the near future. In the second case, a firm facing a positive external shock increasing its sales might interpret it as a gain in competitiveness, even if it is unrelated to its strategy, technology or costs.

Besides, respondents may assess competitiveness referring to its cost component only, or its non-cost component (e.g. quality) only, or a mixture of these two components (see Box 2). Last, among respondents assessing competitiveness from values such as costs or productivity, some may assess competitiveness referring to raw values (either due to misinterpretation of competitiveness or because they do not know the costs and productivity faced by their competitors), whereas others would refer to values relative to their competitors’.

Apart form differences in the definition of competitiveness, subjectivity is another source of heterogeneity among respondents. Some managers may be more or less optimistic than others. Besides, some may be sensible to small variations, whereas others could wait until they observe large, undeniable trends before they consider that their competitiveness has dropped or increased. Lastly, observations suggest that some respondents might interpret the “evolution” of competitiveness as a deviation from some reference level, rather than as a variation from one quarter to the following one.

Therefore, we try to know what firms’ statements about competitiveness truly refer to by comparing those statements to tangible measures from other datasets (see Annex 2).

Our approach is three stages: first, we compare statements about competitiveness to standard accounting data. In particular, we separately control for costs of materials, labor costs, production, and sales. Second, we control for structural factors, which might make firms more or less able to face competition. For instance, we check whether competitiveness is correlated with the age or qualification structure of the workforce. Third, we try to highlight stable behaviors of response in the course of time, i.e. fixed effects. We do so by introducing controls for the situation of firms in 1997 that may capture part of those fixed effects.

6

Box 2 : Competitiveness in Microeconomics

“Competitiveness" intuitively refers to competition. Therefore, a definition could be the “capacity to sustain competition”. In the theoretical framework of a market with homogeneous goods and perfect competition, competitiveness would mean the capacity to produce with a marginal cost below the market price. Then "competitiveness" would be a synonym of "survival". This theoretical framework is rather restrictive. A more relevant theoretical framework is the monopolistic competition model. Within this framework, firms produce differentiated goods, which remain substitutable to a certain extent. Each firm is in the situation of a "monopoly", because it is the only producer of its very type of good, but it faces "competition" since other goods are substitutable to the firm’s product. When there is monopolistic competition, prices are not directly comparable any more. Therefore, competitiveness is not any more a matter of having production costs below or above the “market price”, which result in survival or not. A company may increase its selling prices (for instance, following increases in costs) without its demand falling to zero, since substitution with other goods is only imperfect. In this framework, how can competitiveness be precisely defined? Lorenzi (2002) proposes to define it as the "capacity [of a company] to sell what it produces while making profit on a long term basis ". In other words, a firm is competitive if it meets demand at a price higher than its production cost. However, this definition leaves no room to variation in competitiveness: a firm is or is not competitive, but the definition is unclear about what would be an increase or decrease in competitiveness. Other authors propose to define it as the "capacity to increase market shares". This definition is appealing: variation in market shares can be directly interpreted as variation in “competitiveness”. It can also be understood in a macroeconomic framework: countries replace firms and the world total production replaces the "market". Nevertheless, it can be questionable out of economic grounds. Firms do not primarily aim at increasing market shares: they aim at increasing profits, and they only seek to increase market shares when it implies higher profits. Therefore, true variations in competitiveness are variations in the capacity to increase market shares, not variations in market shares themselves. Thus, observed market shares can be a misleading measure of competitiveness. Actually, most authors fail to give a precise definition of competitiveness because they try to combine too many components . Definitions often combine considerations on costs, effectiveness of the production technology, demand and the firm’s objectives. However, it seems that there should be no precise definition without distinguishing those aspects. For instance, a convenient distinction can be made between the cost and non-cost dimensions of competitiveness. Quite obviously, “cost competitiveness" refers to cost considerations. It thus includes input costs (labor costs and costs of materials) as well as the efficiency of the production technology (the productivity of the inputs). Competitiveness increases when production costs decrease, holding quality constant, independently of the price or the amount of sales. For instance, if two firms use the same technology and if production costs are comparable, a variation of the cost competitiveness of one firm relative to the second one can be defined as a variation of the production costs of this firm relative to the other one’s production costs. On the contrary, "non-cost competitiveness" refers to demand considerations. Non-cost competitiveness increases when demand increases holding prices constant, or when the firm can increase prices holding demand constant. In economic terms, an increase in non-cost competitiveness result in a change in the demand curve, independently of production costs. Therefore, non-cost competitiveness refers to the relative quality of goods, i.e. product differentiation. In the monopolistic competition model, a good measure for non-cost competitiveness is the elasticity of demand. Cost and non-cost competitiveness thus refer to very distinct aspects, one referring to the structure of costs and the other to the shape of demand. There is no straightforward way to combine them into a single, synthetic competitiveness indicator.

7

We regress yearly changes in competitiveness using an ordered multinomial logit model. These regressions remain descriptive. We do not pretend to highlight causalities. Instead, we describe correlations between tangible variables and the « competitiveness » as managers perceive it.

I.1 Accounting Variables

Results of the regression of yearly changes in competitiveness are presented in table 1a. Explanatory variables are a set of accounting variables, including variations in total sales, production, labor costs and costs of materials.

Table1a: Regression of Yearly Changes in Competitiveness Explanatory variables Parameter Std Error

Variables at the firm level:

Change in log total sales 1.079 (0.243)

Change in log total cost of materials -0.714 (0.217)

Change in log total labor cost -1.370 (0.228)

Change in log production 1.504 (0.381)

Change in log employment 0.408 (0.174)

Variables at the sector level:

Change in log total sales 0.733 (0.680)

Change in log total cost of materials -0.144 (0.907)

Change in log total labor cost -0.581 (0.751)

Change in log production 0.161 (1.327)

Change in log employment 0.604 (0.720)

Observations 6,870

Percentage of:

Concordant pairs 58.20

Discordant pairs 41.80

Log likelihood (constant) 18413

Log likelihood (with explanatory variables) 18166

Change in -2*log likelihood 247 Note: Ordered multinomial logit model. Year 1998 to 2001 Dependant variable: Yearly changes in competitiveness on the French market. The dependant variable can take 5 different values (see Annex 2): at least two quarters with competitiveness decreasing (and no increase at all); one quarter of decrease only during the year; stable competitiveness across all quarters; one quarter of increase only during the year; at least two quarters with competitiveness increasing (and no decrease). We do not display estimates for constants in this table. Controls are year and industry dummies (NES16 classification).

All firm-level variables are significantly correlated to changes in competitiveness. There is a positive correlation between an increase in competitiveness and an increase in sales and production, or a decrease in labor costs or costs of materials2.

Results in table 1 illustrate the different aspects of competitiveness. Holding total costs constant

(i.e. labor costs plus costs of materials MpwL m+ ), an increase in the value of production ( )pQ

is equivalent to a decrease in unit costs (pQ

MpwL m+). The correlation with competitiveness

2 We control for employment. Therefore, results would have been the same if variables had been measured per head. In

other words, all other things being equal, competitiveness increases when production per head (i.e. labor productivity) increases and when costs per head (i.e. unit costs) decrease.

8

therefore illustrates the « cost » component of competitiveness. It increases when unit costs

decrease, either because the cost of input (p

ppw mor ) decreases or because productivity

(MQ

LQ

or ) increases. Similarly, holding production and costs constant, competitiveness rises

when sales ( pY ) rise. It is the « non-cost » component of competitiveness: it reflects the ability of

the firm to increase its price p or volume V, holding costs constant.

Those are correlations « all other things being equal », not causal effects. In particular, if there is causality behind those correlations it could be either one way or the other. Explanatory variables include variables that might be either determinants or consequences of a change in competitiveness. For instance, labor costs might rather be a determinant of competitiveness: the negative correlation is likely to mean that an increase in labor costs decreases competitiveness in a firm. On the contrary, the positive correlation between competitiveness and employment may illustrate the other way of causality: a firm with increasing competitiveness may grow, and thus hire new workers.

I.2 Raw Values, Rather than Relative Values, are Correlated to Statements on Competitiveness

Since competitiveness refers to the ability to face competition, it should be correlated to relative, rather than absolute, costs or productivity levels. For instance, competitiveness should be correlated to market shares, i.e. sales relative to the total amount of sales on a market, rather than raw amount of sales by a firm. Firms should not consider an increase in sales as an increase in competitiveness if it stems from an increase in demand that is common to all firms on a market.

We test this by including in the specification explanatory variables (e.g. sales, production, costs, etc) both at the firm- and at the sector-level. For instance, we include both labor costs in the firm and in the aggregate of all firms belonging to the same sector. Sectors are defined at a very narrow level (along NAF 7003 classification, i.e. the whole economy is divided into 700 distinct sectors). We consider that they are equivalent to the « markets » on which firms are selling their goods.

This is equivalent to decomposing competitiveness into a « relative » plus an « absolute » component. If it is correlated to raw (or absolute) variables only, then only firm-level variables will be significant. On the contrary, if competitiveness is correlated to relative variables only, then both firm-level and sector-level variables will be significant, with coefficients of opposite signs.

Results advocate for the first explanation i.e. competitiveness is correlated to raw rather than relative values. For instance, managers perceive an increase in labor costs as a decrease in competitiveness, irrespective of whether labor costs increase more or less for competitors.

There may be several explanations for this. The first one is a lack of information. Respondents (i.e. managers) may not know how competitors’ costs and sales have been evolving over the three months previous to the survey. Thus they may mistake in assessing the evolution of their relative costs and market shares, or they may simply decide that they would only refer to raw sales and costs to state the way competitiveness has been evolving. The second possible explanation is misinterpretation. As we discussed in previous sub-sections, managers may interpret “competitiveness” as the economic situation of their firm or as business conditions, rather than as a position respective to competitors.

Nonetheless, results may suffer two potential drawbacks. First, we consider that sector in the NAF 700 classification provides a good definition for the market. However, although this classification is quite precise, it may not cover the relevant market in some cases. Second, variables (sales,

3 The French NAF 700 classification is very close to the European NACE Rev. 1 classification.

9

production, costs) for the aggregate of competitors on the market are calculated for French firms only. Thus we only observe variations of variables relative to French competitors, not foreign competitors. In markets where French firms mainly face competition from foreign firms, our measured variables for the “aggregate of competitors” may not be relevant.

I.3 Are there Structural Components of Competitiveness?

Managers may interpret competitiveness as a potential, rather than a result. For instance, a manager may think that his firm’s competitiveness has improved because some structural changes have been performed and are likely to enhance the fi rm’s ability to face competition and/or external shocks. In this case, there may be an increase in competitiveness without any change in sales, or costs, or productivity over the year.

There may be several such structural changes. For instance, they may concern the structure of the workforce (e.g. by age or skill), fixed assets (e.g. computerization or purchase of high-technology machinery) or workplace organization (e.g. just-in-time or total quality management).

Table 1b presents results of a larger specification, including controls for changes in the structure of the workforce. We consider the structure by skills (share of workers in unskilled occupations, share of workers in highly skilled occupations and share of interns) and age (share of younger and older workers). We also introduce variables for changes in work intensity (shares of full time workers, average number of hours worked per day by full time workers)4.

The age structure and the occupational structure of the workforce seem correlated with competitiveness. More precisely, an increase in the share of younger workers (aged less than 30) and in the share of highly skilled workers is significantly correlated with an increase in competitiveness.

In the case of age, the correlation may stem from a reverse causality: a rise in competitiveness may favor hirings in a firm. Since most hirings are done among younger people, this may lead to an increase in the share of workers aged less than 30.

I.4 Unobserved Heterogeneity and Fixed Effects

Statements about competitiveness may include permanent component, or « fixed effects ». Some respondents may state that competitiveness is increasing, or decreasing, more often than other. There are two main reasons why we think fixed effects might exist. First, there is subjectivity. Some managers may be more or less optimistic than others. Similarly, some managers may consider small variations in competitiveness, whereas other will consider only large trends.

Second, there may be structural differences between firms, which are not captured by our controls of the structure of the workforce. For instance, changes in competitiveness may be more frequent on markets where there are frequent innovations, or where the volatility of demand is high, or where the production technology is evolving fast. On the contrary, there may be very little change in competitiveness on markets where there are few competitors, a stable demand and no evolution in the technology.

There are numerous standard methods to deal with unobserved fixed effects in panel data econometrics. This would be beyond the scope of this study and we do not apply such methods here.

When fixed effects stem from permanent characteristics of the market, they can be taken into account by including additional controls among explanatory variables. We do so by including in the specification firms characteristics in 1997, i.e. one year previous to the first year of observation (table 1b).

4 Controls for the purchase of innovative assets or the implementation of high-performance workplace practices would be

very interesting as well. Unfortunately such variables are not available in our dataset.

10

Table 1b: Regression of Yearly Changes in Competitiveness. Including more controls

Explanatory variables Parameter Std Error Variables at the firm level:

Change in log total sales 1.146 (0.287) Change in log total cost of materials -0.785 (0.251)

Change in log total labor cost -1.534 (0.279) Change in log production 1.604 (0.448)

Change in log employment 0.391 (0.218) Change in log exports -0.056 (0.039)

Variables at the sector level: Change in log total sales 0.758 (0.763)

Change in log total cost of materials 0.072 (1.082) Change in log total labor cost 0.125 (0.882)

Change in log production -0.138 (1.567) Change in log employment -0.200 (0.863)

Variation of the structure of the firm Change in the wage bill share of interns 0.265 (3.076)

Change in the wage bill share of unskilled workers 0.152 (0.338) Change in the wage bill share of highly skilled workers 1.038 (0.478)

Change in the WBS of workers aged 24 and less 2.210 (0.727) Change in the WBS of workers aged 25 to 29 2.730 (0.945)

Change in the WBS of older workers (age 50 and more) -0.264 (0.730) Change in average hours worked per day (full time

workers) 0.050 (0.053)

Change in the wage bill share of full time workers 0.145 (0.350) Firms characteristics in 1997 Share of interns in employment 1.208 (1.192)

Share of unskilled workers 0.040 (0.158) Share of highly skilled workers 1.349 (0.222)

Share of workers aged 24 and less 1.634 (0.590) Share of workers aged 25 to 29 1.190 (0.515)

Share of older workers (aged 50 and more) -1.783 (0.324) Average hours worked per day (full time workers) -0.109 (0.042)

Share of full time workers -0.072 (0.268) Log of total number of hours worked 0.111 (0.024)

Labor productivity (production/employment ratio) 0.204 (0.072) Share of exports in sales -0.679 (0.120)

Observations 5,459 Percentage of:

Concordant pairs 62.2 Discordant pairs 37.9

Log likelihood (constant) 14692 Log likelihood (with explanatory variables) 14244

Change in -2*log likelihood 448 Note: Ordered multinomial logit model. Year 1998 to 2000 Dependant variable: Yearly changes in competitiveness on the French market. The dependant variable can take 5 different values (see Annex 2): at least two quarters with competitiveness decreasing (and no increase at all); one quarter of decrease only during the year; stable competitiveness across all quarters; one quarter of increase only during the year; at least two quarters with competitiveness increasing (and no decrease). We do not display estimates for constants in this table. Controls are year and industry dummies (NES16 classification), plus a dummy equaling one if the firm has no export during the year.

Competitiveness increases more often in firms with many highly skilled workers and firms with a young labor force (i.e. many younger workers and few older workers). It rises more often in large, productive firms with little export and few hours worked per day.

Those results are hard to interpret. In particular, given the short time dimension of the data, it is difficult to distinguish between true structural characteristics and mere cyclical variations.

There may be several explanations to the fact that competitiveness rises more often in larger firms. It may be due to larger scale economies, a more intensive effort in R&D, or a better knowledge of the market in those large firms.

11

Besides, a larger share of skilled workers results in more frequent innovations. This may explain the reason why competitiveness increases more often in firms with more skilled workers. On the contrary, competitiveness rises less often in firms with many older workers. This is consistent with the fact that older workers are more often matched to old, less productive firms. Another reason could be that innovation is harder when the labor force is old, for instance because of skill obsolescence.

Last, competitiveness increases less often (or decreases more often) in firms with a larger number of hours worked by full time workers in 1997. The explanation behind this correlation is unclear.

I.5 Changes in Competitiveness may be Different on the Domestic and Foreign Markets

In the Quarterly Business Surveys, firms are asked to state the way their competitiveness has been evolving on the French market, as well as on foreign markets (in foreign EU-countries and in foreign non-EU countries). In the data, we find intra-firm variability across markets: in particular, there are firms whose competitiveness rises on the French market whereas it decreases on foreign markets, and vice versa (see Annex 1).

There are several reasons why competitiveness may evolve differently in distinct markets for the same firm. The most straightforward reason is that competitors are not the same. The situation relative to competitors may therefore evolve differently. For instance, let’s imagine that a firm is selling goods in France and in other EU countries, competing with French firms on the domestic market and with international firms in the EU market. Let’s now imagine that an institutional change in France increases labor costs in all French firms. In this case, the firm will face higher labor costs. However, in relative terms, its costs will remain constant relative to competitors on the French market, while they will rise relative to foreign firms. Then competitiveness will remain constant on the French market and decrease on the EU market.

However, such an explanation is inconsistent with results in the previous sub-section. If statements about competitiveness reflect raw values of costs or sales, rather than relative ones, then the fact that there are different competitors on different markets should not matter. A second explanation has to do with the non-cost component of competitiveness. Consumers’ preferences, and therefore the shape of demand, may vary from one market to the other. This could partly explain differences in competitiveness across markets.

Results of regressions on the French, EU and non-EU markets are presented in table 2. In the case of foreign markets, we use to different specifications. Specification 1 has the same explanatory variables as before. Specification 2 includes controls for the way competitiveness has been evolving on the domestic (i.e. French) market.

12

Table 2: Regression of Yearly Changes in Competitiveness French, EU and non-EU markets

Competitiveness on markets Domestic EU countries Non-EU countries Specification [1] [1] [2] [1] [2]

Variables at the firm level: Param Std Err Param Std Err Param Std Err Param Std Err Param Std Err Change in log total sales 1.083 (0.244) 1.053 (0.247) 0.499 (0.257) 1.296 (0.256) 0.855 (0.256)

Change in log total cost of materials -0.737 (0.218) -0.477 (0.223) 0.024 (0.242) -0.486 (0.229) -0.133 (0.238) Change in log total labor cost -1.243 (0.235) -1.164 (0.239) -0.408 (0.250) -1.248 (0.245) -0.631 (0.246)

Change in log production 1.533 (0.380) 0.937 (0.387) -0.142 (0.415) 0.691 (0.397) -0.121 (0.410) Change in log employment 0.505 (0.189) 0.154 (0.185) -0.271 (0.198) 0.070 (0.188) -0.160 (0.196)

Change in log exports -0.052 (0.035) 0.065 (0.036) 0.102 (0.038) 0.060 (0.036) 0.082 (0.037) Observations 6,694 6,712 6,712 6,709 6,709 Percentage of:

Concordant pairs 58 56 77 55 71 Discordant pairs 40 41 20 41 27

Tied pairs 2 4 3 5 2 Log likelihood (constant) 17921 16944 16944 16161 16161

Log likelihood (with explanatory variables) 17676 16771 13732 16016 14393 Change in -2*log likelihood 245 173 3212 145 1768 Note: Ordered multinomial logit models. Year 1998 to 2001 We do not display estimates for constants in this table. Controls are year and industry dummies (NES16 classification), share of exports in total sales, plus a dummy equaling one if the firm has no export during the year. Specification 2 (for EU and non-EU markets) includes 4 dummies equaling one when competitiveness on the French market has decreased at least two quarters during the year, when it has decreased one quarter only, when it has increased one quarter, and when it has increased at least 2 quarters, respectively.

Under specification 1, results are similar on foreign markets and on the French market. A rise in competitiveness is significantly correlated with a rise in total sales, production, and a decrease in both labor costs and costs of materials. Even on foreign markets, competitiveness is correlated with total sales, not exports.

Results are somehow different when we control for what happens on the French market (specification 2). Once this control is included, competitiveness on the EU market is not anymore correlated with total sales, production or costs. Only exports appear to have a significant effect. On non-EU markets, labor costs and total sales remain significant. This would mean that an increase in labor costs has an even worse effect on competitiveness on the global market than on the domestic market.

II - Competitiveness and Macroeconomic Indexes

Quarterly Business Surveys are commonly used for building aggregate indexes. These indexes describe the economic situation of the whole manufacturing industry. They are constructed as a balance, i.e. the difference between the percentages of positive and negative opinions, using each firm’s sample weight.

In the previous section, we analyze firms’ statements about competitiveness at the individual level. From now on we consider the aggregate competitiveness indicator, built from the QBS as the balance between “increasing” and “decreasing” variations in competitiveness. We assess its relevance by comparing it to other macroeconomic indicators usually used as proxies for competitiveness.

There is one fundamental difference between the micro- and macro-level. At the individual level, competitiveness depends on firms’ economic situation relative to all competitors, including other French firms. At the aggregate level, the competitiveness of French firms depends on the aggregate situation of French firms relative to foreign competitors only.

13

Usual macroeconomic indexes of competitiveness are based on exchange rates and a measure of relative prices. We also build indicators that are similar to those used in the previous section: productivity, unit costs, etc. We compare them to the aggregate competitiveness index built from the Quarterly Business Surveys.

Figures 1a, 1b, and 1c: Balance between Positive and Negative Opinions about Competitiveness

On the domestic market

-0,06-0,04-0,02

00,020,040,060,08

0,10,12

Q2/97Q4/9

7Q2/9

8Q4/9

8Q2/9

9Q4/9

9Q2/0

0Q4/0

0Q2/0

1Q4/0

1Q2/0

2

On the European market

-0,18

-0,13

-0,08

-0,03

0,02

0,07

0,12

Q2/97Q4/9

7Q2/9

8Q4/9

8Q2/9

9Q4/9

9Q2/0

0Q4/0

0Q2/0

1Q4/0

1Q2/0

2

On the foreign market(except EU)

-0,2

-0,15

-0,1-0,05

0

0,05

0,10,15

0,2

Q2/97Q4/9

7Q2/9

8Q4/9

8Q2/9

9Q4/9

9Q2/0

0Q4/0

0Q2/0

1Q4/0

1Q2/0

2

Figures 1a, 1b 1c represent the aggregation of firms’ statements about the evolution of competitiveness on the domestic market, on the European market, on foreign markets (i.e. all non-EU countries). It is the balance between the number of firms that declare their competitiveness has increased over the preceding quarter, and the number of firms that declare their competitiveness has decreased. Firms are first weighted by their total sales, in order to have a representative index at the sector level. Then, this sector-based index is weighted by the weight of the sector sales in the manufacturing industry5. The balance of opinions is not the best statistics to sum up the three kinds of answers to the competitiveness question (increasing, decreasing, stable). The first principal component of a Principal Component Analysis would have been more informative than this balance of opinions. Nevertheless, the interpretation of this index is easier even if it certainly does not take account of firms stating a stable competitiveness. It would be all the more relevant as there are few such firms. Unfortunately, this is not the case here. On average 70% of firms declare their competitiveness has not changed over the preceding quarter. To have a benchmark, only 40% or 50% of them declare that their productivity or demand has not changed over the preceding quarter.

The balance of statements about competitiveness has a very erratic evolution (figures 1a-c). Besides, there is only a very gross similarity between aggregate competitiveness on the domestic, EU and non-EU markets. This contradicts the idea that competitiveness index describes the growth of fundamentals such as productivity or production cost. If this were true, the changes in competitiveness on the domestic market would be very close to those on the foreign market, which is not the case.

Since 1997, there has been a decreasing trend of competitiveness, whatever the market. There are three peaks of competitiveness on 1997:Q4, 1999:Q4, 2001:Q4 and three drops of competitiveness on 1999:Q1 and 2001:Q1.

5 The balance of opinions about competitiveness is calculated as ∑ ∑∑∑ ∈∈

−=

sisj

tj

tititi

ss

s

st S

DUSbalance

s

,

,,,

''

).(α

α

where Si,t is the total amount of sales of firm i at quarter t, sα is the weight associated to sector s Ui,t is a dummy which

indicates that firm i declares an increase of its competitiveness. Di,t is a dummy which indicates that firm I declares a decrease of its competitiveness.

14

Box 3: What Do Competitiveness Mean at the Macroeconomic Level?

At the macroeconomic level, we cannot understand the notion of competitiveness in a framework of pure and perfect competition. The point is the same as the one put forward in the previous section. Without heterogeneity and an imperfect substitution between goods, uncompetitive producers would simply disappear.

At the macroeconomic level, competitiveness is not applied to firms but to countries. The competitiveness of a country may describe its amount of imports and exports and its market share.

Armington (1969) builds a macroeconomic model in which he explains how price differences between imported goods may be persistent. It enables to understand market shares of each country by explaining the consumer demand. In Armington’s model, each consumer knows the origin of goods and has preferences over origin. The amount of each product he decides to buy depends on prices and on preferences over the origin of goods. Export and import equations derive from aggregating the consumer demand of each country. As consumers’ preferences, import and export equations depend on relative prices and preferences over origins. The import “price-competitiveness” is a relative price index that we introduce into the import equation. It is calculated as the ratio of the (weighted) price of imported goods divided by the price of domestic goods. The weight of imported goods depends on the domestic market share of the exporting (i.e. foreign) countries. The export “price-competitiveness” is introduced into the export equation and depends on the relative price of the domestic goods on each foreign market.

Thanks to this model, DPAE (Ministry of Economy) build a multilateral competitiveness index, namely the inverse of the real effective exchange rate. It is the ratio of an average price of foreign goods divided by an average price of

domestic goods. Each foreign price j is weighted by djα , which is calculated as follows:

jddd

djdj ≠

−=

∗

∗

,1 α

αα and ∑

∑∈

=

∗

Χ

Χ×

ΧΧ

=Mm

n

p

mp

mj

Md

md

dj

1

α (1)

Where m is an elementary market: there are as many elementary markets as there are manufactured goods multiplied by the number of importing countries. M is the global market, an aggregation of elementary markets

mjΧ are the amount of exports from country j toward the elementary market m

MfΧ =∑

mfΧ is the total amount of exports from the domestic country (in our case, France).

n is the number of competitors. In Armington’s demand model, prices are exogenous. However, by analogy between firms and countries, we can build a supply model. The ability of a country to supply goods at a relatively lower price depends on its relative production cost and productivity. We can build a cost competitiveness index by weighting the unit cost of production

of each foreign country by djα and dividing this average foreign cost by the domestic unit cost of production.

Krugman (1994) criticized the notion of competitiveness understood as a market share and presented as the purpose of economic policy. A new multidimensional definition has been proposed: competitiveness is the ability to increase welfare and depends not only on prices and costs but also on institutions, education and externalities. Multidimensional indexes were built, for instance by the World Economic Forum and the Institute for Management Development. They mix richer characteristics but are also more questionable.

15

II.1 The Euro/Dollar Exchange Rate is Correlated with the Competitiveness on Foreign Markets

Competitiveness on foreign markets should depend on exchange rates. Differences in exchange rates result in differences of the relative price of French goods on foreign markets. It also results in differences of the relative price of non-European goods on the French market. Nevertheless, it might be more important as a determinant of competitiveness on foreign markets, where French firms often face tougher competition than on the domestic market.

Out of Europe, the relative prices of French goods depend on the Euro/Dollar exchange rate. On figure 2, we plot both the change in the Euro/Dollar exchange rate and the competitiveness index built from the QBS (competitiveness in non-EU countries). Since 1997, both indexes have been evolving closely.

Figure 2: Euro/dollar Exchange Rate and the Competitiveness Index Built from the QBS Competitiveness on non-European markets

-0,15

-0,1

-0,05

0

0,05

0,1

0,15

Q2/97

Q4/97

Q2/98

Q4/98

Q2/99

Q4/99

Q2/00

Q4/00

Q2/01

Q4/01

Q2/02

the euro/dollar exchange rate

competitiveness on the non European market

Table 4 present Spearman rank-order correlations between the Euro/Dollar exchange rate and our competitiveness index. Indeed we find a positive, significant correlation between the competitiveness on foreign markets and the Euro/dollar exchange rate. This positive correlation means that if the dollar rate increases, French goods are relatively less expansive. Consequently, French firms are more competitive than foreign non-European competitors on non-European markets.

On European or domestic markets, the Spearman correlation between the Euro/dollar exchange rate and competitiveness is not significant.

Table 4: Spearman Rank-Order Correlation: Competitiveness Indexes and Euro/Dollar Exchange Rate Correlation (p-value)

Competitiveness on The domestic

market

The domestic market (quarterly lagged series6)

The European market

The non EU market

0.25466 0.16431 0.33371 0.55957 The euro/dollar exchange rate (0.253) (0.465) (0.129) (0.007) Note: The change in the euro/dollar exchange rate is the change between quarter t and quarter t+1 in the euro equivalent for

one dollar.

6 Due to a data collection issue, there may be a time difference between the competitiveness index built from the QBS and

other indexes of productivity. Therefore, we also present results using a quarterly lag, in order to assess the robustness of our results.

16

II.2 Changes in Competitiveness are Very Close to Changes in the Real Effective Exchange Rate

The Euro/dollar exchange rate is a gross index of price differences between European goods and non-European goods. It focuses on two currencies only. Moreover, it does not take account of price differences between French goods and other European goods. Therefore, we use a more informative price index: the real effective exchange rate. It is calculated as a weighted average of French firms’ prices compared to those of their main trade competitors (see Box 3). Weights are the market shares of each country in each sector.

We only have a half-yearly index of this real effective exchange rate. Our time period is therefore quite short, and it is very hard to compare it properly to the quarterly competitiveness index built from the QBS. This stresses the need for further work with larger data.

We plot the inverse of the real effective exchange rate on figure 4. This index relates foreign prices to French prices. It evolves the same way as the competitiveness index built from the QBS. However, its correlation with the competitiveness on the domestic market is not significant (table 5). The correlation is stronger with competitiveness on foreign markets.

Figures 4a, 4b and 4c: Changes in the Inverse of the Real Effective Exchange Rate and Competitiveness

Competitiveness on the domestic market l Competitiveness on the European market Competitiveness on the non EU market

Note: the competitiveness indexes are built from the French Quarterly Business Survey. The inverse of the real effective exchange rate is

calculated by Division de la Prévision et de l’Analyse Economique (Ministry of the Economy, Finance and Industry).

Table 5: Spearman Rank-Order Correlation: Competitiveness Indexes and the Inverse of the Real Effective Exchange Rate

Correlation (p-value)

Competitiveness … On the domestic

market

On the domestic market (quarterly

lag variable)

On the European market

On the non EU market

0.44545 0.32121 0.48182 0.50909 The inverse of the real effective exchange rate (0.170) (0.366) (0.133) (0.110) 11 six-month periods

-0,035

-0,005

0,025

0,055

0,085

0,115

S1/

98S

2/98

S1/

99S

2/99

S1/

00S

2/00

S1/

01S

2/01

S1/

02S

2/02

changes in the inverse of the realeffective exchange rate

changes in the competitivenesson the domestic market

-0,120

-0,070

-0,020

0,030

0,080

S2/

97

S2/

98

S2/

99

S2/

00

S2/

01

S2/

02

changes in the inverse of the realeffective exchange rate

changes in the competitivenesson the Euopean market

-0,175

-0,075

0,025

0,125

S2/

97

S2/

98

S2/

99

S2/

00

S2/

01

S2/

02

changes in the realf effectiveexchange rate

changes in competitiveness onthe non EU market

17

II.3 Labor Productivity is Correlated with Competitiveness

Other indexes than price indexes are commonly used as proxies for competitiveness, e.g. production cost and labor productivity. At the firm level, they are well correlated with competitiveness (see Section 1). We now examine this correlation at the macroeconomic level.

Figures 5 show the evolution of two measures of labor productivity: output per filled job (figure 5a) and value added per hour worked (figure 5b). The output per filled job is the ratio of the output divided by the number of filled jobs. This measure is provided by the Quarterly National Accounts. The output per hour worked is the ratio of Gross Value Added divided by the total actual hours worked in a week, including paid overtime.

Figures 5a and 5b: Changes in Productivity Growth and Competitiveness Output per filled job (growth rate) Gross Value Added per hour worked (growth rate)

The left scale is associated with competitiveness index and the right scale is associated with productivity growth. Source: Insee, QBS and Direction de la Prévision et de l’Analyse Économique (Ministry of Economy)

Both measures for labor productivity evolve more or less the same as the competitiveness index. We observe three periods of increase and two periods of decrease in the output per filled job. They correspond to the three peaks and two drops of the competitiveness indicator. Up to 2000, the Gross Value Added per hour evolves as the competitiveness index but then their evolution differs.

This is confirmed by Spearman rank-order correlations (table 6). Both productivity indexes are significantly and positively correlated with the three competitiveness indicators (on domestic market, on European market, on global market).

However, productivity (whether it is calculated as the output per filled job or as the output per hour worked) may increase because of an increase in the utilization rate of productive capacities: it cannot always be interpreted as the result of technical innovation. In this case, the correlation between competitiveness and our measure for productivity would mainly stem from a correlation to business cycles.

-10%

-5%

0%

5%

10%

15%

Q2/97Q4/9

7Q2/9

8Q4/9

8Q2/9

9Q4/9

9Q2/0

0Q4/0

0Q2/0

1Q4/0

1Q2/0

2

-2,0

-1,0

0,0

1,0

2,0

3,0

4,0

competitiveness on the domesticmarketoutput per filled job

-10%

-5%

0%

5%

10%

15%

Q2/97Q4/9

7Q2/9

8Q4/9

8Q2/9

9Q4/9

9Q2/0

0Q4/0

0Q2/0

1Q4/0

1Q2/0

2

-3,0%

-2,0%

-1,0%

0,0%

1,0%

2,0%

3,0%

4,0%

competitiveness on the domesticmarketvalue added per hour worked

18

Table 6: Spearman Rank-Order Correlation: Competitiveness Indexes and Productivity Growth Correlation (p-value)

Competitiveness On the domestic

market

On the domestic market (quarterly lagged variable)

On the European market

On the non EU market

0.55844 0.36646 0.58329 0.57086 Output per filled job (0.007) (0.094) (0.004) (0.006) 0.36533 0.20909 0.48617 0.47149 Gross Value Added per hour worked (0.095) (0.363) (0.022) (0.027)

Note: Spearman rank-order correlation. 22 observations.

II.4 Absolute Production Costs are not a Good Indicator of Competitiveness whereas Relative Production Costs are

Competitiveness includes other components than labor productivity, e.g. labor costs and production costs. We propose three cost series. The first series are the unit cost of production (figure 6a). It is the ratio of the aggregate production cost (labor cost and intermediate consumption cost, i.e. cost of materials) divided by aggregate production. The second one is the intermediate consumption cost (figure 6b). The last one is the evolution of the average real wage (figure 6c).

Figures 6a, 6b and 6c: Changes in Competitiveness and Production Costs

Unit cost of production (growth rate) Intermediate consumption cost (growth rate) Average real wage (growth rate)

Source : Insee, QBS (competitiveness) and Quarterly National Accounts (cost series)

The three cost series reach a top at the end of 1999, at the same time as the competitiveness index. The intermediate cost series evolve roughly as the competitiveness index. The conclusion is that periods of sharp increase in competitiveness are also periods of increase in costs. This result is very surprising. Indeed, the opposite was expected.

Moreover, this result is confirmed by the Spearman rank-order correlation (table 7). Intermediate consumption costs are positively and significantly correlated with the competitiveness index, although French firms should be all the more competitive as their costs are low. We observe no significant correlation between the average real wage and the competitiveness of French firms.

The unit labor cost is negatively and significantly correlated with competitiveness. However there is nothing new here: it is very likely to stem from the correlation between labor productivity and competitiveness (see previous sub-section). Labor productivity is actually a component of the unit labor cost, since it is the denominator of it.

-0,10

-0,05

0,00

0,05

0,10

0,15

Q2/97Q1/9

8Q4/9

8Q3/9

9Q2/0

0Q1/0

1Q4/0

1Q3/0

2

-1,0

-0,5

0,0

0,5

1,0

1,5

2,0

competitiveness on the domesticmarketintermediate consumptio cost

-0,10

-0,05

0,00

0,05

0,10

0,15

Q2/97Q1/9

8Q4/9

8Q3/9

9Q2/0

0Q1/0

1Q4/0

1Q3/0

2

-0,4-0,20,00,20,40,60,81,01,21,41,6

competitiveness on thedomestic marketaverage wage

-0,10

-0,05

0,00

0,05

0,10

0,15

Q2/97Q1/9

8Q4/9

8Q3/9

9Q2/0

0Q1/0

1Q4/0

1Q3/0

2

-1,0-0,8-0,6-0,4-0,20,00,20,40,60,81,01,2

Competitiveness on thedomestic marketUnit cost of production

19

Tableau 7: Spearman Rank-Order Correlation: Competitiveness Indexes and Growth Rates of Cost Series

Correlation (p-value)

Variable: Competitiveness … On the domestic

market

On the domestic market (quarterly lagged variable)

On the European market On the non EU market

0.00169 0.01299 -0.03331 0.23885 Unit cost of production (0.994) (0.954) (0.883) (0.284) 0.4568 0.45567 0.51101 0.6533 Intermediate consumption cost (0.033) (0.033) (0.015) (0.001)

-0.69848 -0.48391 -0.68041 -0.59684 Labor cost per unit produced (0.000) (0.023) (0.001) (0.003) 0.05364 -0.10446 0.10672 0.35178 Average wages (0.813) (0.644) (0.636) (0.108)

-0.38227 -0.16431 -0.28628 -0.38001 Contributions per unit produces(0.079) (0.465) (0.197) (0.081)

-0.08413 -0.35743 -0.3179 -0.21965 Taxes per unit produced (0.710) (0.103) (0.149) (0.326) 0.65455 0.28485 0.72727 0.59091 “cost competitiveness” (0.029) (0.425) (0.011) (0.056)

Note the “cost competitiveness” is defined in box 3. 22 observations except for cost competitiveness (11 observations)

All these cost series describe the evolution of the absolute production costs faced by French firms. But average production costs may increase without competitiveness decreasing, if average production costs also increase in foreign countries.

Therefore we use a relative production cost index (figure 7). It is similar to the real exchange effective rate, with costs instead of prices. This cost competitiveness index is actually the ratio of a weighted foreign cost index divided by the average production cost in France. Weights are the same as used to build the real effective exchange rate, i.e. market shares of each country in each sector.

This cost competitiveness index is positively and significantly correlated with the competitiveness index built from the QBS (table 7). It means that the higher foreign costs are relatively to French costs, the more likely French firms state that they have an increasing competitiveness.

Figures 7a, 7b et 7c: Changes in Competitiveness and “Cost Competitiveness” Competitiveness on the domestic market Competitiveness on the European market Competitiveness on the non EU market

Note: the “cost competitiveness” index is built as the ratio of a weighted foreign cost index divided by the average production cost in France

(see box 3 for details) Source: Direction de la Prévision et de l’Analyse Économique (Ministry of Economy)

-0,035

-0,005

0,025

0,055

0,085

0,115

S1/

98

S1/

99

S1/

00

S1/

01

S1/

02

Cost competitiveness

Competitiveness on theDomestic Market

-0,120

-0,070

-0,020

0,030

0,080

S2/

97

S2/

98

S2/

99

S2/

00

S2/

01

S2/

02

Cost Competitiveness

Competitiveness on the EUmarket

-0,175

-0,075

0,025

0,125

S2/

97

S2/

98

S2/

99

S2/

00

S2/

01

S2/

02

Cost competitiveness

Competitiveness in non-EUcountries

20

Conclusion

Giving a relevant definition for competitiveness is a difficult task. In the economic literature, definitions vary a lot and are often quite unsatisfactory. This mainly stems from the fact that competitiveness is a very general notion, covering many distinct aspects, such as cost, technology and demand considerations. It seems that there is no way to give a precise definition of competitiveness, that could be linked to standard variables in economics, without making a distinction between the components of competitiveness.

Quarterly Business Surveys do not deal with this conceptual issue. They ask managers to state the way their firms’ competitiveness has been evolving without giving any definition to it. Therefore, it is up to managers to interpret what it means precisely. The measure of competitiveness in QBS is thus highly subjective, but it is a direct measure of the way managers perceive competitiveness.

At the firm level, self-stated competitiveness is correlated with higher sales and production, and lower labor costs and costs of materials. However, raw values of these variables seem to matter, not relative values. This result is counter-intuitive. Given that competitiveness refers to competition, only variables relative to competitors should matter. Nonetheless, it is difficult to explain our results. It may mean that managers misunderstand competitiveness, but it may also mean that they lack immediate information about their competitors, or that only foreign competitors matter, not French competitors.

At the macroeconomic level, we build a “competitiveness indicator” as the balance between positive and negative statements about changes in competitiveness (i.e. “increasing”, resp. “decreasing”). This indicator is fairly well correlated to aggregate productivity or cost indicators. The correlation is weaker with the inverse real exchange rate.

Competitiveness is correlated to raw variables at the firm level and to relative indicators at the macroeconomic level. It may sound paradoxical. However, the differences between the two levels should be stressed. They do not cover the same aspects of competitiveness. At the firm level, the components of competitiveness are identified from individual variability between French firms. At the macro level, they are identified from temporal variability, between the aggregate of French firms relative to the aggregate of foreign firms.

This study mainly aimed at assessing the quality of statements about competitiveness in QBS. These statements seem relevant: they are fairly well correlated with other standard proxies for competitiveness, such as costs and productivity. However, it seems more relevant in the intra-firm (i.e. in the panel dimension) than in the inter-firm dimension. This remark is motivated by the fact that there exists unobserved heterogeneity between firms and that there is a bias in the way managers assess the relative evolutions of their competitors. In microeconomic studies, e.g. public policy evaluation, statements about competitiveness in QBS are complementary, rather than substitutable, to standard proxies for competitiveness. They could be used as dependent variables in order to assess the competitiveness effects of policies, but they should be compared to the effects of policies on costs and productivity.

QBS also provide an aggregate competitiveness indicator that seems quite relevant. Nonetheless, further work is necessary, given the short time dimension in our data.

21

References

Aubert, Patrick and Marie Leclair, 2004, “La compétitivité exprimée dans les enquêtes trimestrielles sur la situation et les perspectives dans l’industrie”, Document de travail Insee-DESE, G2004/01

CPCI, 2002, « Compétitivité industrielle : des inquiétudes pour l’avenir », dans « L’industrie française en 2001/2002 », Rapport de la Commission permanente de concertation pour l’industrie, pp.45-49

Direction de la Prévision, 2002, « Compétitivité des entreprises françaises », dans « Rapport économique, social et financier », pp.162-165

Donzel, F., 1998, « Spécifications de dépouillement des enquêtes mensuelles et trimestrielles sur l’activité dans l’industrie », Note n°236/G123 de la Direction des Études et Synthèse Économiques (Insee)

Gregoir, Stéphane, et Maurel, Françoise, 2003, « Les indices de compétitivité des pays : interprétations et limites », in « Compétitivité », Rapport n°40 du Conseil d’Analyse Economique, Michèle Debonneuil et Lionel Fontagné, 2003

Krugman, Paul, 1994, « Competitiveness: a dangerous obsession », Foreign Affairs, vol 73, n°21, March/April

Lorenzi, Jean-Hervé, 2002, « L’intégration européenne, moteur de la compétitivité française », dans « Vingt ans de transformation de l’économie française », Cahiers français n°311, pp.3-10

22

Annex 1: Rate of Response to Questions about Competitiveness in QBS

Questions about competitiveness have been included in QBS from the third quarter of 1997. Our dataset therefore includes all firms that answered this question from 1997:Q1 to 2002:Q4.

The reference unit is a good produced by a firm. Each firm in the dataset produces at least one good. It can produce more than one. In this case, there are as many responses to the survey as there are goods. In particular, firms state the way their competitiveness has been evolving on each good separately. There are 6,380 goods in the survey.

Table 1: Number of quarters in the dataset, for each good Number of quarters

in the dataset Number of goods Percentage Sum of number of

goods Sum of percentages

1 708 11.10 708 11.10 2 525 8.23 1 233 19.33 3 493 7.73 1 726 27.05 4 424 6.65 2 150 33.70 5 386 6.05 2 536 39.75 6 264 4.14 2 800 43.89 7 270 4.23 3 070 48.12 8 203 3.18 3 273 51.30 9 215 3.37 3 488 54.67 10 190 2.98 3 678 57.65 11 168 2.63 3 846 60.28 12 165 2.59 4 011 62.87 13 151 2.37 4 162 65.24 14 155 2.43 4 317 67.66 15 170 2.66 4 487 70.33 16 166 2.60 4 653 72.93 17 176 2.76 4 829 75.69 18 192 3.01 5 021 78.70 19 195 3.06 5 216 81.76 20 235 3.68 5 451 85.44 21 312 4.89 5 763 90.33 22 617 9.67 6 380 100.00

(Each firm can be in the QBS dataset for up to 4 different goods). Maximum: 22 quarters.

The rate of response is quite good. For 4,759 goods (74,6 % of the total 6,380), respondents answer the question about competitiveness on the French market for all quarters. Only 3,3 % of respondents do never answer the question about competitiveness. There are slightly more missing data for competitiveness on the EU and non-EU market.

Table 2: Missing data for question on competitiveness Competitiveness on the French market

Always missing At least one quarter with

non-missing data All observations 212 6168

At least one quarter with missing data 212 1409 Competitiveness on the

French market Never missing 0 4759 Always missing 212 762

At least one quarter with missing data 0 5406

At least one quarter with missing data 212 3424

All questions about competitiveness (three

markets)

Never missing 0 2744 All observations = 6,380 goods, in the dataset for at least one and at most 22 quarters.

23

Among the 212 goods for which statements about competitiveness are always missing, 125 are only present in the dataset for one quarter only and 183 are present for less than 3 quarters.

Table 3: Competitiveness on the French and on the EU market Competitiveness one the EU market

Increasing Stable Decreasing Missing Total

Increasing

1 877 2.94 27.88 44.57

2 610 4.08 38.77 8.01

583 0.91 8.66 6.17

1 662 2.60 24.69 9.39

6 732 10.53

Stable

1 976 3.09 4.35 46.92

28 430 44.46 62.62 87.24

4 719 7.38 10.39 49.96

10 276 16.07 22.63 58.07

45 401 71.0

Decreasing

317 0.50 4.13 7.53

1 396 2.18 18.17 4.28

4 093 6.40 53.27 43.33

1 878 2.94 24.44 10.61

7 684 12.02

Missing

41 0.06 0.99 0.97

153 0.24 3.71 0.47

51 0.08 1.24 0.54

3 880 6.07 94.06 21.93

4 125 6.45

Competitiveness one the French

market

Total 4 211 6.59

32 589 50.97

9 446 14.77

17 696 27.68

63 942 100

Number of observations: 63 942 goods*quarters In each box, figures represent: number of observations, percentage of total, row percentage, column percentage.

Respondents state that their competitiveness has been stable much more often than increasing or decreasing. However, there remains some va riability between competitiveness on different markets. In particular, there are observations with competitiveness increasing (resp. decreasing) on the French market while decreasing (resp. increasing) on the EU market. Similarly, there are observations with competitiveness increasing (resp. decreasing) on the EU market while decreasing (resp. increasing) on the non-EU market.

24

Annex 2: The Data

Datasets

Our dataset merges three files:

Quarterly Business Surveys in manufacturing industries (« enquêtes trimestrielles sur la situation et les perspectives dans l’industrie ») aim at knowing managers’ opinion about their firm’s activity results and prospects. Firms are interrogated on a quarterly basis, in January, April, July and October (see Box 1).

The DADS (Déclarations Annuelles de Données Sociales) is an administrative dataset built out of employers’ mandatory reports. It covers all workers in all firms within the private and semi-public sector. We use a version of this dataset that is aggregated at the firm level. It includes information about the structure of the workforce in firms, e.g. shares of occupations, age groups, women, full time workers, interns, etc.

The BRN file (Bénéfices Réels Normaux) consists of firms’ balance sheets and is collected by the tax administration. It is very rich and provides information on sales, exports, production, value-added, fixed assets, labor costs, costs of material, etc. Moreover, although it is not exhaustive, it includes all large firms and covers about 95 % of the total of sales in the French economy. From this dataset, we can therefore calculate the aggregate of costs, production, sales, etc. for various sectors (defined by the NAF 700 classification).

Aggregation from quarterly to yearly variables

QBS are performed on a quarterly basis, whereas the DADS and the BRN are yearly data. In order to merge all datasets, we first aggregated the data from QBS into yearly data.

There may be many problems arising from this « aggregation ». We therefore perform robustness tests to check whether results are robust to the aggregation method (see Annex 3).

A limit comes from the availability of data. Competitiveness is observed in QBS from 1997:Q3. The BRN are available until 2001 and the DADS are available from 1994 to 2000. This leaves us with only three year of observations: 1998, 1999 and 2000. However, we also ran regressions including only variables from the QBS and from the BRN. This provides us an additional year of observation (2001).

QBS are performed in January, April, July and October. Thus there may be two ways to aggregate quarters into yearly data: we can consider that the last quarter of the year corresponds to the October survey or to the January survey of the following year. Since most firms fill the reports that help building the DADS and the BRN file with information that describe the situation of the firm at December 31st, it seems more natural to choose the January survey as corresponding to the last quarter of the year. This survey is performed in the first weeks of January and statements therefore correspond to change in competitiveness between early or mid October the previous year to early or mid January. This period seems to us the closer to the last quarter of the year (i.e. October, November, December). In practice and for instance, we consider that yearly changes from 1998 to 1999 should be built out of the 1998 April, July and October surveys, plus the 1999 January survey.

25

Table 4: Merging datasets Dataset Number of observations

QBS (1998 to 2001) 7 513 BRN (1998 to 2001) 12 396 DADS (1998 to 2000) 6 664

QBS merged with BRN (1998 to 2001) 7 233 QBS merged with BRN and DADS (1998 to 2000) 5 780

An observation correspond to a good*year.

Statements about competitiveness are qualitative, trimodal variables (competitiveness can be either increasing, stable or decreasing). Therefore, aggregation is an uneasy task, since we have no information about the scale of variations. In particular, the evolution of competitiveness in firms with both an increase and a decrease within the same year is unknown. Fortunately, there are very few such ambiguous cases with both increases and decreases in competitiveness during the same year. We therefore eliminate these cases from the dataset.

Once these observations have been eliminated, firms in the dataset only have a combination of « stable » and « increasing » quarters, or a combination of « stable » and « decreasing » quarters, or four quarters with competitiveness remaining stable. We can therefore define the yearly change in competitiveness as « increasing » if there is at least one quarter with an increase in competitiveness, « decreasing » if there is at least one quarter with a decrease, and « stable » otherwise.

This definition is quite restrictive. In particular, we do not use the information about the number of quarters with an increase (or a decrease) in competitiveness, although it could give information about the size of the increase. In order to perform robustness tests, we also construct other variables capturing yearly changes in competitiveness.

- Specification (1): 3 modes: equals 1 if there is at least one quarter with an increase (and all other quarters with stable competitiveness), 0 is competitiveness remains stable all over the year, and -1 if there is at least one quarter with an decrease (and all other quarters stable).

- Specification (2): 3 modes: equals 1 if there is at least two quarters with an increase (and all other quarters with stable competitiveness), 0 is competitiveness remains stable all over the year, and -1 if there is at least two quarters with an decrease (and all other quarters stable).

- Specification (3): 5 modes: equals 2 if there is at least two quarters with an increase, 1 if there is only one quarter with an increase, -1 if there is one quarter with a decrease, -2 if there is at least two quarters with a decrease, and 0 otherwise.

- Specification (4): 9 modes: equals 4 if there are 4 quarters with an increase, 3 if there are three quarters with an increase, etc until -4 (four quarters with a decrease).

Results of the robustness analysis are presented in Annex 3.

26

Table 5: Aggregating Quarterly Data into Yearly Data

Number of observations Competitiveness on the French market

Competitiveness on the non-EU market

QBS (goods*year, 1998 to 2001) 7 513 7 513 Observations with:

At least two quarters with increasing competitiveness (the other ones stable)

745 380

One quarter with increasing competitiveness (the other ones stable)

1 598 872

Always stable 3 805 4 232 One quarter with decreasing competitiveness (the

other ones stable) 1 775 2 093

At least two quarters with decreasing competitiveness (the other ones stable) 903 1 205

Ambiguous cases: Some quarters with an increase, other with a

decrease within the same year) Theses observati ons are eliminated

335 (4.5 %) 316 (4.2 %)

Source: Quarterly Business Surveys in manufacturing industries, 1998 to 2000

27

Annex 3: Robustness Tests

The following table presents results of the same regression performed on our different specifications for the yearly competitiveness variable (see Annex 2).

Table 6: Results of robustness tests Dependent variable (definition of

specification in Annex 2) Specification (1) Specification (2) Specification (3) Specification (4)

Variables Parameter Std Error Parameter Std Error Parameter Std Error Parameter Std Error

Change in log total sales 1.025 (0.249) 1.181 (0.288) 1.087 (0.244) 1.057 (0.236)

Change in log total cost of materials -0.646 (0.222) -1.030 (0.265) -0.725 (0.218) -0.744 (0.216)

Change in log total labor cost -1.193 (0.241) -1.500 (0.284) -1.234 (0.235) -1.210 (0.228)

Change in log production 1.407 (0.385) 1.980 (0.453) 1.512 (0.380) 1.534 (0.374)

Change in log employment 0.497 (0.194) 0.612 (0.225) 0.508 (0.189) 0.476 (0.181)

Change in log exports -0.046 (0.035) -0.080 (0.043) -0.053 (0.035) -0.056 (0.035) Exchange rate €/$ * share of exports in

total sales -1.178 (0.943) -1.800 (1.172) -1.262 (0.938) -1.335 (0.936)

Observations 7005 7005 6694 6694

Percentage of:

Concordant pairs 59 62 59 32

Discordant pairs 41 38 42 68

Log likelihood (constant) 13963 9480 17915 21368

Log likelihood (with explanatory variables) 13740 9245 17664 21117

Change in -2*log likelihood 223 235 250 251 We do not display estimates for constants in this table. Controls are year and industry dummies (NES16 classification), share of exports in total

sales, plus two dummies equaling one if the firm has no export during the year, resp. the previous year.

Specifications are: (1) = 3 modes: at least one quarter with an increase, stable all quarters, at least one quarter with a decrease

(2)= 3 modes: at least two quarters with an increase, stable or at most one quarter with variation, at least two quarters with a decrease (3)= 5 modes: at least two quarters with an increase, one quarter with an increase, stable all quarters, one quarter with a decrease, at least two

quarters with a decrease (4) = 9 modes: number of quarters of with an increase, resp. minus number of quarters with a decrease