Price-cost Margins and Shares of Fixed Factors Jozef Konings, Werner Roeger and Liqiu Zhao∗

∗

Konings: Vives, Faculty of Business & Economics, Katholieke Universiteit Leuven, Naamsestraat 61, 3000 Leuven, Belgium (e-mail: [email protected]); Roeger: DG Economic and Financial Affairs, European Commission, BU-1, 3/163 Wetstraat/Rue de la Loi 200, B-1049 Brussels, Belgium (e-mail: [email protected]); Zhao: Vives, Faculty of Business & Economics, Katholieke Universiteit Leuven, Naamsestraat 61, 3000 Leuven, Belgium (e-mail: [email protected]). We would like to thank Jan De Loecker, Klaus Desmet, Damiaan Persyn, Patrick Van Cayseele, Stijn Vanormelingen, Frank Verboven, Frederic Warzynski and seminar participants at K.U.Leuven and Louvain La Neuve, for useful comments and suggestions. Jozef Konings and Liqiu Zhao gratefully acknowledge the financial support from the research council (LICOS excellence center) and VIVES.

1

Abstract We exploit properties of the primal and the dual Solow residual for estimating price-cost margins using the observed variation in output and input factors. Our approach has a number of advantages over other commonly used methodologies. First, we distinguish between variable and fixed input factors and develop an approach that allows us to estimate both price-cost margins and the shares of fixed factors in production, which are usually not observed in firm or sector level data. We show that ignoring shares of fixed factors in estimating price-cost margins will result in a bias. Second, our approach does not suffer from endogeneity problems caused by productivity shocks that can affect output or input growth. And third, we do not need to rely on price deflators for obtaining real values of output and input factors and hence our approach does not suffer from unobserved price heterogeneity in sectors with differentiated products or in multiple-product firms. Our findings can have important implications for a wide range of papers that have documented the effect of changes in the operational environment of firms, such as trade liberalization and deregulation, on price-cost margins. We illustrate our approach to estimate price-cost margins for Belgian firms operating in manufacturing sectors and compare them with firms that operate in service sectors. While it has generally been claimed that services are less competitive than manufacturing, reflected in higher estimated price-cost margins, we show that once fixed input factors are taken into account, price-cost margins of firms in services are comparable to those in manufacturing. Keywords: Price-cost margin;

Shares of fixed inputs;

residual JEL Classification: L11, L13, L60, L80

2

Solow

1

Introduction

The economic implications of institutional change, trade liberalization and anti-trust policy on market power have been widely conjectured and researched. One prominent concern has been how to estimate market power. When very detailed market-level data on product prices, quantities sold, product characteristics and consumer demand information for narrowly defined sectors are available, structural demand models as introduced by Rosse (1970), Just and Chern (1980) and Bresnahan (1981, 1982, 1989), have proven to be a useful tool, in particular for policy simulations, like merger simulation. The estimates of price-cost margins obtained from these structural demand approaches, however, seem rather high, for example, 0.504 for the food processing industry (Lopez, 1984), 0.40 for the railroads industry (Porter, 1983) and between 0.10 and 0.69 for the car industry (Goldberg, 1995; Verboven, 1996). However, the detailed data requirements for structural demand estimation are often not available, yet from a policy point of view it can still be important to have some assessments of how market power is affected when institutional changes, such as deregulation or trade liberalization take place. To this end, alternative approaches, inspired by Hall (1986), making use of the observed variation in output and input factors, have been developed and applied to assess the impact of changes in the operating environment of firms on markups. For instance, Levinsohn (1993), Harrison (1994), Konings and Vandenbussche (2005) use this framework to analyze the effects of trade policy on price-cost margins, Konings, Cayseele and Warzynski (2005) analyze the effects of privatization and Kee and Hoekman (2007) show how competition policy has had an impact on markups. But the estimated price-cost margins in these approaches are usually also rather high, mostly ranging between 0.10 and 0.35. 3

Largely overlooked in this literature, however, has been the issue of how price-cost margins are affected by the presence of fixed costs. This is not surprising as fixed costs of production are usually not observed, while in the structural demand models mostly marginal cost shifters are used for identification. However, changes in the environment in which firms operate cannot only affect price-cost margins, but also the scale of operations and hence the share of fixed costs in production. Not taking into account fixed costs when estimating markups by focusing on prices and marginal cost is an important shortcoming of these methods which limits their practical use in antitrust applications for example. Lindenberg and Ross (1981) as well as Elzinga and Mills (2011) point to the problem that a deviation between price and marginal cost does not necessarily indicate the “degree of monopoly” but could also arise from the need to cover fixed costs. A few papers have analyzed price-cost margins allowing for the unobservable adjustment costs in capital. Klette (1999) points out that due to the adjustment costs in capital the marginal product of capital does not equal its user cost. He develops a model in which capital does not fully adjust to its equilibrium value and applies this to estimating price-cost margins using Norwegian plant level data. He obtains relatively low price-cost margins, ranging between 5 and 10 percent. De Loecker (2011b) and De Loecker and Warzynski (2011) extend Klette’s work in which the markup is obtained as the ratio of the output elasticity of variable input (e.g. material) to the share of input’s cost in turnover. The output elasticity is obtained from estimating a production function, using a control function like the one proposed by Olley and Pakes (1996) or Levinsohn and Petrin (2003), which does not require measuring the adjustment cost of capital. However, both of the approaches suffer from unobserved price heterogeneity by using deflated sales as a proxy 4

for physical quantity which may lead to a bias in the estimation of the output elasticity (Klette and Griliches, 1996; De Loecker, 2011a) and hence in the estimated markup1 . Another concern with the control function approach is the monotonicity condition between intermediate inputs and the unobserved productivity shock, which can break down if an increase in productivity results in a substantial increase in the markup, by such an amount that it would result in a decrease in the firm’s input usage (Levinsohn and Melitz, 2006). In fact, the typical variation in the firm level markups using this approach is quite substantial, due to the high heterogeneity that exists between firms in terms of cost shares and estimated output elasticities. As in De Loecker and Warzynski (2011), we start in this paper, from the framework introduced by Hall (1986, 1988, 1990) and further extended by Roeger (1995). But while De Loecker and Warzynski (2011) use the control function approach to deal with the endogeneity problem, we exploit the properties of the primal and dual Solow residual to solve the endogeneity of output and input growth to productivity shocks. In addition, we allow for the joint estimation of price-cost margins and shares of fixed factors in production, using standard production data on expenditures of inputs and revenue at the firm level. An important advantage of our approach is that, unlike most other approaches, we do not need to rely on unobserved product price data for deflating firm level sales or deflating input factors such as material costs. In addition, our estimation methodology allows not only for the flexible treatment of capital (either fixed, variable or both) but also for the flexible 1

De Loecker and Warzynski (2011) argue that this is likely to bias the output elasticities downwardly as the correlation between inputs and prices is expected to be negative. This would mean that the markups are underestimated. However, when input prices reflect quality differences as documented by Kugler and Verhoogen (2008), for instance, in the case that intermediate inputs are imported, the correlation could go the other way.

5

treatment of other input factors, like labor. And it allows solving the endogeneity problems concerning productivity shocks. To this end, we only need to make two mild assumptions, i.e., constant returns to scale in the variable input factors of production and the variable capital adjusts freely to changes in the user cost of capital. We show that ignoring fixed costs of production results in a bias in the estimates of price-cost margins. Hence, when evaluating changes in the policy environment that firms face, both the impact on price-cost margins and the impact on shares of fixed factors need to be taken into account. We illustrate this in the context of the ‘Service Directive’ of the European Union aimed at reducing administrative entry barriers and increase cross border service flows2 , which was inspired by the observation of high estimated markups in service sectors3 . Using our approach, however, we show that, once we take into account fixed costs in production, competitive pressure, as measured by price-cost margins, is comparable for firms in services and manufacturing. Our approach also explains why the difference between primal and dual Solow residuals with revenue weights is so strongly correlated with the difference between primal and dual Solow residuals with cost weights. In the absence of fixed costs, the former contains a markup component, while the latter equals zero, suggesting that there is no correlation between these differences. As shown below the presence of fixed costs can in principle explain the high correlation between these differences, because in both differences there appears a component related to fixed costs. In our empirical analysis we ex2

The ‘Services Directive’ was adopted in 2006 aiming to further enhance the ‘Single Market’ for services by reducing the barriers to cross-border trade, principally by doing away with the service industry regulations of individual EU Member States. 3 See Siotis (2003), Christopoulou and Vermeulen (2008), Martins, Scarpetta and Pilat (1996) for example.

6

ploit this relationship. The paper is organized as follows. The next section summarizes Hall’s and related approaches and then extends them to allow for fixed input factors. Section 3 describes the firm level data used and presents and discusses the empirical results. Section 4 provides the applications of this approach to the service sectors. Section 5 gives some concluding remarks and discussions.

2 2.1

Methodology The Hall and related approaches

Consider a production function Qit = F (Kit , Lit , Mit )Θit for firm i at period t, where Kit , Lit and Mit are capital, labor and material inputs, respectively, Θit is an index of technical progress. Under the assumptions of perfect competition and constant returns to scale, the Solow residual is given by SRQR it ≡ ∆qit − PMM Wit Lit it Lit ∆lit − Pitit Qitit ∆mit −(1− W Pit Qit Pit Qit

−

MM Pit it )∆kit Pit Qit

and captures the growth rate

of total factor productivity (Solow, 1957), where ∆qit , ∆lit , ∆mit and ∆kit are the growth rates of output, labor, material and capital inputs, respectively. Wit Lit Pit Qit

and

MM Pit it Pit Qit

are the shares of labor cost and material cost in turnover,

respectively. By relaxing the condition that price equals marginal cost, Hall (1988) shows that the Solow residual can be decomposed into a markup and a productivity factor: SRQR it = Bit (∆qit − ∆kit ) + (1 − Bit )∆θit where Bit is the price-cost margin defined as Bit ≡ related to the markups via µit = 1/(1 − Bit ).

7

Pit −M Cit , Pit

(1) which is directly

This approach has been used to obtain an estimate of the average pricecost margin. An estimate of Bit significantly larger than 0 rejects the model of perfect competition. This approach has been used in many papers using industry level or firm level data (e.g. Domowitz, Hubbard and Petersen, 1988; Waldmann, 1991; Morrison, 1992; Levinsohn, 1993; Norrbin, 1993; Harrison, 1994; Basu and Fernald, 1997; Klette, 1999; Konings, Cayseele and Warzynski, 2001). An important problem, however, in estimating equation (1) is that unobserved productivity shocks may be positively correlated with output (or input) growth. Thus instrumental variables are required to estimate Bit , but it has often turned out to be difficult to find good instruments, especially when firm level data are used. In addition, when the impact of policy changes is analyzed, not only price-cost margins may be affected, by also productivity and productivity growth (Harrison, 1994), which can bias the estimated change in price-cost margins. Finally, in equation (1) deflated sales are used to proxy for physical output, but with firm heterogeneity and multiple-product firms, this can introduce a bias (see De Loecker, 2011a; De Loecker and Warzynski, 2011). One way to deal with these problems is the extension suggested by Roeger (1995), who first derives the dual price-based Solow residual by solving the cost minimization problem. In particular, the dual Solow residual is: SRPitR ≡

P M Mit M Wit Lit PitM Mit Wit Lit ∆wit + it ∆pit + (1 − − )∆rit − ∆pit Pit Qit Pit Qit Pit Qit Pit Qit

= −Bit (∆pit − ∆rit ) + (1 − Bit )∆θit (2) where ∆pit , ∆wit , ∆pM it and ∆rit are the growth rates of product price, wage, material price and the rental price of capital, respectively. By subtracting the dual from the primal Solow residual unobserved productivity shocks cancel 8

out and the average price-cost margin Bit can be estimated consistently using equation (3), in which we added an i.i.d. error term it . R SRQR it − SRPit = B[(∆qit + ∆pit ) − (∆kit + ∆rit )] + it

(3)

Equation (3) has the additional advantage that all variables are expressed in nominal terms, so that price deflators are not required for estimating Bit consistently. In equation (3), however, all factors of production are fully flexible and adjust to the equilibrium values, which is not realistic. Typically, there are substantial overhead costs which are fixed and not directly attributable to the short-run variations in output. We therefore adjust our framework to allow for fixed factors of production, which will allow us to estimate both price-cost margins and the shares of (unobserved) fixed input factors in a consistent way.

2.2

Primal and Dual Solow Residuals in the Presence of Fixed Factors of Production

2.2.1

Primal and Dual Solow Residuals with Revenue-based Shares

We start from a standard ‘short run’ production function with constant returns to scale in the variable factors for firm i at period t: Qit = F (Kitv , Lvit , Mit )Θit

(4)

where output Qit is produced with variable capital Kitv ≡ Kit − Kitf , variable labor input Lvit ≡ Lit − Lfit and material Mit . This implies that fixed capital and fixed labor input does not directly enter into the ‘short run’ production function. We implicitly assume that material inputs are fully flexible and

9

that firms are price-takers in their input markets4 . Θit is the unobserved productivity term. Kitv (Lvit ) is the fraction of total capital (labor) which is adjusted to current demand and cost changes without friction. While Kitf (Lfit ) is the type of capital (labor) which is not adjusted within a period to current demand and cost changes. We can interpret the fixed shares of capital and labor input as overhead costs, such as the rent for buildings or the salaries of executives, etc.(Stigler, 1939). In typical firm level dataset there is information on the total amount of labor or capital used, but no explicit distinction can be made between the variable and fixed component of capital (labor). Let svitk and svitl denote the share of variable capital Kitv /(Kitv + Kitf ) and the share of variable labor input Lvit /(Lvit + Lfit ), respectively, which capture the production technology that firms apply but are unobservable to the econometrician. The shares svitk and svitl are firm specific and can vary over time. As will be shown below, the average shares can be estimated in the model simultaneously. Under imperfect competition, the first order condition and Euler’s law imply that the output growth is determined by a weighted sum of the input growth and the growth rate of productivity. Input weights are given by the corresponding shares of variable costs in revenue adjusted by markups. Pit ∆qit = M Cit where



 svitk Rit Kit svitl Wit Lit v PitM Mit v ∆kit + ∆lit + ∆mit + ∆θit (5) Pit Qit Pit Qit Pit Qit

kR K svit sv l W L it it , Pitit Qitit it Pit Qit

and

MM Pit it Pit Qit

are shares of variable capital cost, variable

labor cost and material cost in turnover, respectively. Constant returns to scale implies that the total variable cost is Citv = M Cit · Qit = svitk Rit Kit + svitl Wit Lit + PitM Mit . 4

Crepon, Desplatz and Mairesse (2010) extend Hall (1988)’s approach relaxing the condition that the labor market is perfectly competitive. For applications of this approach see Dobbelaere (2004), Dobbelaere and Mairesse (2008).

10

As in Hall (1988), the primal Solow residual with revenue-based shares is defined as SRQR it ≡ ∆qit −

P M Mit Wit Lit PitM Mit Wit Lit ∆lit − it ∆mit − (1 − − )∆kit (6) Pit Qit Pit Qit Pit Qit Pit Qit

Substituting equation (5) into (6), we get the primal Solow residual with revenue-based shares. (1 − svitl )Wit Lit (1 − svitl )Wit Lit ∆kit − ∆lit + Pit Qit Pit Qit svitk Rit Kit sv l Wit Lit (∆kitv − ∆kit ) + it (∆litv − ∆lit ) + (1 − Bit )∆θit Pit Qit Pit Qit (7)

SRQR it =Bit (∆qit − ∆kit ) +

Following Roeger (1995), we apply the dual price-based Solow residual to eliminate the growth rate of productivity in equation (7). The dual cost minimization problem gives Citv =

M) G(Wit ,Rit ,Pit Qit Θit

which corresponds to the

production function (4). So the marginal cost is M Cit =

M) G(Wit ,Rit ,Pit . Θit

Loga-

rithmic differentiation of marginal cost and using Shepard’s lemma gives: Pit ∆pit = M Cit



svitl Wit Lit PitM Mit M svitk Rit Kit ∆rit + ∆wit + ∆pit Pit Qit Pit Qit Pit Qit

 − ∆θit (8)

Substituting equation (8) into the dual Solow residual with revenue-based shares defined as equation (9), we obtain equation (10). SRPitR ≡

Wit Lit PitM Mit Wit Lit P M Mit M ∆wit + it ∆pit + (1 − − )∆rit − ∆pit (9) Pit Qit Pit Qit Pit Qit Pit Qit

SRPitR = −Bit (∆pit −∆rit )−

(1 − svitl )Wit Lit (1 − svitl )Wit Lit ∆rit + ∆wit +(1−Bit )∆θit Pit Qit Pit Qit (10)

By subtracting equation (10) from (7), the growth rate of productivity, (1 −

11

Bit )∆θit , is eliminated. The difference of the primal and dual Solow residual with revenue-based shares is, R SRQR it − SRPit =Bit [(∆qit + ∆pit ) − (∆kit + ∆rit )]+

(1 − svitl )Wit Lit (1 − svitl )Wit Lit (∆kit + ∆rit ) − (∆lit + ∆wit )+ Pit Qit Pit Qit Wit Lit Rit Kit (∆kitv − ∆kit ) + svitl (∆litv − ∆lit ) svitk Pit Qit Pit Qit (11) Equation (11) shows that the difference of the primal Solow residual and the price-based dual Solow residual can be explained by capital (labor) fixity and imperfect competition5 . In contrast to equation (3), four extra terms appear in equation (11) which make the prediction of the direction of the estimation bias of Roeger (1995) approach impossible. The omission of terms Wit Lit (∆kit Pit Qit

+ ∆rit ) and

the omission of

Wit Lit (∆wit Pit Qit

Rit Kit (∆kitv Pit Qit

+ ∆lit ) leads to a downward bias, while

− ∆kit ) and

Wit Lit (∆litv Pit Qit

− ∆lit ) leads to an upward

bias. In addition, equation (11) cannot be used to estimate B, the average share of fixed capital sf k ≡ 1 − sv k and the average share of fixed labor input sf l ≡ 1 − sv l either, as the growth rates of variable inputs ∆kitv and ∆litv are unobservable in the firm level data. And ∆kitv and ∆litv are likely to be positively correlated with the growth rate of output, which may lead to an upward bias in the estimate of the price-cost margins using equation (11). In order to avoid making assumptions about what could be correlated with the unobserved growth rates in variable input factors, we introduce a similar derivation, based on the difference of the primal and dual Solow residual but 5

Shapiro (1987) focuses on the capital fixity to explain why the primal Solow residual is poorly correlated to the dual Solow residual and he uses the marginal product of capital in terms of quantities as the price of capital, while Roeger (1995) stresses the imperfect competition in explaining the difference between the primal Solow residual and dual Solow residual.

12

with cost-weighted shares to eliminate the unobservable terms6 . 2.2.2

Primal and Dual Solow Residuals with Cost-based Shares

Hall (1990) proposes a cost-weighted TFP measure as a way of avoiding the bias caused by imperfect competition. However, in the presence of fixed inputs, the cost-weighted Solow residual captures not only productivity growth but also the fixity of inputs. In this section, we derive cost-weighted primal and dual Solow residuals allowing for the presence of fixed inputs. Similarly, the growth rate of output can be written as a cost-weighted average of the growth rate of variable inputs plus the growth rate of productivity. Input weights are given by the corresponding shares of variable costs in total variable cost. ∆qit =

svitk Rit Kit svitl Wit Lit v PitM Mit v ∆k + ∆lit + ∆mit + ∆θit it Citv Citv Citv

(12)

The primal Solow residual with cost-based shares SRQC it is defined as follows: SRQC it ≡ ∆qit −

Wit Lit P M Mit Rit Kit ∆lit − it ∆mit − ∆kit Cit Cit Cit

(13)

Substituting equation (12) into (13), we have Rit Kit Wit Lit (∆qit − ∆kit ) + (1 − svitl ) (∆qit − ∆lit )+ Cit Cit Rit Kit Wit Lit Cv svitk (∆kitv − ∆kit ) + svitl (∆litv − ∆lit ) + it ∆θit Cit Cit Cit (14)

k SRQC it =(1 − svit )

The dual cost minimization problem implies that the growth rate of price can 6

Roeger and Warzynski (2004) derive a similar estimation equation. However, they assume that the unobservable growth rate of variable capital in equation (11) can be proxied by the growth rate in labor productivity and they do not allow for the quasi-fixed labor input.

13

be written as a cost-weighted average of the growth rate of inputs’ prices minus the growth rate of productivity. ∆pit =

svitl Wit Lit PitM Mit M svitk Rit Kit ∆r + ∆w + ∆pit − ∆θit it it Citv Citv Citv

(15)

The dual Solow residual with cost-based shares is then Rit Kit Wit Lit P M Mit M ∆rit + ∆wit + it ∆pit − ∆pit Cit Cit Cit Rit Kit Wit Lit Cv = −(1 − svitk ) (∆pit − ∆rit ) − (1 − svitl ) (∆pit − ∆wit ) + it ∆θit Cit Cit Cit (16)

SRPitC ≡

By subtracting (16) from equation (14), the growth rate of productivity v Cit

Cit

∆θit is eliminated. The difference of the primal and dual Solow residual

with cost-based shares is, Rit Kit [(∆qit + ∆pit ) − (∆kit + ∆rit )]+ Cit Wit Lit (1 − svitl ) [(∆qit + ∆pit ) − (∆lit + ∆wit )]+ Cit Rit Kit Wit Lit svitk (∆kitv − ∆kit ) + svitl (∆litv − ∆lit ) Cit Cit

C k SRQC it − SRPit =(1 − svit )

2.2.3

(17)

Difference-in-differences Approach

We find that in equation (11) and (17) the unobservable parts are similar except for the denominator. Multiplying both sides of equation (11) by Pit Qit and multiplying both sides of equation (17) by Cit give: R (SRQR it − SRPit )Pit Qit = Bit · Pit Qit [(∆qit + ∆pit ) − (∆kit + ∆rit )]+

(1 − svitl )Wit Lit (∆kit + ∆rit ) − (1 − svitl )Wit Lit (∆lit + ∆wit )+ svitk Rit Kit (∆kitv − ∆kit ) + svitl Wit Lit (∆litv − ∆lit )

14

(18)

C k (SRQC it − SRPit )Cit = (1 − svit )Rit Kit [(∆qit + ∆pit ) − (∆kit + ∆rit )]+

(1 − svitl )Wit Lit [(∆qit + ∆pit ) − (∆lit + ∆wit )]+ svitk Rit Kit (∆kitv − ∆kit ) + svitl Wit Lit (∆litv − ∆lit ) (19) By subtracting equation (18) from equation (19), the unobserved parts svitk Rit Kit (∆kitv − ∆kit ) and svitl Wit Lit (∆litv − ∆lit ) can be cancelled out. R R C (SRQC it − SRPit )Cit − (SRQit − SRPit )Pit Qit = −Bit · Pit Qit [(∆qit + ∆pit ) − (∆kit + ∆rit )]+

sfitk Rit Kit [(∆qit + ∆pit ) − (∆kit + ∆rit )] + sfitl Wit Lit [(∆qit + ∆pit ) − (∆kit + ∆rit )] (20) The left side of equation (20) can be calculated from the firm level data and the parentheses terms on the right side of equation (20) all refer to growth rates of nominal values of output and input factors. Thus, all variables in equation (20) are observable except Bit , sfitk and sfitl , so equation (20) can easily be estimated with firm level data to obtain the estimates of the average pricecost margin B, the average share of fixed capital sf k and the average share of fixed labor input sf l . Since the left-hand side of equation (20) is the difference of the difference of the primal and dual Solow residual with cost-based shares and the difference of the primal and dual Solow residual with revenue-based shares, we call it a “difference-in-differences” (DID hereafter) approach. Notice that the right hand sides of equation (18) and (19) have some common components capturing fixed inputs, which implies a positive correR C C 7 lation between (SRQR it − SRPit )Pit Qit and (SRQit − SRPit )Cit . If a firm’s

markup can exactly cover fixed cost, i.e., no excess markups Bit Pit Qit = 7

In the absence of fixed costs, the difference between the primal and dual Solow residual with cost weights is 0. Thus the correlation between the difference of primal and dual Solow residuals with revenue-based shares and that of primal and dual Solow residuals with cost-based shares is 0.

15

sfitk Rit Kit + sfitl Wit Lit , the right hand side of equation (20) is equal to 0 and R R C thus (SRQC it − SRPit )Cit = (SRQit − SRPit )Pit Qit . In particular, the cor-

relation between the difference of the primal and dual Solow residuals with revenue-based shares and the difference of the primal and dual Solow residuals with cost-based shares is stronger if markups go in the same direction as fixed costs. The advantages of our approach are (i) it solves the endogeneity problem between productivity shocks and growth in output or input factors, (ii)we do not need to rely on price deflators, which solves the problem of unobserved price heterogeneity, particularly for multiple-product firms, (iii)we obtain an estimate of the unobserved shares of fixed input factors. One potential problem, however, is the measurement errors in input factors. Since our model is estimated in first differences, it may exacerbate measurement errors, which leads to a downward bias of the estimates as suggested by Griliches and Hausman (1986) and Griliches and Mairesse (1995). However, this conclusion rests on the classical errors-in-variables model under strict exogeneity, as well as other assumptions. So whether the bias in first differences is larger than that in OLS, or vice versa, is unknown (Wooldridge, 2002).

3 3.1

Data and Results The Data

The data used in this paper are drawn from the Belfirst database collected by Bureau van Dijk. The database includes the full income statements of every Belgian firm that has to report to the tax authorities. We have data of firms active in manufacturing and firms that are active in services and have a panel

16

with observations running from 1999 through 2008. The variables used for the analysis are turnover, number of employees (in full time equivalents), wage bill of full-time equivalents employees, material costs (raw materials, consumables and services) and tangible fixed assets. Our final sample consists of an unbalanced panel of 8,972 firms operating in manufacturing sectors with a total of 43,404 observations and 51,508 firms operating in service sectors with a total of 191,741 observations (See Appendix A for a detailed description of the dataset and cleaning process). Table 1 provides some summary statistics of the main variables. The median manufacturing firm has 15 employees, 0.42 million Euro tangible fixed assets, earns a revenue of 3.44 million Euro and faces staff cost of 0.61 million Euro per year. Firms in service industries seem to be smaller and have a larger variation. The median service firm has 3 employees, 99 thousand Euro tangible fixed assets, earns a revenue of 0.68 million Euro and faces staff cost of 87 thousand Euro per year. [Table 1 about here.] Since we need a measure of the nominal value of the capital stock, we introduce a measure of capital cost. We use the one that is commonly used in literature and is based on Hall and Jorgenson (1967). For capital, we use the book value of the tangible fixed assets. The rental price of capital is calculated by the standard formula R = PI (r − π + δ), where PI denotes the index of investment goods prices, r refers to the nominal interest rate, π is the inflation rate, and δ is the depreciation rate on fixed assets which we assume to be 20% for every sector8 . 8

We experimented with depreciation rates of 15%-25%, but results remained similar.

17

3.2 3.2.1

Results Basic Results

We start by estimating the average price-cost margins for the manufacturing industry as a whole without taking into account the fixed inputs using the traditional Roeger (1995) approach. The results shown in column (1) and (2) of Table 2 indicate that the average price-cost margin is 0.091, suggesting that the average markup is 1.10 for manufacturing firms in Belgium. We report results based on both OLS and firm fixed effects estimators, but the results are virtually the same. In column (3) and (4), we apply our new methodology to estimate the average price-cost margins and the average shares of fixed inputs. The fixed effects model gives similar results as OLS. The average price-cost margin is 0.04 (or a markup of 1.04) which is much lower than the estimate using the traditional approach. The average share of fixed capital is 25% and the average share of fixed labor input is 7.4%. [Table 2 about here.] In order to assess the reliability of the estimates from our approach, we conduct a series of tests, for example, the magnitude of the estimated price-cost margins and the estimated fixed costs, the comparison between the estimation bias predicted from the model and the bias calculated from the data. 3.2.2

Do Profits Cover Fixed Costs?

The estimated price-cost margin using DID approach is much lower than the one when shares of fixed input factors are ignored. This gives rise to one concern: are the profits high enough to cover fixed costs? To address this

18

question, we compare the calculated profits and fixed costs based on the coefb × Pit Qit , ficients estimated. The estimated profit is π cit = (Pit − M Cit )Qit = B dk Rit Kit + sf cl Wit Lit . We define the cit = sf and the estimated fixed cost is F cit )/Pit Qit . excess profit margin as (c πit − F Figure 1 shows the kernel probability density estimates of the excess profit margin. The average excess profit margin is 0.0081 in the sample suggesting that on average firm’s profits can cover fixed costs, and a T-test shows that the average excess profit margin is significantly greater than zero. While around 75% firms make positive excess profits. [Figure 1 about here.] The concentrated excess profit margin suggests a strong correlation between the difference of the primal and dual Solow residuals with revenue weights and the difference of the primal and dual Solow residuals with cost weights. The data shows that this correlation coefficient is 0.82 for manufacR C turing firms, while the correlation between (SRQC it − SRPit )Cit and (SRQit −

SRPitR )Pit Qit is 0.98. 3.2.3

Price-cost Margins, Fixed Costs and Estimation Bias

In this section, we investigate the relations between price-cost margins, fixed costs and the estimation bias through estimating equation (20) by sectors. Table 3 shows the estimates of price-cost margins and shares of fixed inputs for each NACE 2-digit manufacturing sector in Belgium. Column (1) reports the price-cost margins estimated by the traditional approach. Column (2) to (4) show the results estimated by our approach. Here we only focus on the sectors with at least 500 observations in order to obtain statistically reliable results. 19

We find that the average price-cost margins vary across sectors. On average, price-cost margins estimated by Roeger (1995) approach are larger than that estimated by DID approach for all sectors except three – pulp, paper and paper products; other non-metallic mineral products; radio, TV, and communication equipment – with significantly higher shares of fixed labor input. The price-cost margin estimated by DID is positively correlated with that by Roeger (1995) with the correlation coefficient of 0.42. The results in column (2) also show that other non-metallic mineral products, machinery and equipment n.e.c., publishing and printing are among the highest markups sectors. [Table 3 about here.] Regarding the share of fixed capital, the results shown in Table 3 are in line with our expectations, as the share is larger in sectors where we would expect fixed costs to be high. Basic metals (0.8) and motor vehicles (0.80) have the highest shares of fixed capital, followed by wood and wood products (0.63), machinery and equipment n.e.c.(0.54) and chemicals and chemical products (0.39). Turning to the share of fixed labor input, 7 out of 18 sectors have a significantly positive share of fixed labor input. Radio, TV, and communication equipment has the highest share of fixed labor input (0.36). [Figure 2 about here.] dk RK + sf cl W L) would have As expected, sectors with higher fixed costs (sf higher price-cost margins, i.e., price-cost margins increase with the share of fixed costs in turnover. Figure 2 provides strong evidence for it. Sectors with a higher share of fixed costs in turnover are likely to charge higher markups. In addition, we are also interested in the relation between the estimation bias and the shares of fixed inputs. Since all estimates vary across sectors, 20

we are able to have rough pictures of them by looking into the estimates by sectors. As discussed in Section 2, equation (11) suggests that the bias introduced by ignoring fixed costs is negatively correlated with the share of fixed labor input9 . However, the relation between the estimation bias and the share of fixed capital is more complex. There should exist an inverted U-shaped relationship between the estimation bias and the share of fixed capital: when Kit the share of fixed capital is 0, svitk RPititQ (∆kitv − ∆kit ) in equation (11) equals 0, it Kit when the share of fixed capital is increased to be 1, svitk RPititQ (∆kitv − ∆kit ) also it

equals 0, but if the share of fixed capital falls in the range (0, 1), svitk is positive and

Rit Kit (∆kitv −∆kit ) Pit Qit

is positively correlated with [(∆qit +∆pit )−(∆kit +∆rit )]

which leads to an upward bias. The scatter plots of the bias and the share of fixed capital (share of fixed labor input) are shown in Figure 3. The bias is strongly negatively correlated with the share of fixed labor input, which is in line with our expectation. However, the relation between the bias and the share of fixed capital is weak and not clear. [Figure 3 about here.] 3.2.4

Different Production Technology: High-tech Manufacturing Sectors

To examine whether the production technology matters in the estimation of price-cost margins, manufacturing firms are split into high-technology and lowtechnology sectors following Eurostat’s definition of high-technology, medium high-technology, medium low-technology and low-technology according to technological intensity (see Appendix C for details). Table 4 reports the results for 9

In equation (11), sf l is the coefficient of both downward bias terms and a lower sf l implies a lower downward bias. In addition, (1 − sf l ) is also the coefficient of an upward bias term and a lower sf l implies a higher upward bias.

21

high-tech and low-tech sectors separately. The average price-cost margin for high-tech sectors is 0.046, the average share of fixed capital and the average share of fixed labor input are 37% and 10%, respectively. The results indicate that high-tech manufacturing sectors have higher shares of fixed capital and labor input, and charge higher price-cost margins, which are in line with the fact that high-tech sectors have substantial R&D investment (or R&D workers) and thus need to charge higher price-cost margins to cover the investment. On the contrary, low-tech sectors have lower shares of fixed capital and labor input and charge lower price-cost margins. [Table 4 about here.]

4

An Application: Manufacturing Vs. Service Industries

4.1

A Debate over Service Industries

Since the mid 1990s, the productivity gap between Europe and the United States has increased dramatically: GDP per hour worked in the EU has decreased from 98.3 percent of the U.S. level in 1995 to 90.0 percent in 2006. Van Ark, O’Mahony and Timmer (2008) show that the productivity slowdown in European countries is largely the result of slower productivity growth in service sectors, particularly in trade, finance, and business services10 . They further argue that the lack of flexibility and competitiveness in labor and product markets in service sectors in the EU is one of the causes of the trend11 . 10

That is, productivity levels in manufacturing are relatively similar across countries compared to intermediate services. 11 The productivity in service sectors is important for the whole economy. First, services account for 70% of GDP and employment in most EU Member States. Second, competition

22

Desmet and Parente (2010) also suggest that the European service sectors can benefit (productivity gains) from the increase in the competition and spatial concentration in the service sectors. In particular, the network utilities, such as post and telecommunications, air transport, are still highly regulated in Europe. For example, incumbent operators are largely protected from competition in most EU countries through their monopolies or other regulations. So increasing the flexibility in the labor market and strengthening the competition in service product markets within and across countries are claimed to be important to improve productivity growth in European service sectors. Unlike manufacturing, services are less exposed to international competition because of the nature of service exchange and the restrictiveness of services trade policies, suggesting that there is less competitive pressure in service sectors. A number of empirical studies find that service industries have higher markups than manufacturing industries (e.g. Siotis, 2003; Christopoulou and Vermeulen, 2008; Martins, Scarpetta and Pilat, 1996). Nevertheless, since the 1990s, the EU has implemented a series of policies to encourage competition between European service producers so as to strengthen competition and foster efficiency in service sectors, for example, the ‘Single Market’ and ‘Services Directive’. However, despite the great efforts made by the EU Commission, Badinger (2007) shows that the markups in most service industries have even slightly increased since the early 1990s in EU countries. Hence, if the competition is as low as shown by the estimated markups in service sectors, liberalization and deregulation of services should have a pro-competitive effect. However, this crucially relies on the reliability of the estimation of markups. We therefore apply our new methodology to investigate whether in the Service Sector is a major determinant of the performance of manufacturing firms (Francois and Hoekman, 2010).

23

service industries charge high price-cost margins after allowing for the fixed factors of production, i.e., the shares of fixed inputs.

4.2

Price-cost margins in Manufacturing and Service Industries

As shown in Figure 4, the price-cost margins calculated following Collins and Preston (1969) are higher in service sectors12 , especially the knowledgeintensive services (KIS hereafter), than manufacturing sectors (see Appendix D for the definition of KIS and LKIS). There is a large variation in service industries in terms of sales, employment, tangible fixed assets etc.(see Table 1). As the fixity of capital input may matter more in KIS sectors13 , we focus on these sectors in this section. [Figure 4 about here.] The first two columns in Table 5 show the results using the Roeger (1995) approach, suggesting that the estimated price-cost margin in KIS sectors is 0.15, which is almost the double of that in manufacturing sectors (0.091). The last two columns in Table 5 report the results using our approach. The estimated price-cost margin is 0.048, only slightly higher than that in manufacturing sectors (0.040) and similar as that in high-tech manufacturing sectors (0.046). The average share of fixed capital is 0.52, which is much higher than that in the manufacturing sectors (0.25), but the estimate of the share of the fixed labor input is insignificant. 12

The price-cost margin (also referred to as the Lerner Index) is defined as Turnover-Material Cost-Cost of Employees . Turnover 13 The share of fixed capital is likely to be higher in air transport or telecommunications sector than that in hotels and restaurants sector.

24

[Table 5 about here.] The results in Table 5 imply that the price-cost margins of KIS sectors are overestimated by the traditional approach not taking into account of the fixity of capital, which is high in KIS sectors. After taking into account the fixity of inputs, the price-cost margin in KIS sectors is only slightly higher than that in manufacturing sectors. The KIS sectors have relatively higher markups but also a higher share of fixed capital. We therefore check whether KIS sectors have higher excess profit margins compared to manufacturing sectors. As in Figure 1, we depict the kernel probability density estimates of the excess profit margin for service sectors in Figure 5. The average excess profit margin is -0.0094 in the sample and the left hand tail is longer compared to the distribution for manufacturing industry. While around 70% of the firms make positive excess profits. Figure 5 suggests that the KIS sectors do not make higher excess profit margins than manufacturing sectors. [Figure 5 about here.]

5

Conclusions

In this paper, we propose a new methodology to simultaneously estimate the price-cost margins and the shares of fixed inputs extending the work of Hall (1986, 1988, 1990) and Roeger (1995). Our approach has a number of advantages over other commonly used methodologies. In particular, our approach allows an estimate of the share of fixed factors in production, which are usually not observed in sector or firm level data. Furthermore, we do not need to rely on price deflators for obtaining real values of output and input factors 25

and hence our approach does not suffer from omitted unobserved price variable bias, particularly for multiple-product firms. Also our approach does not suffer from endogeneity problems caused by unobserved productivity shocks that can affect output or input growth. Applying our methodology to Belgian firm level data for the period 19992008, we find that the average price-cost margin is 0.040 which is much lower than that using the traditional approach, the average share of fixed capital in total capital usage is 0.25 and the average share of fixed labor input in total labor usage is 0.074 for manufacturing firms in Belgium. We also find that sectors with a higher share of fixed costs in turnover charge higher markups. Finally, we apply our methodology to firms in services and find that the price-cost margins in service industries are overestimated under the traditional framework because of fixed costs. After taking into account the fixity of inputs, the price-cost margin in knowledge-intensive service industries is only slightly higher than that in the manufacturing industry, suggesting that production technologies may cause the high estimated markups found in earlier studies, rather than the lack of competition. The methodology we propose is particularly promising for analyzing the price-cost margins of sectors with high shares of fixed costs, for example, hightechnology sectors and knowledge-intensive service sectors. More interestingly, it can be applied to assess the impact of changes in operating environment of firms on markups as well as production technologies, i.e., the share of fixed inputs.

26

A

Data Cleaning process

The firm-level data used in this paper are provided by Bureau van Dijk. Firm level data often contains outlier observations that may bias the estimated coefficients. Hence, we carefully clean the original dataset to handle the missing observations and outliers. Several cleaning procedures are applied to the sample: • We work with unconsolidated accounts only. • Observations with extreme values (smaller than 1st percentile and larger than 99th percentile) of the variables used in the empirical analysis are dropped, as well as all observations with missing information on some variables and observations that were based on irregular reports or unreasonable data values in the levels of variables (such as negative values of material cost, negative values of staff cost). • We restrict our data to the manufacturing sectors based on NACE Code (15-37) and service sectors based on NACE Rev.1 Sectors G to K and M to O (OECD). This leaves us with 17.74 percent of the registered manufacturing firms, accounting on average for about 46.28 percent of aggregated value added in the manufacturing sector, and 7.55 percent of the registered service firms, accounting on average for about 24.32 percent of aggregated value added in the service sectors.

B

Definition of Variables • P Q = Turnover 27

• P M M = Raw materials, consumables, services and other goods • W L = Staff costs of full-time equivalents employees + benefits in addition to wages • K = Tangible fixed assets • R = PI (r − π + δ), where PI is the index of investment goods prices, r is the nominal interest rate, π is the inflation rate and δ is the depreciation rate (20%). The investment goods price index is taken from the E.U. AMECO database.

C

Definition of High-tech and Low-tech manufacturing sectors [Figure 6 about here.]

D

Definition of KIS and LKIS [Figure 7 about here.]

28

References Badinger, Harald. 2007. “Has the EU’s Single Market Programme Fostered Competition? Testing for a Decrease in Mark-up Ratios in EU Industries.” Oxford Bulletin of Economics and Statistics, 69(4): 497–519. Basu, Susanto, and John G. Fernald. 1997. “Returns to Scale in U.S. Production: Estimates and Implications.” Journal of Political Economy, 105(2): 249–283. Bresnahan, Timothy F. 1981. “Departures from Marginal-cost Pricing in the American Automobile Industry: Estimates for 1977-1978.” Journal of Econometrics, 17(2): 201–227. Bresnahan, Timothy F. 1982. “The Oligopoly Solution Concept is Identified.” Economic Letters, 10(1-2): 87–92. Bresnahan, Timothy F. 1989. “Handbook of Industrial Organization.” , ed. R. Schmalensee and W. Robert, Chapter Empirical Studies of Industries with Market Power, 1011–1057. Amsterdam: North-Holland. Christopoulou, Rebekka, and Philip Vermeulen. 2008. “Markups in the Euro Area and the US Over the Period 1981-2004: A Comparison of 50 Sectors.” ECB Working Paper No. 856. Collins, Norman R., and Lee E. Preston. 1969. “Price-cost Margins and Industry Structure.” The Review of Economics and Statistics, 51(3): 271– 286. Crepon, Bruno, Rozen Desplatz, and Jacques Mairesse. 2010. “Contributions in Memory of Zvi Griliches.” , ed. Jacques Mairesse and Manuel Trajtenberg, Chapter Price-Cost Margins and Rent Sharing: Evidence from 29

a Panel of French Manufacturing Firms, 583–610. Annales D’Economie et de Statistique, 79-80, July-December 2005. De Loecker, Jan. 2011a. “Product Differentiation, Multi-Product Firms and Estimating the Impact of Trade Liberalization on Productivity.” Econometrica. forthcoming. De Loecker, Jan. 2011b. “Recovering Markups from Production Data.” International Journal of Industrial Organization, 29(3): 350–355. De Loecker, Jan, and Frederic Warzynski. 2011. “Markups and FirmLevel Export Status.” American Economic Review. forthcoming. Desmet, Klaus, and Stephen L. Parente. 2010. “Bigger is Better: Market Size, Demand Elasticity and Innovation.” International Economic Review, 51(2): 319–333. Dobbelaere, Sabien. 2004. “Estimation of price-cost margins and union bargaining power for Belgian manufacturing.” International Journal of Industrial Organization, 22(10): 1381–1398. Dobbelaere, Sabien, and Jacques Mairesse. 2008. “Panel Data Estimates Of The Production Function And Product And Labor Market Imperfections.” NBER Working Paper No. 13975. Domowitz, Ian, Glenn R. Hubbard, and Bruce C. Petersen. 1988. “Market Structure and Cyclical Fluctuations in U.S. Manufacturing.” The Review of Economics and Statistics, 70(1): 55–66. Elzinga, Kenneth G., and David E. Mills. 2011. “The Lerner Index of Monopoly Power: Origins and Uses.” American Economic Review: Papers & Proceedings, 101(3): 558564. 30

Francois, Joseph, and Bernard Hoekman. 2010. “Service Trade and Policy.” Journal of Economic Literature, 48(3): 642–692. Goldberg, Pinelopi Koujianou. 1995. “Product Differentiation and Oligopoly in International Markets: The Case of the U.S. Automobile Industry.” Econometrica, 63(4): 891–951. Griliches, Zvi, and Jacques Mairesse. 1995. “Production Functions: The Search for Identification.” NBER Working Paper No. 5067. Griliches, Zvi, and Jerry A. Hausman. 1986. “Errors in variables in panel data.” Journal of Econometrics, 31(1): 93–118. Hall, Robert E. 1986. “Market Structure and Macroeconomic Fluctuations.” Brookings Papers on Economic Activity, 17(2): 285–338. Hall, Robert E. 1988. “The Relation between Price and Marginal Cost in U.S. Industry.” Journal of Political Economy, 96(5): 921–947. Hall, RobertE. 1990. “Growth, Productivity, Unemployment: Essays to celebrate Bob Solow’s Birthday.” Chapter Invariance Properties of Solow’s Productivity Residual, 71112. Cambridge, MA: MIT Press. Hall, Robert E., and Dale Jorgenson. 1967. “Tax Policy and Investment Behavior.” American Economic Review, 57(3): 638–655. Harrison, Ann E. 1994. “Productivity, imperfect Competition and Trade Reform.” Journal of International Economics, 36(1-2): 53–73. Just, Richard E., and Wen S. Chern. 1980. “Tomatoes, Technology, and Oligopsony.” The Bell Journal of Economics, 11(2): 584–602.

31

Kee, Hiau L., and Bernard Hoekman. 2007. “Imports, entry and competition law as market disciplines.” European Economic Review, 51(4): 831–858. Klette, Tor J. 1999. “Market Power, Scale Economies and Productivity: Estimates from a Panel of Establishment Data.” Journal of Industrial Economics, 47(4): 451–476. Klette, Tor J., and Zvi Griliches. 1996. “The Inconsistency of Common Scale Estimators When Output Prices Are Unobserved and Endogenous.” Journal of Applied Econometrics, 11(4): 343–361. Konings, Jozef, and Hylke Vandenbussche. 2005. “Antidumping Protection and Markups of Domestic Firms.” Journal of International Economics, 65(1): 151–165. Konings, Jozef, Patrick Van Cayseele, and Frederic Warzynski. 2001. “The dynamics of industrial mark-ups in two small open economies: does national competition policy matter?” International Journal of Industrial Organization, 19(5): 841–859. Konings, Jozef, Patrick Van Cayseele, and Frederic Warzynski. 2005. “The Effects of Privatization and Competitive Pressure on Firms’ PriceCost Margins: Micro Evidence from Emerging Economies.” The Review of Economics and Statistics, 87(1): 124–134. Kugler, Maurice, and Eric A. Verhoogen. 2008. “Product Quality at the Plant Level: Plant Size, Exports, Output Prices and Input Prices in Colombia.” Discussion Paper No. 0708-12, Columbia University. Levinsohn, James. 1993. “Testing the Imports-as-Market-Discipline Hypothesis.” Journal of International Economics, 35(1-2): 1–22. 32

Levinsohn, James, and Amil Petrin. 2003. “Estimating Production Functions Using Inputs to Control for Unobservables.” Review of Economic Studies, 70(2): 317–341. Levinsohn, Jim, and Marc J. Melitz. 2006. “Productivity in a Differentiated Products Market Equilibrium.” Mimeo, Princeton University. Lindenberg, Eric B, and Stephen A Ross. 1981. “Tobin’s q Ratio and Industrial Organization.” Journal of Business, 54(1): 1–32. Lopez, Ramon E. 1984. “Measuring Oligopoly Power and Production Responses of the Canadian Food Processing Industry.” Journal of Agricultural Economics, 35(2): 219–230. Martins, Joaquim Oliveira, Stefano Scarpetta, and Dirk Pilat. 1996. “Mark-up Pricing, Market Structure and the Business Cycle.” OECD Economic Studies, , (27): 71–105. Morrison, Catherine J. 1992. “Markups in U.S. and Japanese Manufacturing: A Short-Run Econometric Analysis.” Journal of Business & Economic Statistics, 10(1): 51–63. Norrbin, Stefan C. 1993. “The Relation between Price and Marginal Cost in U.S. Industry: A Contradiction.” Journal of Political Economy, 101(6): 1149–1164. Olley, Steven G., and Ariel Pakes. 1996. “The Dynamics of Productivity in the Telecommunications Equipment Industry.” Econometrica, 64(6): 1263–1297. Porter, Robert H. 1983. “A Study of Cartel Stability: the Joint Executive Committee, 1880-1886.” Bell Journal of Economics, 14(2): 301–314. 33

Roeger, Werner. 1995. “Can Imperfect Competition Explain the Difference between Primal and Dual Productivity Measures? Estimates for U.S. Manufacturing.” Journal of Political Economy, 103(2): 316–330. Roeger, Werner, and F. Warzynski. 2004. “Estimating Markups and Productivity Growth in EU Manufacturing.” Working paper 04-17, Department of Economics, Aarhus School of Business. Rosse, James N. 1970. “Estimating Cost Function Parameters without Using Cost Data: Illustrated Methodology.” Econometrica, 38(2): 256–275. Shapiro, Matthew D. 1987. “Are Cyclical Fluctuations in Productivity Due More to Supply Shocks or Demand Shocks?” American Economic Review, Papers and Proceedings, 77(2): 118–124. Siotis, Georges. 2003. “Competitive Pressure and Economic Integration: an Illustration for Spain, 1983-1996.” International Journal of Industrial Organization, 21(10): 1439–1459. Solow, Robert M. 1957. “Technical Change and the Aggregate Production Function.” The Review of Economics and Statistics, 39(3): 312–320. Stigler, George. 1939. “Production and Distribution in the Short Run.” Journal of Political Economy, 47(3): 305–327. Van Ark, Bart, Mary O’Mahony, and Marcel P. Timmer. 2008. “The Productivity Gap between Europe and the United States: Trends and Causes.” Journal of Economic Perspectives, 22(1): 25–44. Verboven, Frank. 1996. “International Price Discrimination in the European Car Market.” The RAND Journal of Economics, 27(2): 240–268.

34

Waldmann, Robert J. 1991. “Implausible Results or Implausible Data? Anomalies in the Construction of Value-Added Data and Implications for Estimates of Price-Cost Markups.” Journal of Political Economy, 99(6): 1315– 1328. Wooldridge, Jeffrey M., ed. 2002. “Econometric Analysis of Cross Section and Panel Data.” , ed. Jeffrey M. Wooldridge. MIT Press, Cambridge, MA.

35

Table 1: Summary Statistics for the Sample of Belgian Firms Industry

Variable

Mean

Manufacturing

Turnover 16,467.17 Employment 57.06 Tangible fixed assets 2,368.98 Material Costs 12,871.05 Wage bill 2,637.24

Service

Turnover Employment Tangible fixed assets Material Costs Wage bill

4,676.62 14.45 774.17 3,872.67 561.73

Median

Std. Dev.

3,444 15 416 2,254 605

56,049.13 354.98 8,989.84 47,476.05 8,051.71

681 3 99 448 87

21,322.77 506.72 5,836.93 18,710.85 2347.21

Note: Turnover, tangible fixed asset, material cost and wage bill are expressed in thousands of Euro.

36

Table 2: Estimation of Price-cost Margins and Shares of Fixed Inputs (Manufacturing) Roeger (1) OLS (∆p + ∆q) − (∆k + ∆r)

∗∗∗

0.091 (0.0041)

DID

(2) FE

(3) OLS

(4) FE

∗∗∗

0.091 ( 0.0043) ∗∗∗

∗∗∗

−P Q[(∆p + ∆q) − (∆k + ∆r)]

0.041 (0.0040)

RK[(∆p + ∆q) − (∆k + ∆r)]

0.27 ( 0.062)

W L[(∆p + ∆q) − (∆k + ∆r)]

0.072 (0.022)

0.074 (0.023)

∗∗∗

∗∗∗

0.040 (0.0044) ∗∗∗

0.25 (0.067) ∗∗∗

Year dummy

Yes

Yes

Yes

Yes

Industry dummy

Yes

Yes

Yes

Yes

Observations Number of Firms R2

44,760 8,972 0.17

44,760 8,972 0.17

43,404 8,972 0.13

43,404 8,972 0.12

Note: Robust standard errors adjusted for clustering at the 4-digit industry level are in parentheses. *** significant at the 1 percent level. ** significant at the 5 percent level. * significant at the 10 percent level.

37

Table 3: Estimates of Price-cost Margins and Shares of Fixed Inputs by Sectors Nace 2-Digit

Description

15

Food and Beverages

16

(1) PCM(Roeger) ∗∗∗

(2) PCM(DID) ∗∗∗

(3) sf k ∗

(4) sf l

(5) Obs. ∗∗

0.10 (0.0095)

0.018 (0.0059)

0.24 (0.15)

Tobacco

0.066 (-)

0.067 (-)

-2.25 (-)

0.52 (-)

101

17

Textiles

0.062 (0.015)

0.082 (0.22)

0.12 (0.11)

2,161

18

Wearing apparel; fur

0.057 (0.0050)

19

Leather and leather products

0.12 (0.031)

20

Wood and wood products

0.097 (0.0035)

21

Pulp, paper, and paper products

0.061 (0.023)

22

Publishing and printing

0.11 (0.019)

23

Coke, refined petroleum products, nuclear fuel

0.045 (0.0055)

24

Chemicals and chemical products

0.083 (0.0096)

25

Rubber and plastic products

0.086 (0.011)

26

Other non-metallic mineral products

0.082 (0.013)

27

Basic metals

0.090 (0.016)

28

Fabricated metal products

0.095 (0.0065)

29

Machinery and equipment n.e.c.

0.096 (0.0076)

30

Office machinery and computers

0.039 (0.0057)

31

Electrical machinery and apparatus n.e.c.

0.098 (0.018)

32

Radio, TV, and communication equipment

0.081 (0.029)

33

Medical, precision, and optical instruments

0.092 (0.0098)

34

Motor vehicles

0.059 (0.0082)

35

Other transport equipment

0.11 (0.019)

36

Furniture/manufacturing n.e.c.

0.070 (0.011)

37

Recycling

0.10 (0.0056)

∗∗∗

∗∗∗

∗

∗∗∗

∗∗

∗∗∗

∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗

∗∗∗

∗∗∗

∗∗

∗∗∗

∗∗∗

∗∗

Note: Robust standard errors adjusted for clustering at the 4-digit industry level are in parentheses. *** significant at the 1 percent level. ** significant at the 5 percent level. * significant at the 10 percent level.

38

∗∗

0.034 (0.012) ∗∗∗

0.032 (0.0068) ∗∗∗

0.087 (0.00056) ∗

0.078 (0.034) ∗∗∗

0.076 (0.015)

∗∗∗

0.083 (0.020)

∗∗

-0.0026 (0.00023) ∗∗∗

0.039 (0.0098) ∗∗∗

0.055 (0.010)

∗∗∗

0.089 (0.023)

∗∗∗

0.058 (0.0095) ∗∗∗

0.053 (0.012)

∗∗∗

∗∗

-0.57 (0.16)

-0.12 7,540 (0.049)

∗∗

0.20 (0.057)

866

0.21 (0.087)

110

0.64 (0.14)

0.073 (0.17)

1,599

0.14 (0.27)

0.30 (0.18)

945

0.27 (0.13)

0.23 (0.13)

4,323

-0.046 (0.18)

-0.13 (0.11)

110

0.39 (0.23)

0.037 (0.098)

2,589

0.41 (0.22)

0.11 (0.055)

0.32 (0.26)

0.23 (0.071)

∗∗∗

1.82 (0.073) ∗∗∗

∗

∗

∗∗∗

∗

∗∗∗

1,781 2,953

∗∗

0.80 (0.22)

0.15 1,298 (0.065)

0.18 (0.20)

0.12 (0.050)

∗∗

∗∗

6,970

∗∗∗

0.084 (0.0089)

0.54 (0.24)

0.21 3,145 (0.064)

0.056 (0.042)

-0.45 (1.72)

0.28 (0.44)

270

0.058 (0.024)

0.10 (0.46)

0.18 (0.10)

1,074

0.086 (0.038)

-0.38 (0.70)

0.36 (0.079)

702

0.087 (0.045)

0.37 (0.75)

0.26 (0.17)

1,039

-0.0096 (0.012)

0.80 (0.14)

-0.11 (0.033)

548

0.049 (0.027)

0.17 (0.77)

0.024 (0.082)

355

0.025 (0.0095)

-0.040 (0.22)

0.011 2,385 (0.042)

0.030 (0.014)

0.17 (0.047)

-0.19 (0.25)

∗∗

∗∗

∗∗

∗∗

∗

540

Table 4: Different Production Technology: High-tech & Low-tech Low-tech (1)OLS ∗∗∗

High-tech

(2)FE ∗∗∗

(3)OLS ∗∗∗

(4)FE ∗∗∗

−P Q[(∆p + ∆q) − (∆k + ∆r)]

0.038 (0.0043)

RK[(∆p + ∆q) − (∆k + ∆r)]

0.27 (0.065)

W L[(∆p + ∆q) − (∆k + ∆r)]

0.046 (0.025)

0.058 (0.026)

0.12 (0.048)

0.10 (0.048)

Year dummy

Yes

Yes

Yes

Yes

Industry dummy

Yes

Yes

Yes

Yes

Observations Number of firms R2

33,840 7,105 0.14

33,840 7,105 0.13

9,564 1,867 0.11

9,747 1,867 0.11

∗∗∗

∗∗

0.038 (0.0048) ∗∗∗

0.22 (0.071) ∗∗

0.048 (0.0095) ∗∗

0.31 (0.14) ∗∗

0.046 (0.0099) ∗∗∗

0.37 (0.16) ∗∗

Note: Robust standard errors adjusted for clustering at the 4-digit industry level are in parentheses. *** significant at the 1 percent level. ** significant at the 5 percent level. * significant at the 10 percent level.

39

Table 5: Price-cost Margins in Knowledge-intensive services (KIS) Roeger (1)OLS (∆p + ∆q) − (∆k + ∆r)

∗∗∗

0.15 (0.0088)

DID

(2)FE

(3)OLS

(4)FE

∗∗∗

0.15 (0.0092) ∗∗∗

∗∗∗

−P Q[(∆p + ∆q) − (∆k + ∆r)]

0.048 (0.0079)

RK[(∆p + ∆q) − (∆k + ∆r)]

0.52 (0.088)

0.52 (0.094)

W L[(∆p + ∆q) − (∆k + ∆r)]

-0.0067 (0.024)

-0.0060 (0.025)

∗∗∗

0.048 (0.0077) ∗∗∗

Year dummy

Yes

Yes

Yes

Yes

Industry dummy

Yes

Yes

Yes

Yes

Observations Number of Firms R2

67,326 18,592 0.25

67,326 18,592 0.25

63,382 18,498 0.12

63,382 18,498 0.12

Note: Robust standard errors adjusted for clustering at the 4-digit industry level are in parentheses. *** significant at the 1 percent level. ** significant at the 5 percent level. * significant at the 10 percent level.

40

Figure 1: The Kernel Density Estimates of Excess Profit Margin (Manufacturing). Observations below the 5th and above the 95th percentiles of excess profit margin are dropped.

41

Figure 2: Price-cost Margins and the Share of Fixed Costs in Turnover. Numbers denote 2-digit NACE industries.

42

Figure 3: The Estimation Bias and the Shares of Fixed Inputs. Numbers denote 2-digit NACE industries.

43

Figure 4: Sectoral Difference in Lerner Index. Observations below the 5th and above the 95th percentiles of excess profit margin are dropped.

44

Figure 5: Excess Profit Margin Kernel Density Estimates (KIS). Observations below the 5th and above the 95th percentiles of excess profit margin are dropped.

45

Figure 6: Breakdown of NACE manufacturing sectors depending on their technological intensity

46

Figure 7: Breakdown of NACE service sectors depending on their technological intensity

47