Innovation and Advertising : The Role of Competition

Innovation and Advertising : The Role of Competition Philippe Askenazy, Thomas Breda and Delphine Irac 29 mars 2008 Abstract : Advertising and innovation are two engines for firms to escape competition through a cost advantage or a better attraction power toward consumers. We propose a model that encompasses both the static and dynamic interactions between R&D, advertising and competitive environment. This model provides two main predictions. First, for a given competitive environment, technological leaders spend more in advertising than Neck and Neck firms or technological followers in order to extract maximal rents from a double monopolist positions. Second, more competition pushes Neck and Neck firms to advertise more to attract the larger share of consumers on their products or services. Empirical evidence from a large panel of 59,000 French firms over 1990-2004 strongly supports these two properties. 1

1 Introduction Advertising and innovation are two engines for firms to escape competition through a cost advantage or a better attraction power toward consumers. Advertising enables to acquire a reputation by publicizing a better quality, intensity in innovation or even fashionableness of products or services. Aghion et al. (2005) show an escape competition effect of R&D, whereby competition exerts pressure on firms to spend in R&D in order to strengthen their technological and market position. Moreover, advertising and R&D are two intangible assets of firms. Competition appears indeed as an important determinant of firms choices. Though in-depth firm-level empirical investigations are relatively scarce, it seems well established that a more competitive environment induce firms to advertise more. On the other hand, competition should favour R&D through a mechanism of escape competition or the creation of barriers for entry, but when it is too harsh, it challenges incentives to innovate. A rich literature has shown that both R&D and advertising enhance profits (see Bagwell 2005 for a review on advertising). The latter has clear positive short-run consequences. However, there are still controversies on the relative returns of these two intangible investments and their interplay. Favouring advertising may induce a substitution and thus a reduction of the R&D effort. On the contrary, it is not necessarily possible to disentangle these two types of investments. Advertising and R&D may be complements. Advertising should be associated with improving quality, since a famous firm is reluctant to lose its reputation by offering an odd or outdated product (Fogg-Meade 1901). As shown by Grossmann (2007), advertising also increases sunk costs and makes entry more difficult. This in turn induces higher market concentrations with larger firms and enhances R&D investments (since R&D is more profitable to large firms that are able to spread R&D costs over higher sales). Advertising should be more efficient if the firm proposes innovative or less costly goods or services (Nelson 1974, Fluet 2002). Advertising should also improve the information of consumers on the true quality of firm output. More generally, advertising may generate short-term rents that help to finance long-run investments. These interpretations provide explanation for the high advertising spending in R&D intensive sectors, like drugs. We propose a model that encompasses both the static and dynamic interactions between R&D, advertising and competitive environment. The model is composed of two 2

blocks, a static one and a dynamic one. In a given sector, we consider two firms that compete on a market composed of a continuum of consumers. Two shares of the latter have a predilection for the production from each firm. The lower these shares, the lower the differentiation, the higher the proportion of undecided consumers and then the larger the room for price competition between the duopolists. The two firms could use costly advertising to convince undecided consumers. The sector is either levelled - both firm are technologically Neck-and-Neck and thus faces similar production costs - or unlevelled - one firm being a technological leader with lower production costs than the follower. In order to introduce a dynamic trade-off between innovation and predilection advantages for firms, this first block of the model is plugged to a standard framework à la Aghion et al. (2005). It allows us to endogenise the relationships between competition and advertising R&D decisions. This model provides two main predictions. First, for a given competitive environment, technological leaders spend more in advertising than Neck and Neck firms or technological followers in order to extract maximal rents from a double monopolist positions. There is thus a static complementarity between past R&D efforts that stochastically determines the innovation position of the firms. Second, more competition pushes Neck and Neck firms to advertise more to attract the larger share of consumers on their products or services. More generally, endogenising the state of a sector levelled versus unlevelled exhibits a positive relationship between competition toughness and advertising expenditures. But the relation between competition and R&D is inverted-u shape. For high competitive environments, current R&D may be statistically negatively correlated but this negative correlation hinges only upon changes in competition and is not due to substitution effect in the firm decisions on R&D and advertising efforts. Using a large unbalanced panel of around 59,000 French firms over the 1990-2004 period, we test most of these assertions. The Centrale des Bilans database from the Banque de France provides very detailed data on firm performance and firm expenditures or investments including R&D and advertising. Within sectors, most productive firms seem to spend more in advertising. Similarly, current advertising spending are positively correlated to past R&D efforts. These results are consistent with dynamic complementarity between R&D and advertising. Estimations also support the monotonic impact of competition on advertising. This article is organized as follows. Section 2 lays out the basic theoretical static frame- 3

work. Section 3 introduces the dynamic R&D process and study the impact of advertising costs on the flows of innovation. Our main predictions are then derived analytically from this model. Section 4 provides a description of the data and presents our main empirical findings. Section 5 concludes with directions for further research. 2 Static theoretical framework This section presents a static theoretical framework to capture the basic connections between competition and advertising for a technological level of firms. The dynamic interaction with R&D will be dealt with in section 3. 2.1 Basic market structure We consider markets as duopolies with firms A and B producing differentiated goods or services. The market can be in a neck-and-neck situation where there is no technological gap between A and B or in an unlevelled situation where a technogical leader (say A) and a follower (say B) coexist. In the leader-follower case, the leader enjoys a technological gap reducing its production cost by a factor 1 ɛ, where ɛ is given parameter. Then, we assume that ɛ also represents the ex ante valuation advantage firms have on specific consumers. These consumers have an initial preference for the goods from A or B. For example, young adults will prefer trekking with friends in Nepal while retired will favour tour operators. Adults between 25 and 65 are initially neutral. Other examples include the wine vs. beer US market of alcohol. Indeed recent Gallup polls show that upperclass male Americans that are above 45 and very fond of European culture give a prominent place to wine whereas less well-to-do and younger drinkers favour beer. In-between these two categories, people may be classified as indifferent. Similarly, artistic professions favour Mac computers whereas scientific professions are more inclined to buy PC, with neutral users in between. Segmentation of consumers can also come from geographic constraints, e.g. customers prefer to buy in stores located in their neighbourhood. We formalize this ex ante inclination of consumers by the existence of segments of captive consumers. The size of these segments is inversely proportional to the degree of competition. To escape competition on the non captive segments, firms can advertise to attract a share of the initially neutral consumers. 4

2.2 Consumers 2.2.1 Without advertising Consider a continuum of consumers of mass one indexed by i. Their utility follows : u i = 1 0 ln x ij dj where x j is the aggregate of two outputs A and B from two firms on the market j defined by : x ij = v i (x Aj, x Bj ) where v symmetric and homogenous of degree 1. A and B are imperfect substitutes : x ij = (1 + ɛ) k ij/2 x Aj + (1 + ɛ) k ij/2 x Bj (1) On each market, consumers are aligned on the segment [0,1] by increasing order of preference for good B. The fraction f j [0, 1/2] of ex ante non indifferent consumers is defined such as : 1 if i I A = [0, f j ] k ij = 0 if i I 0 =]f j, 1 f j [ 1 if i I B = [1 f j, 1] For simplicity we suppress the notation for industry index from here on. The log-preference assumption made in the first equation implies that individuals spend the same amount on each basket x j. We normalize this common amount to unity by using expenditure as numeraire for the prices p Aj and p Bj at each date. Thus each consumer i chooses x Aj and x Bj to maximise v i (x Aj, x Bj ) subject to the budget constraint : p Aj x Aj + p Bj x Bj = 1. Due to the special form of v, the demand function facing firm A is : x Aj = 1 if p Aj /p Bj < 1 + ɛ 1/2 if p Aj /p Bj = 1 + ɛ 0 if p Aj /p Bj > 1 + ɛ The demand function facing firm B is trivially obtained by inverting A and B in the expression above. 5

2.2.2 With advertising Advertising is viewed as a mean of modifying consumers preferences by affecting their marginal rate of substitution. More formally, let a A (resp a B ) denote the set of consumers who receives ads from firm A (resp firm B). The utility function of a consumer i is still given by equation 1 with k i now given by : k i if ki { 1, 0, 1} k i = 1 if ki < 1 1 if ki > 1 where k i = 1 {i IA } 1 {i IB } + 1 {i aa } 1 {i ab }. As ɛ is small, we work at first order terms in ɛ from herein. Consequently : (1 + ɛ) 2 = 1 + 2ɛ, 1/(1 + ɛ) = 1 ɛ and 1/(1 + ɛ) 2 = 1 2ɛ. Chart 1 sums up the MRS depending on if a consumers received ads from firms A and B. Fig. 1 Advertising and consumers valuations Ads received from none or both firms Ads received from A 2.3 Firms : equilibrium prices and profits 2.3.1 Without advertising Firms use labour as the only input, according to a constant-return production function, and take the wage rate as given. Thus the unit costs of production c A and c B of the two firms in an industry are independent of the quantities produced. Firms are supposed to be able to set differentiated prices for the various segments of customers. For example, price promotion for family or students or free delivery for faraway consumers. Due to the perfect substitution between the two goods of each industry in the utility function of consumers, duopolies compete in prices for each consumer, arriving at a Bertrand equilibrium. We now derive the explicit form of prices and profits depending on the technology configuration of the market. 6

a) Levelled sector In this case, firms are neck-and-neck and production costs are equal. On I 0, firms will trivially set their price equal to their cost c = c A = c B. On I A, firm A will use its comparative advantage and choose the maximum price such as firm B cannot steal the market I A from A without making a negative profit. That is, p A,IA = c (1 + ɛ) η where η is infinitely small 1. Firm B acts on I B as A on I A and gets this market. The infinitesimal profit made on each i [0, 1] is Π A,i = p A,i x A,i c A x A,i. Given that p A,i x A,i = 1 if i chooses to buy firm A s good, Π A,i = 1 c A p A,i if i chooses to buy firm A s good. Finally the overall profit flow of firms A and B in the neck and neck case is : Π 0 = f(1 1 1 + ɛ ) = fɛ b) Unlevelled sector In this second case, one of the two firms is leader and has a cost advantage equal to ɛ, that is, the ratio of the other firm cost to its cost is equal to 1 + ɛ. Without loss of generality, we assume that when the sector is unleveled, firm A is the leader and firm B the follower. Again, on each segment, firm A will use its comparative advantage and choose the maximum price such as firm B cannot steal him the market without making a negative profit. Hence, firm A prices p A,IA = c B (1 + ɛ) = c A (1 + ɛ) 2 = c A (1 + 2ɛ), p A,I0 = c B = c A (1 + ɛ) and p A,IB = c B /(1 + ɛ) = c A. Firm A and B share the segment I B (since the MRS is equal to the ratio of prices) and do not make any profit on this segment. The follower total profit Π 1 is equal to 0 and the leader s total profit is : Π 1 = f(1 1 1 ) + (1 2f)(1 ) = 2fɛ + (1 2f)ɛ = ɛ 1 + 2ɛ 1 + ɛ 2.3.2 With advertising Firms are now given the opportunity to advertise their product. We suppose that firms can target their ads, that is, they can choose the consumers who will receive it, for example they choose medias, TV programs that are devoted to particular segments of consumers. For touching subsegments of consumers of size a, they incur a quadratic cost φa 2. In addition, firm are assumed to target first ex ante neutral consumers that are adjacent to their captive segment, e.g. consumers with a close age -the 25-35 year old if the youth are captive- or living in a neighbour city. 1 for simplicity, similar infinitesimal terms will be ignore in the reminder of this section 7

The equilibrium prices, profits depends on the amount of advertising realised by each firm which is function of the cost of advertising φ. In the reminder of this section, we first state two lemma about advertising strategies by firms before solving completely the model. Lemma 1. Firms never advertise on their captive segment, that is, firm A never advertises on I A and firm B never advertises on I B. Proof : See appendix B Lemma 2. There is no overlapping in advertising, that is firms A and B have no interest to advertise toward the same consumers. Proof : See appendix B Now that the general pattern of advertising strategies has been described, we turn to the resolution of the model for φ = ɛ. The general results are then quickly given for φ = ɛ/2 and, to make the presentation lighter, advertising strategies for general values of φ are presented in chart 2. From Lemma 1 and 2, firms advertise only when advertising shifts consumers preferences and enables to make additional positive profit. In these cases, the additional profit generated for a firm generated by advertise on a subsegment of size a is : Π a = ɛa φa 2 (2) First order condition of equation?? gives, if it exists, the optimal interior size of targeted subsegment a = ɛ/2φ with associated profit Π a = ɛ 2 /4φ. For tractability, we exhibit solutions for a benchmark case ɛ = φ, that corresponds to a moderate cost of ads. a) ɛ = φ : Levelled sectors Neck-and-neck firms only advertise on the segment I 0 (see proof of Lemma 1 and 2 for details) i.e. a share 1 2f of the consumers. Since optimal interior advertising is equal to 1/2 (ɛ = φ) and firms advertise to different consumers, the segment I 0 will be completly concerned. From the no overlapping property, the symmetric equilibrium advertising strategies (a A, a B ) = (1/2 f, 1/2 f). Profits are then Π 0 = ɛf + ɛ(1/2 f) ɛ(1/2 f) 2 = 8

ɛ(1/4 + f f 2 ). b) ɛ = φ : Unleveled sectors Recall that the follower never advertises. The leader can advertise up to a fraction (1-f) of consumers including ex ante neutral consumers and captive ones of the follower. Since 1/2 1 f for all f [0, 1/2], the leader s advertising target is the interior optimal level 1/2. Associated total profits for the leader and the follower are Π 1 = 0 and Π 1 = ɛ + ɛ/4 = 5ɛ/4. So advertising expenditures non decline with the technological advantage of firms and advertising expenditures non decrease with competition. Actually, these results also hold when φ ɛ. When φ > ɛ, the interior optimal level of advertising is lower and the upper bounds 1/2 f for a neck-an-neck firm and 1 f for a leader firm are becoming less restrictive. Hence the leader s advertising will keep being equal to its optimal level ɛ/2φ < 1 f and its profit will remain unchanged. However, neck-and-neck firms advertising will deviate from its saturating level 1/2 f for small values of f. Similarly, when φ < ɛ, the interior optimal level of advertising is higher and the upper bounds for neck-and-neck firms and leader firms become more restrictive. Hence, neck and neck firms keep on their constrained level 1/2 f and their profit remain unchanged. However, leader firms deviate from their previously unconstrained level of advertising ɛ/2φ and set at the saturation level 1 f for large value of f. For φ < ɛ/2, the optimal unconstrained level of advertising is greater than 1 ; thus the upper bound is always binding, even for small values of f and leader s advertising is always equal to 1 f. Chart 2 sums up these results by picturing the leader s and Neck-and-neck s advertising levels when f varies between 0 and 1/2. In words, the static theoretical model predicts : Property 1 : Advertising expenditures increase with the technological advantage of firms, that is, a L a N a F where L,N and F stand for leader, neck-and-neck and follower. Intuitively, because it faces lower production costs, the technological leader has in- 9

Fig. 2 Firms advertising level for different costs of advertising terest to try to capture both ex ante neutral and unfavourable segments. In addition it does not face the advertising competition of its competitor. So it advertises more than neck-and-neck firms for a given level of competition f. Property 2 : For a given state of the sector (levelled or unlevelled) advertising expenditures are increasing with competition, that is, a L, a N and a F are decreasing with f. Intuitively, when competition is tougher i.e. non captive market is large, all firms try to escape competition through an increase in their advertising effort. But computing the aggregated levels of advertising for different degrees of competition and so to determine the relation between advertising and competition requires to determine the proportion of levelled and unlevelled sectors for a given degree of competition. 3 Dynamics of R&D investment and innovations Firm can develop dynamic strategy to escape competition through becoming a technological leader and thus innovation. We thus have to introduce the dynamic of R&D investment. The standard Aghion et al. (2005) framework is adopted, because it naturally generates unleveled and neck and neck sector. We assume the main subcase of? : knowledge spill-overs between leader and follower are such that the maximum sustainable technological gap is 1, leading to a maximal relative cost advantage ɛ. If a firm is one step ahead and it innovates the follower will 10

automatically copy the leader s previous technology and so remain only one step behind. Formally, the technology level of a duopoly firm in some industry j is denoted by k, that is, one unit of labour currently employed by firm X generates an output flow equal to : A X = (1 + ɛ) k h, X = A, B The state of an industry is then fully characterized by a pair of integers (l, m) where l is the leader s technology and m = 1 is the technology gap of the leader over the follower ; m = 0 when firms are neck and neck. We define by π 1 (respectively π 1 ) to be the equilibrium profit flow of a firm m step ahead of (respectively behind) its rival 2. The R&D cost of firm moving one technological step ahead with a Poisson hazard rate of n is n 2 /2. We call n the innovation rate or R&D intensity of the firm. We assume that follower firm can move one step ahead with hazard rate h even if it spends nothing on R&D, by copying the leader s technology. Thus n 2 /2 is the R&D cost of a follower firm moving ahead with a hazard rate n + h. 3.1 Bellman equations We now derive general equations for R&D investments. Let V denote the steady state value of the firm. We have the following Bellman equations : rv 1 = π 1 + (n 1 + h)(v 0 V 1 ) n 2 1 /2 rv 1 = π 1 + (n 1 + h)(v 0 V 1 ) n 2 1 /2 rv 0 = π 0 + n 0 (V 1 V 0 ) + n 0 (V 1 V 0 ) n 2 0 /2 The annuity value rv 1 of currently being a technological leader in an industry with gap 1 at date t equals the current profit flow π 1 minus the current R&D cost δn 2 1 /2, minus the expected capital loss (n 1 + h)(v 0 V 1 ) from having the follower catch up with the leader. Similar arguments leads to equations for the value of a follower and a neck and neck firm. Given that profitability is only dependant on the gap between leader and follower, no innovation will be undertaken by the leader i.e. n 1 = 0. Now, using the fact that each firm 2 As proved in the previous section, the profit in the industry depends only on the gap m between two firms and not on absolute levels of technology 11

chooses its own R&D intensity to maximize its current value, i.e. to maximize the RHS of the corresponding equation, we obtain the first order conditions : n 1 = V 0 V 1 n 0 = V 1 V 0 n 1 = 0 According to these first order conditions, as an increase in market competition diminishes profits of a levelled firm, and consequently its market value, V 0 decreases and one could expect an increase of n 0 and a decline in n 1. Equations (1) and (2) solve for n 1 and n 0. Eliminating the V s between these equations yields the reduced form R&D equations : n 2 0 2 + (r + h)n 0 (π 1 π 0 ) = 0 n 2 1 2 + (r + h + n 0)n 1 (π 0 π 1 ) n2 0 2 = 0 This system is recursive, as the first equation solves for n 0, and then given n 0 the second equation solves for n 1.We obtain : n 0 = r h + (r + h) 2 + 2(π 1 π 0 ) (3) n 1 = (r + h + n 0 ) + (r + h + n 0 ) 2 + n 2 0 + 2(π 0 π 1 ) (4) Using equation (3) to substitute (r + h + n 0 ) 2 in equation (4) yields the alternative expression : n 1 = (r + h + n 0 ) + (r + h) 2 + n 2 0 + 2(π 1 π 1 ) (5) The R&D investment n 0 of a neck and neck firm is increasing in (π 1 π 0 ) : the larger the difference between neck and neck firms and leader firms profit flows, the larger the incentive for a neck and neck firm to become a leader and thus the larger its R&D investment. Interpretation of equation 5 is less straightforward. First, n 1 is increasing in (π 1 π 1 ) : the larger its incentive to become a leader (represented by (π 1 π 1 )), the greater the follower s R&D investment. But it requires two successful investments for the follower to become a leader, and its profit in the intermediate situation of Neck and Neck should also matters. This is captured by the presence of n 0 in equation 5. It is easy to show that n 1 is decreasing in n 0 and thus decreasing in (π 1 π 0 ) : the smaller π 0, the 12

larger (π 1 π 0 ) and the smaller the follower s R&D investment. The innovation rate of a sector is 2n 0 if the sector is levelled and n 1 if the sector is unlevelled. In addition, the average innovation rate of a sector in steady state also depends of the fraction of time a sector spends being levelled or unlevelled. Formally, let µ 1 (resp. µ 0 ) denote the steady state probability of being an unlevelled (resp. neck and neck) industry. During any unit time interval, the steady state probability that a sector moves from being unlevelled to levelled is µ 1 n 1, and the probability that it moves in the opposite direction is 2µ 0 n 0. In steady state, these two probabilities must be equal : µ 1 n 1 = 2µ 0 n 0 This together with the fact that µ 1 + µ 0 = 1 implies that the average flow of innovation is : I = µ 0 2n 0 + µ 1 n 1 = 2µ 1 n 1 = 4n 0n 1 2n 0 + n 1 (6) 3.2 Relationships between competition, innovation and advertising Profit flow of firms A and B calculated in section 2 depend on the degree of competition f, the ratio 1 + ɛ of valuation for goods A and B for a consumer in [0, f] and the ratio λ of firms A and B costs. If A is leader, λ = 1 ɛ and if A and B are neck and neck, λ = 1. We thus get the profits π 1, π 0 and π 1 as functions of f, ɛ and γ. As a consequence, I is a function of exogenous parameters f, ɛ, r and h. As in Aghion et al. (2005), the general form of I is a U-inverted shape. The escape competition effect dominates when competition is not too harsh. Figure 3 illustrates this property when r = 0.04, h = 2.5, ɛ = 0.25. 3 It shows the plot of I as a function of f [0, 0.5]. We also plot advertising as a function of f. It exhibits a positive relation between competition and advertising effort. Actually this property is general : Property 3 : In the dynamic framework, advertising expenditures are still increasing with competition, that is, a L, a N and a F are decreasing with f. 3 h has been chosen according to (Aghion et al., 2005). ɛ is set such that average markup of prices over costs in the model is roughly equal to the average markup observed in the economy in the past two decades. 13

Fig. 3 Average sectoral R&D and advertising investments as a function of competition R&D investment Advertising investment Proof : See appendix B. So as in the static environment, competition enhances advertising. Theoretically, if the competitive environment is harsh (f small), we may observe that a firm facing an even more competition reduces current R&D and increases current advertising. But, this mechanism driven by competition does not mean that advertising and R&D are substitute. Actually, the static results still hold : for a given competitive environment, innovative firms advertise more and firms innovate more when advertising is possible. 4 Empirical evidences 4.1 Data We use a subset of the FIBEN dataset provided by the Observatoire des enterprises at the Banque de France. Data from FIBEN are collected on a voluntary basis. Clerks in the different local establishments of the Bank of France contact firm to complete a survey. The Fiben database is based on firms tax forms and includes all businesses with more than 500 employees and a fraction of smaller firms. It covers about 57% of employment for manufacturing but less for service sectors. A subset of FIBEN, the so called Centrale des Bilans contains more detailed information on firms expenditures that are specifically devoted to increase their potential sales, with two special items on advertising and R&D 14

expenditures 4. The clear value-added of these micro data compared to other sources on R&D is to include firms that have episodic R&D and advertising activities or novel firms and to provide in the same time their advertising efforts. R&D can be considered either as expenditure or as investment in the French legal accounting setting. Broadly speaking, R&D costs concerning a well defined project and yielding almost certain return can be declared as investments whereas R&D expenditures linked to more uncertain projects have to be considered as current expenditures. In this paper, we add these two categories together. A Lerner index for each firm can be built using these data. We only observe sectoral price provided by the INSEE, but we have detailed information on costs. The Lerner index is supposed to measure the market power of the firm by the difference between price and marginal costs (which equals the negative inverse of demand elasticity). Since neither price nor marginal costs are available at the firm level, we compute the index using value-added net of depreciation and provisions minus the financial cost of capital (cost of capital*capital stock) over sales (in line with Aghion et al., 2005). The Fiben database contains very detailed balance sheet information that enables to compute these Lerner indicators. 5 In our model the Lerner index is decreasing with f, the measure of competition. We also compute a total factor productivity index (TFP) for each firm defined as the [to be completed] We finally have lerner index, total R&D and marketing expenditures available for an unbalanced panel of 59 thousands firms from 1990 to 2004. This final sample contains around 480,000 firm-year observations, the number of firms present each year is around 30,000 and is relatively stable over time. In average a firm is observed in our sample during around 7 years. Appendix A2 shows some descriptive statistics. Unsurprisingly, retail trade, food industry and consumer goods exhibit high levels of advertising (more than 2000 euros per employee) ; whereas high level of R&D are observed in cars, equipment goods and energy 4 These items have a precise counterpart in the official accounting plan (plan comptable général). Advertising comes from category 623, whereas R&D expenditures are the sum of elements in categories 61, 62 and 64 5 Lerner=(value added-depreciation-cost of capital.capital stock-provision)/sales Using the standard mnemonics of French tax forms : Lerner=[VA-(AQ+AS+AU+AW+AY- AQ-1-AS-1- AU-1-AW-1-AY-1)-0.085.capital-(DR-DR-1)]/FL. 15

sectors. Two sectors have very high level of advertising compared to R&D : food industry, which partly reflects the downstream margin effects, and real estate. 4.2 Results Raw statistics are consistent with the model. Figure 4 plots the average of R&D and advertising efforts as a function of the 20-ciles of firm lerners. R&D effort appears U- shaped in the measure of competition whereas average advertising is clearly increasing with competition. The similarity with figure 3 is striking. Fig. 4 Competition and firms advertising and R&D investments in France Source : FIBEN/Centrale des Bilans Two key predictions of the model can also be statistically tested. First, our model shows that there is an increasing relationship between advertising and competition (see figure 1). Second, as chart 2 summarizes, the technological leader always has a level of advertising that is higher than the rest of the firms. Two types of estimations are applied on the panel data : a simple robust OLS with controls for year, sector and firm size, a firm fixed effect estimator with robust coefficients. Table A tests the first prediction. Two alternative specifications are used : The measure of the inverse of competition is either a firm level lerner or a sector level lerner. To cope with simultaneity issue, the firm level index is taken alternatively contemporanous or lagged. Advertising effort is measured as the ratio of advertising in 2004 euro over the number of employees in the firm. Controlling for years and for firm size, the regression 16

coefficient between advertising and the inverse of competition is clearly negative. Note that these results are not driven by large firms since they still hold when restricting the sample to firms below 250 employees. Tables B and C test the second prediction. In order to identify potential leaders, we first make the reasonable assumption that leaders enjoy a better total factor productivity (TFP). Table B shows that higher TFP (coincident or lagged) is correlated with higher advertising. Assuming that the technological position can also be described by cumulative past R&D efforts, we build a rough proxy for a R&D stock by adding R&D expenditures over the past 4 years. The average R&D stock is around 950 euros per employee. 17

Table C shows that higher lagged R&D stock per employee is correlated with higher advertising. Unsurprisingly, the relationship between current R&D stock and advertising is much more blurred, since this former includes current R&D expenditures which should equal 0 for the technological leader according to the model. Finally, columns 5 and 6 of tables B and C pull together the tests of the two predictions, that are again confirmed. These findings seem to clearly support a complementarity between advertising efforts and the past innovation efforts of firm or their current technological level, in line with the prediction of our model. 18

5 Conclusion [to be written] 19

Références Aghion, P., N. Bloom, R. Blundell, R. Griffith, and P. Howitt (2005) : Competition and Innovation : An Inverted-U Relationship, Quarterly Journal of Economics, 120(2), 701 728. Bagwell, K. (2005) : The Economic Analysis of Advertising, Handbook of Industrial Organization, 3. Fluet, C., and P. Garella (2002) : Advertising and prices as signals of quality in a regime of price rivalry, International Journal of Industrial Organization, 20(7), 907 930. Fogg-Meade, E. (1901) : The Place of Advertising in Modern Business, The Journal of Political Economy, 9(2), 218 242. Grossmann, V. (2007) : Advertising, in-house R&D, and growth, Oxford Economic Papers. Nelson, P. (1974) : Advertising as Information, The Journal of Political Economy, 82(4), 729 754. 20

Appendix A 21

Appendix B Proof of Lemma 1 : See appendix B1 Without loss of generality, we will show that A never advertises on I A. Notice first that there is absolutely no gain for firm A to advertise on I A if the other firm do not. This is because advertising has no effect on consumers who already prefer the advertised good. If firms are neck-and-neck or if A is leader, firm B cannot set a price lower than c B c A without making a negative profit. Whatever A does, firm B cannot make consumers better than indifferent when it advertises on I A. Due to the perfect competition in prices on each segment, B cannot make a positive profit on I A and thus advertising on I A is a strictly dominated strategy for firm B and, from what said above, A do not advertise neither on I A. If A is follower, B can have an incentive to advertise on I A if the cost of advertising is not too high. Without advertising, firms A and B share I A while A prices c A and B prices c B = c A (1 ɛ). If B advertises on I A and set a price equal to c A η, with η infinitely small, it gets the market I A and realises a positive profit ɛ on consumers from I A. Thus, if the cost of advertising do not overpass the expected additional profit, B will advertise on I A. To conclude, we still have to check that even if firm B advertises on I A, firm A do not. Firm A cannot do better than getting back its ex ante advantage of ɛ when it advertises towards consumers on I A, which is not enough to realise any profit. Consequently, firm A never advertises on A. Proof of lemma 2 : From lemma 1, overlapping can only happen on I 0. When firms are neck-and-neck and advertise on I 0 toward a same consumer, they happen to keep this consumer being indifferent between the two goods and make no profit on this consumer. Meanwhile, they incur the cost to advertise toward this consumer. Hence both firms have interest to deviate and overlapping cannot be an equilibrium. If firm B is leader, we have already seen that there is no way for firm A to make any profit by advertising on I A when firm B does. As firm A do not have any advantage on I 0, the situation on this segment is even worse for it and there is no way to realise profits for him by advertising. 22

Proof of property 3 : Consider the benchmark case φ = ɛ that corresponds to moderate advertising costs [To be written] 23