Organizational Capital and Plant Location

Transcription

1 Organizational Capital and Plant Location Peter Egger, Raymond Riezman and Benedikt Rydzek August 21, 2014 Preliminary working paper. Please do not cite without authors permission Abstract In this paper we present a formal model of organizational capital for multi-plant firms and the plant location decision. The optimal number of plants depends on the firm s productivity as well as the costs of transferring organizational capital to plants. As the number of plants increases, the required investment into organizational capital increases and more productive firms invest more into organizational capital and have more plants. If firms can decide to open a plant in a foreign country, they will not only consider the wage differential between the domestic and foreign country, but as well the relative costs of transferring organizational capital. If wages in the foreign country are lower, but the costs of transferring organizational capital is higher, e.g. due to cultural difference, it can be optimal for the firm to have plants in both countries. The theoretical model in this paper is consistent with many empirical findings, e.g. most firms have only a few plants (products), but a few firms with many plants account for a huge share of total output. JEL classification : F23, L11, M11 Key Words : MNC, MNE, Outsourcing, Organizational capital Affiliation: ETH Zürich, CEPR, CESifo, Leverhulme Centre for Research on Globalisatin and Economic Policy (GEP) at the University of Nottingham, and Oxford University Centre for Business Taxation (OUCBT). Address: ETH Zürich, KOF, Leonhardstrasse 21, LEE G 116, 8092 Zurich, Switzerland. egger@kof.ethz.ch. Affiliation: University of Iowa, CESifo. Address: University of Iowa, Iowa City, IA , USA. raymond-riezman@uiowa.edu. Affiliation: ETH Zürich. Address: ETH Zürich, KOF, Leonhardstrasse 21, LEE G 131, 8092 Zurich, Switzerland. rydzek@kof.ethz.ch. Financial support by the Swiss National Science Foundation (SNSF) is gratefully acknowledged. 1

2 1 Introduction Today multinational firms dominate global markets. Bernard et al. (2010) and Bernard et al. (2011) show the importance of multi-national and multi-product firms in global market and international trade. Although only a small percentage of firms are multi-national firms, these firms account for a great share of value added and output, as well as international trade. The recent literature on multi-product firms is mostly based on the concepts of Eckel and Neary (2010) and Nocke and Yeaple (2013). Firms have a core competence for which they are most productive, once firms differentiate into new products they become less and less productive. The management literature has identified this effect in many case studies of expending firms. Nonaka and Takeuchi (1995) and Bray (2007) find that it becomes increasing difficult for firms to management their firm specific knowledge as the firm expends. They find as well that firms that invest more into management of organizational capital are more productive and successful. One reason might be that for firms with higher investments into organizational capital, it is easier to integrate new products into the existing production processes as well as to transfer known production technologies to new products. This effect might aggravate if parts of the production are in foreign countries, where language and cultural difference make it even harder to transfer knowledge. In this paper develop a formal model with organizational capital and multi-product (multiplant) firms. In contrast to the multi-product firm literature we do not consider the export decision of multiproduct firms, but the location decision of products or plants. Firms face different costs creating and transferring OC to plants in different countries. In contrast to in Grossman and Helpman (2005) and Grossman and Rossi-Hansberg (2012) the firm s decision to outsource parts of each production or products to other countries does not only depend on trade costs or wage and technology difference between two countries, but as well on the costs of transferring organizational capital to plants in different locations. The importance of knowledge capital or organization capital has being emphasized by Markusen (2001) with respect to intellectual property rights, where organizational capital is almost an physical input, that is embedded into workers. In this paper we take a different approach, much closer to Oshima et al. (2008) where the firm can create organizational capital (formalize informal organizational knowledge). In the model a firm produces horizontal differentiated goods under monopolistic competition. The firm decides about the scope of the production (how many products (plants) to have), where to locate each plant and how much to invest into organizational capital for each product 2

3 (plant). The model relies on work of Arkolakis and Muendler (2010), Eckel and Neary (2010), Nocke and Yeaple (2013) and Mayer et al. (2014). In contrast to the aforementioned models, we focus on the location decision of a plant and not on its export status. Our model is closest to Nocke and Yeaple (2013), who already introduce organizational capital in a multi-product firm (MPF) model. We contribute to the literature in several dimensions. First, in our model firms are heterogenous in terms of labor productivity and organizatinal capital invesment. Consequently the number of products (plants) differ between firms. Moreover, the share of foreign production differes between firms. Second, increasing the number of firms has no direct negative spillover on the producitivty of other plants. Second, the plants are heterogenous in terms of organizational capital investment, which depends on the plant- and location-specific transfer costs. Moreover, the model provides a micro foundation why marginal costs increase for products (plants) further away from the firm s core competence. transferring organizational capital to new plants becomes increasingly expensive with the number of plants, e.g. operating and supervise one plant demands less organizational structure than operating and supervising 10 plants. Within the model we find that more productive firms produce more output, have more plants, and earn higher profits. The more productive firms spend more resources for organizational capital inplants producing further away from the "core competence" of the firm. Within the model we show that most of the firms in a country are single product (plant) firms and only a few firms are multi-product (multi-plant) firms, which reconciles with the Bernard et al. (2010). Only the most productive firms will produce in foreign countries. The decision to produce in a foreign country depends on the wage differences as well as on the relative costs to transfer organizational capital in the domestic and foreign country. As in Caliendo and Rossi-Hansberg (2013) changes in exogenous variables such as entry costs, wages and the possibility to outsource parts of the production, changes the allocation of organizational capital and the structure of firms in the economy. Entry costs, setup costs of plants and organizational capital transfer costs, have strong effects on the entry decision as well as the firm-plant distribution. Allowing for outsourcing increases number of firms in the market and increases competition as firms can introduce cheaper products produced in the foreign country. Having plants in foreign countries makes firms more profitable, which in turn makes entry more attractive. Last, the share of foreign plants increases monotonically with the firm size as well as their total sales. The paper is structured as follows: In section 2 we develop a multi-product (multi-plant) 3

4 model for a country in autarky. In section 3 we numerically solve the model in autarky and show how changes in key variables change the distribution of plants per firm as well as the entry decision. In section 4 we introduce a second country focus on the outsourcing decision of the firm. In section 5 we numerically solve the model for the two country case. Section 6 concludes. 2 Model in autarky In this section we develop the multi-product (multi-plant) model. We start with the baseline model for a country in autarky, where the firms do not decide about the location of each product (plant). Later we extend the model for the two country case. Considering only one country is closely related to (trade) models of multi-product firms, where firms decide how many products to produce. Utility: All individuals have the following utility function U = u(x) γ u(b) 1 γ, (1) where u(b) gives the utility from the consumption of a good b produced in the outside sector with constant returns to scale and u(x) is an aggregator function with constant elasticity of substitution (CES) over a set Ω of products x((g)). ( u(x) = x((g)) σ 1 σ d(g) (g) Ω ) σ σ 1, (2) where σ > 1 and x((g)) indicates a variety g produced by firm. Ω is the set of all aviable varieties produced by all firms in the market. The (residual) budget constraint is (g) Ω x((g))p((g)) d(g) Y, (3) where the share of total income spend on x-goods is given by Y = γr and R = wl is the total income of all individuals in the economy. w is the wage rate exogenously fixed by the constant returns to scale outside sector. L is the total labor in the economy. The demand for a good x((g)) is given by x((g)) = p((g)) σ P 1 σ Y, (4) 4

5 ( 1 where P = Ω d(g)) p((g))1 σ 1 σ is the common price index. Firm: A firm consists of several plants, denoted by g. Each plant produces a variety g. The first plant of the firm is g = 0, which corresponds to the core competence of the firm. Firm profits are given by π() = G g=0 x((g))p((g)) MCx((g)) f)dg + F e, (5) where f are the variety specific setup costs, F e are fixed entry costs at the firm level. The fixed entry costs F e are sunk costs for the firm, while the setup costs f only occure, if the firm decides to produce a variety g. MC denotes the marignal costs of producing a variety g. is labor productivity of a firm and drawn from a known random distribution. Plant: Each plant of a firm is uniquely identified by g. For the moment we drop the index to simplify notation. The profits at the plant level are given by π(g) = x(g)p(g) w x(g)1+β o(g) β β w o(g)(g + h)ɛ f, (6) where o(g)(g + h) ɛ β is the cost of providing organizational capital, o(g), to the plant g. Higher organizational capital (OC) reduces the marginal costs of producing x(g). On the other hand, it becomes more costly to provide organizational capital for plants with a higher g. In contrast to common multi-product models, see Arkolakis and Muendler (2010), Eckel and Neary (2010), Nocke and Yeaple (2013), the productivity of a product (plant) does not decline per se, but only due to the fact that the transfer of organizational capital becomes more expensive when the firm expends, which is consistent with the findings in the management literature. The firm will always start to produce the variety with the lowest g first, as organizational capital is cheaper for plants with lower g. We interprete g = 0, similar to Eckel and Neary (2010), as the core competence of the firm. Varieties with a lower g are closer to the core competence or business model of the firm. The parameter h indicates the (country-specific) costs of transferring organizational capital to a plant. 1 We assume that the variety specific setup costs are independent of the number of variety. Moreover, β > 0 and ɛ > 0. 1 h will vary with the location of the plant. If a plant is in a region/country very different from the location of the firm, h will be higher and hence the costs of transferring organizational capital will be higher. 5

6 Firm decisions: The firm faces several decisions. 1. Pay the fixed entry costs, f e, and draw a productivity parameter. 2. Decide to enter the market given. 3. Decide how many varieties to produce, G. 4. Decide about the optimal pricing of each variety. 5. How much organizational capital to allocate to each plant, o(g). We solve the firm s problem by backward induction. Optimal organizational capital: The optimal level of organizational capital is derived at the plant level. The firm wants to maximizes profits at the plant level given the number of varieties produced by the firm, the price of each variety and the decisions of all other firms in the market. π(g) o(g) = 0 o(g) β 1 x(g) 1+β = (g + h) ɛ. (7) o(g) = x(g)(g + h) ɛ 1+β g. (8) Substituting this back into the profit equation of the plant yields. The (optimal) marginal costs of a plant are π(g) = x(g)p(g) (1 + β) w βɛ (g + h) 1+β x(g) f. (9) MC = (1 + β) w βɛ (g + h) 1+β. (10) The marginal costs are constant for each plant, but the marginal costs are higher for plants with a higher g. Optimal variety pricing: The firm has monopoly power for all varieties it produces. Given the demand for each variety 6

7 in equation (4) the firm sets the price p(g) to maximzes the plant level profits. The common optimal price for CES utility and constant marginal costs is p(g) = σ σ 1 MC = The profits for each plant are given by π(g) = 1 σ σ σ 1 (1 + β)w βɛ (g + h) 1+β. (11) ( ) σ 1 σ MC 1 σ P σ 1 Y f. (12) σ 1 To simplify notation and allow for tractability, we assume that ɛ = 1+β β π(g) = 1 σ 2 σ 1 > 0. ( σ 1 σ ( ) w 1 σ (1 + β) σ 1) 1 σ (g + h) 2 P σ 1 Y f. (13) Everything else equal, plants further away from the core competence face higher marginal costs and consequently are less profitable. Total firm profits are given by π() = π(g) dg = 1 ( ) σ 1 σ ( ) w 1 σ G (1 + β) 1 σ P σ 1 Y (g +h) 2 dg fg F e. g=0 σ σ 1 g }{{} (14) G Optimal scope: π() = h ( w ) 1 σ G (G + h) fg F e. (15) The firm maximizes its total profits with respect to the number of varieties, G, considering the optimal pricing rule and organizational capital levels as given. π() G = 0 ( w ) 1 σ (G + h) 2 = f, (16) and the optimal scope, G, of the firm is G = The optimal scope decreases in f, w, and h, and increases in. ( ( ) ) 1 w 1 σ 2 h. (17) f 7

8 Substituting G back into the firm level profit equation yields ( π() = 1 h ( ) ) 1 2 w 1 σ 2 f 1 2 h F e, (18) which again is decreases in f, w, and h, and increases in. Entry decision: A firm will only enter the market if its productivity draw,, is sufficiently high to have at least on plant, which we denote by g = 0. As fixed entry costs are sunk, this implies that there is some such that 1 h ( ( ) ) 1 2 w 1 σ 2 f 1 2 h = 0. (19) Solving for yields = w ( fh 2 ) 1 σ 1, (20) which is increasing in w, f, and h. Moreover is a function the price index P. Equilibrium: In equilibrium the expected profits of a firm have to be equal to the fixed entry costs F e. For simplicity we assume that z = σ 1 2 is Pareto distributed with θ scale paramter and α > 2 as shape paramter. The scale parameter θ < z = σ 1 2. Expected profits are E(π()) = 1 ( ( ) ) 1 h E w 1 σ 2 ( = 1 var h = 1 h ( ) ) 1 w 1 σ hf 1 2 = F e ( E ( ) ) 1 2 w 1 σ 2 f 1 2 h = F e ( w 1 σ var( σ 1 2 ) + w 1 σ (E( σ 1 2 )) 2 2hf 1 2 (w 1 σ) ) 12 E( σ 1 2 ) + fh 2 = F e = var(z) + (E(z)) 2 2h 2 ( fw σ 1 z 2 α 2 2h2 ( fw σ 1 ) 1 2 ) 1 2 E(z) = h wσ 1 (F e fh) z = α 1 h α wσ 1 (F e fh), (21) 8

9 which defines the the price index as a function of the known cutoff value. 2 Last, note that the number of firms in the market, M, is given by where E(π()) = F e by equation (21) and MσE(π()) + M e F e = Y, (22) M e = M 1 G(z), (23) where G(z) is the probability density function of a Pareto distribution with scale parameter θ < z and shape parameter α. Solving for M yields M = Y (1 G(z)) (σ(1 G(z)) + 1)F e, (24) z is determined in equation (21) and is a function of entry costs, setup costs, transfer costs, wages, etc. the number of firms also depends indirectly on all these variables. 3 Simulation in autarky To illustrate how fundamental variables of the model, such as transfer costs of organizational capital, h, variety setup costs, f, or fixed entry costs, F e, affect the number of firms, and the distribution of plants, we solve the model numerically using the following baseline calibration given in Table 1: Variable Baseline value Y w 1 f 2 F e 2 σ 5 β 1.5 α 3 h 2 θ 0.3 Table 1: Baseline calibration values. We solve the model for the entry cutoff, using equations (20) and (21). From equation (24) and the entry cutoff we determine the number of firms in the market and from equation (23) we obtain the number of firms drawing a productivity from the Pareto distribution. 2 Note that the cutoff value and the expression in equation (21) both depend on P as is a function of P. Hence equation (21) is an implicit function of P only, and can be solved numerically. 9

10 Then we draw a productivity for each of the potential entrant and calculate the optimal number of plants from equation (17). From this we obtain the distribution of plants by firm. We repeat this exercise 100 times and average the distribution to make the histograms for different calibrations comparable. Baseline calibration: Number of firms with G plants > Number of plants, G Figure 1: Baseline model Figure 1 shows the distribution of firms with G plants. In the baseline calibration 857 firms draw a productivity from the Pareto distribution. Finally 829 firms have a productivity draw higher than the cutoff value of and enter the market. 289 of the firms in the market have only one plant, g = 0. On the other hand there are 4 firms with more than 50 plants. Clearly, the vast majority of firms are single product (plant) firms. This squares with the empirical findings of Bernard et al. (2010) and Bernard et al. (2011). Higher transfer costs: Now we consider the case with h = 3. Figure 2 shows the histogram. First note that, now 850 firms draw from the Pareto distribtion, and 830 firms are entering the market. The entry cutoff is higher, firms have only on plant and 5 firms have more than 50 plants. As transferring organizational capital becomes more expensive, opening new plants, 10

11 Number of firms with G plants > Number of plants, G Figure 2: h = 3 becomes more expensive, as shown by equation (8), plants with a higher g demand more resources for organizational capital. These additional costs are the reason why less firms draw a productivity, but the higher entry cutoff, ensures that firms have on average higher productivity and hence more plants. Higher variety setup costs: Now we consider the case with f = 3. Figure 3 shows the histogram. 850 firms want to enter the market, and in the end 830 firms have a productivity draw higher than and enter the market. 188 firms have only one plant and 4 firms have more than 50 plants. Variety setup costs have the same effect as increasing organizational capital transfer costs. Introducing a new product (plant) becomes more expensive and hence less firms are in the market having less products (plants). Higher fixed entry costs: Now we consider the case with F e = 3. Figure 4 shows the histogram. With higher fixed entry costs only 578 firms draw a productivity and 551 firms have draw higher than and enter the market. 250 firms have only one plant and only 3 firms have more than 50 plants. Higher entry costs reduce the number of firms in the market directly, but interest- 11

12 Number of firms with G plants > Number of plants, G Figure 3: f = Number of firms with G plants > Number of plants, G Figure 4: F e = 3 12

13 ingly they have as well a negative effect on the entry cutoff. As there are less (potential) entrants the price index P increases relative to the baseline simulation, which makes entry more profitable and hence is associated with a lower entry cutoff. Higher total income: Last we consider the case with Y = Figure 5 shows the histogram. Now 1713 firms Number of firms with G plants > Number of plants, G Figure 5: Y = draw from the Pareto distribution and 1657 firms enter the market as their productivity draw is higher than the cutoff of firms have only one plant and 5 firms have more than 50 plants. Doubling the total income, doubles the number of potential entrants and as well the number of firms in the market, and the entry cutoff remains unchanged. 4 Model with two countries Now consider the case, when the firm can decide not only on the number of plants, but as well on the location of each plant. For simplicity we assume that the firm only sells its products in the domestic market, no matter where the plant is located. We consider free trade between the two contries. 3 3 Alternatively, we could interprete the wage rate in the country as trade cost adjusted wage rate. 13

14 A firm will only locate a plant in the foreign country, if the marginal costs of production are lower for some plants in the foreign country. This means that the marginal cost for plants in the domestic has to be lower for lower g and the marginal costs for plants in the domestic country have to be lower for plants greater a cutoff plant ĝ, i.e. the two marginal cost curves have to intersect. We consider a foreign country, that has a lower wage rate, but (due to cultural differences) higher transfer costs of organizational capital, w > w and h < h. 4 Foreign variables are starred. Figure 6 plots two sample marginal curves, which intersect at ĝ. At ĝ it is optimal for the firm to open the next plant in the foreign country. In general these curves must not intersect and it might be optimal for a firm to produce always domestically or in the foreign country. Note that the marginal cost function increases in g and there exists a plant, ĝ, for which the marginal costs are equalized. (1 + β) w (ĝ + h ) η = (1 + β) w (ĝ + h)η, where η = 2 σ 1. w (ĝ + h ) η = w(ĝ + h) η, (25) ĝ = (h h)w 1 η, (26) w 1 η w 1 η The outsourcing cutoff, ĝ, depends only on the wage differential and the costs of transferring organizational capital between the two countries. Only a firm which would produce in more than ĝ plants domestically (if outsourcing would not be available) will have some plants in the foreign country. Moreover, note that everything else equal, the number of plants G is an increasing function of the productivity parameter and hence more productive firms will have more plants in the foreign country. Only if G for the autarky case in equation (17) is higher than the outsourcing cutoff, ĝ, a firm will produce in the foreign country. From this condition we can derive the outsourcing cutoff in terms of productivity of the firm (holding all other variables constant). ( ( ) ) 1 w 1 σ 2 G = h (h h)w 1 η = ĝ, f w 1 η w 1 η ˆ h2 f (h h) w σ 1 η + 1. h w 1 η w 1 η (27) Then the outsourcing productivity cutoff, is decreasing in h, w, and Y, and increasing in f, 4 The two marginal cost curves only intersect if w w < ( ) h η ( h and w ) 1 w η 1 > g+h. g+h 14

15 MC Domestic production Foreign production w φ 1 + β h η w 1 + β hη φ g g Figure 6: Marginal cost curves for domestic and foreign production. w, and h. Optimal scope with foreign production: The profits of a firm that produces in the domestic and foreign country are given by π() = ĝ g=0 G π d (g) dg + ( (w ) 1 σ ĝ = g=0 ĝ π f (g) dg ( w (g + h) 2 ) 1 σ ) G dg + (g + h) 2 dg fg F e, ĝ (28) where π d (g) are the profits from plants in the domestic country and π f (g) are the profits of plants in the foreign country. Note that the firm s profits are perfectly separable into domestic and foreign profits. As shown in equation (27) only firms with > ˆ will have plants in the foreign country. These firms will add plants in the foreign country unit πf () G implies that the optimal scope of the firm is given by = 0. This ( G = ( f f ( w ) 1 σ ) 1 2 h, if < ˆ. ( ) ) 1 w 1 σ 2 h, if ˆ. (29) Note that the entry decision of a domestic firm only depends on the costs of producing at 15

16 the first domestic plant, g = 0, and hence the entry cutoff,, is still given by = w ( fh 2 ) 1 σ 1. Equilibrium with foreign production: The total expected profits in the two country case are given by E(π()) = ( ) 1 fw h w1 σ var(z) + (E(z)) 2 + 2h 2 σ 1 2 (E(ẑ) E(z)) var(ẑ) (E(ẑ)) 2 + ( ) 1 fw h w (1 σ) var(ẑ) + (E(ẑ)) 2h 2 (σ 1) 2 E(ẑ) fh = F e, where the expression in the first line corresponds to the expected domestic profits, which we denote by E(π d ()), and the second line are the expected profits from foreign plants, which we denote by E(π f ()). z and ẑ are given by equations (20) and (27), respectively. Moreover note that the cutoff productivity for outsourcing can be written as function of the entry productivity cutoff. ẑ = (h h) h w 1 η w 1 η w 1 η (30) + 1 w 1 σ 2 z, (31) which can be used to express the equation (30) only in terms of z. z is a function of the price index given by equation (20). Then we can write the above expression in equation (30) as an implicit function of P only and solve for the price index numerically. In equilibrium the income spend on varieties in the domestic country corresponds to the number of firms paying the fixed entry costs plus total revenues of domestic plants plus profits from foreign plants. where M e = M 1 G(z). Y = MσE(π()) + M e F e, (32) 5 Simulation with outsourcing We solve the model with the possibility of outsourcing as previously. Once the price index is determined, all other variables can be recursively recovered. To make the two scenarios comparable we assume that suddenly the option of outsourcing becomes available, so we 16

17 only add a second country with w = 0.9 and h = 3, while leaving all other variables unchanged. This allows to analyse how the structure of firms in the domestic country changes under outsourcing. As previously we calculate the firm distribution for 100 draws of labor productivity from the known Pareto distribution and average the distributions. Figure 7 shows the distribution of firms by number of plants. At first sight the histogram looks fairly similar to the baseline simulation. But, with outsourcing more firms are drawing a productivity paramter, 988. Having plants in the foreign country disproportionally increases the profits of firms and hence more firms want to enter the market. This results in a lower entry cutoff, Under outsourcing, firms are bigger than in autarky, which is obvious now only 252 firms have one plant. Firms with more than 5 plants start to open plants in the foreign country, which corresponds to an outsourcing productivity cutoff of Share of plants in foreign country > Number total plants Figure 7: Firm distribution with outsourcing. For illustratitive reasons we plot the share of foreign plants in total plants in Figure 8. As firms grow, the number of plants in the foreign country increases dramatically. Firms with 100 plants have 95% of their plants in foreign countries. Figure 9 shows the total sales by firm size in terms of number of plants and the share of total sales produced by plants in the foreign country for one random draw from the Pareto distribution. Total sales increase steeply with the number of plants. So only a few firms with many plants can account for a huge share of total production of the economy. Moreover, there is a sharp 17

18 Share of plants in foreign country Number total plants Figure 8: Share of foreign plants in total plants. 4.5 x Total sales Share of sales from foreign plants in total sales Number total plants Number total plants Figure 9: Total sales and share of total sales produced by foreign plants. 18

19 discontinuity in the share of foreign produced output. Once a firm decides to open plants in the foreign country, it will produce at least 60% of its ouput in the foreign country. Notice, that firms with 100 plants have about 95 foreign plants, the share of total output produced by foreign plants is significantly lower, about 85%. 6 Conclusion In this paper we developed a tractable model of multi-plant firms. The model stresses the importance of organizational capital in multi-product (multi-plant) firms. We develop a micro foundation for the commonly used assumption that marginal costs of products further away from the core competence of the firm increase. In the model it becomes more costly for the firm to transfer organizational capital to an extending network of plants. As organizational capital is used as an input in the production, the marginal costs of plants further away from the core competence of the firm increase. Changes in the transfer costs of organizational capital affect the entry decision of firms in the market as well as the number of plants a firm manages. Everything else equal, countries with lower transfer costs of organizational capital have firms with more plants. From the firm perspective, firms with higher productivity have more plants, still the median firm has only a few plants and multiplant firms with many plants are relatively rare. Moreover, the firms with many plants have a disproportional share in total output. These two findings reconcile with the empirical findings of Bernard et al. (2010) that almost all firms in the United States are single product firms, but most value added comes from multi-national and multi-product firms. Last, we find that difference in the costs of transferring organizational capital are a further comparative adavantage of countries, besides classical wage differences. This implies that firms are more likely to open plants in countries that have a similar language and culture, as organizational transfer costs are lower in these countries. Further research needs to link the theoretical model with the empirical distribution of firms and plants. As well as the (observed) outsourcing decision of firms. Firm this we should be able to recover the structural parameters of the model and use the model for counterfacutal analysis. References Arkolakis, Costas and Marc-Andreas Muendler, The Extensive Margin of Exporting Products: A Firm-level Analysis, National Bureau of Economic Research Working Paper Series, 2010, No ,. 19

20 Bernard, Andrew B., Stephen J. Redding, and Peter K. Schott, Multiple-Product Firms and Product Switching, American Economic Review, 2010, 100 (1), ,, and, Multiproduct Firms and Trade Liberalization, The Quarterly Journal of Economics, 2011, 126 (3), Bray, David A., Literature Review - Knowledge Management Research at the Organizational Level, Technical Report, Emory University Caliendo, Lorenz and Esteban Rossi-Hansberg, The Impact of Trade on Organization and Productivity, Quarterly Journal of Economics, 2013, 127(3), Eckel, Carsten and J. Peter Neary, Multi-Product Firms and Flexible Manufacturing in the Global Economy, The Review of Economic Studies, January 2010, 77 (1), Grossman, Gene M. and Elhanan Helpman, Outsourcing in a Global Economy, Review of Economic Studies, 2005, 72 (1), and Esteban Rossi-Hansberg, Task Trade Between Similar Countries, Econometrica, 2012, 80 (2), Markusen, James R., Contracts, intellectual property rights, and multinational investment in developing countries, Journal of International Economics, 2001, 53 (1), Mayer, Thierry, Marc J. Melitz, and Gianmarco I. P. Ottaviano, Market Size, Competition, and the Product Mix of Exporters, American Economic Review, 2014, 104 (2), Nocke, Volker and Stephen Yeaple, Globalization and Multiproduct Firms, National Bureau of Economic Research Working Paper Series, 2013, No ,. Nonaka, I. and H. Takeuchi, The Knowledge-creating Company: How Japanese Companies Create the Dynamics of Innovation, Oxford University Press, Oshima, Atsushi, B. Ravikumar, and Raymond Riezman, Entrepreneurship, Organization Capital and the Evolution of the Firm, Technical Report