Preservation of trade secrets and multinational wage premia.

Bernhardt, Dan ; Dvoracek, Vladimir

I. INTRODUCTION

The notions that multinational corporations (MNCs) pay higher wages than domestic firms in less developed countries (LDCs) and that MNCs have superior technologies are so prevalent and widely held that they are often given as statements of fact without citation or reference (e.g., United Nations 1994). (1) The wage differences far exceed the minor differences in rental payments for land and buildings, or prices paid for domestic raw materials by foreign firms compared with their local counter-parts (Greenaway, Hine, and Wright 2000). While our paper focuses on MNCs in LDCs, there is also evidence of wage premiums and knowledge transfer through trained workers in wealthy countries. Balsvik (2006) examines labour mobility from MNCs to domestic firms in Norwegian manufacturing during the 1980s. For domestic firm workers, he finds that workers with at least 3 years of experience at an MNC command a wage premium of 5% compared with workers with similar experience at a domestic firm. He also finds that workers with MNC experience are 20-25% more productive. Gorg & Strobl (2005) examine technological spillovers from foreign to domestic firms in Ghana. They focus solely on the movement of workers as the mechanism for technology spillovers. Specifically, they compare the productivity levels of domestic firms headed by someone who has worked for an MNC with those headed by someone who has not. They find that firms run by owners who worked for multinationals in the same industry just prior to opening up their own firm are more productive. They argue that this is because these entrepreneurs bring knowledge accumulated in the MNC to their new domestic firm.

This paper starts with the premise that MNCs have superior technologies, and that domestic firms in LDCs lack informational access to this technology, and derives the implications for how the MNC compensates its workers and how it divides job tasks among workers. (2) We identify those conditions under which an MNC pays employees wage premia to discourage competing domestic firms from poaching them in order to retain a monopoly over the informational aspects of its superior technology--the knowledge, intangible capital, and know-how that includes production secrets, management organizational techniques, and marketing skills.

This analysis is more generally relevant for any firm with an informational monopoly over its superior technology that other firms can only gain access to by hiring away its trained workers. We pose the analysis within the context of an MNC investing in a less developed country because it is easy to motivate the informational advantage for the MNC and the data indicate its empirical relevance.

Two specific observations suggest that the wage differential results from the MNC's desire to prevent the dissemination of information about their superior technology. The first observation is that while technology transfer and development by MNCs have been central to the economic growth of developing countries (e.g., in disrupting established practices and introducing innovations as in Wilkins, 1974), MNCs appear to transfer this information to LDCs reluctantly. LaPalombara and Blank (1979) observe that behind the acknowledgment that MNCs do upgrade the quality of labor pools there lurks the suspicion that they hold back on the inculcation of certain essential skills. There is a residual belief that MNCs know more than they are willing to reveal and that the most useful wisdom is held back. Smarzynska (1998) provides additional evidence, documenting that weaker protection of intellectual property rights is a primary determinant of foreign direct investment in transition economies in Eastern Europe. (3) The second observation is that MNCs in developing countries have lower labor turnover than domestic firms. For instance, Gerschenberg (1987) finds that labor mobility for managers employed by MNCs in Kenya is lower than for those employed by domestic firms.

To capture the essence of the MNC's informational advantage, we assume that a worker trained by the MNC on its superior technology can communicate the trade secrets about the technology to a domestic firm. If the domestic firm acquires full information about the MNC's technology, the domestic firm can implement the technology itself. The MNC would like to discourage this information transfer because it lowers the domestic firm costs. This lowers the MNC's profits in the domestic market; Landefeld and Mataloni (2004) find that on average two-thirds of the sales of foreign affiliates go to local markets, and Smarzynska (1998) finds that in Poland, 84.5% of MNC revenues are from local sales.

We characterize the conditions under which an MNC with a superior technology will strategically offer its workers wages that exceed the going rate at domestic firms, and when MNCs will strategically divide job tasks among their workers. It is a well-established observation that an MNC pays a wage premium only if gross domestic profits are higher when the MNC retains its informational advantage. (4) We go beyond this observation to characterize how the industry structure and the initial technology gap between the MNC and the domestic firm affect the MNC's incentives to pay wage premia to preserve its technological superiority. We first prove that if the technology gap is so great that the MNC is initially a monopolist, then it always pays high wages to preserve its monopoly. We then prove that with output competition, a duopolist will pay higher wages if and only if its technology is sufficiently better than the domestic firm's. Thus, we uncover the potentially counterintuitive finding that if the MNC's original technological advantage is small, then it never pays the wage premium necessary to preserve its trade secrets--even though the requisite wage premium is small. This finding is consistent with the empirical observation that MNCs invest only if their technological advantage is sufficient for it to be worth preserving. Relatedly, Lipsey, Blomstrom, and Kravis (1990) and Yu (1990) document that firms that invest in LDCs are more R&D intensive than firms investing in developed countries.

We then determine how the structure of the domestic industry and the nature of competition between the MNC and domestic firms affect outcomes. Ronde (2001) examines how the level of competition affects the incentive of a technologically superior firm to protect its trade secrets. Ronde uses a single parameter to model the degree of competition in the industry. We go beyond Ronde by fleshing out the competitive environment to generate richer predictions:

* We prove that the MNC is more willing to pay its workers wage premia to defend its technological superiority if more domestic firms would otherwise gain access to the MNC's technology. However, the MNC is less willing to pay the wage premia if there are more domestic firms in the industry.

* We find that with quantity competition in the domestic market, the closer substitutes are the two goods, the more willing is the MNC to pay the wage premium. If, instead, the MNC and domestic firm engage in price competition, the MNC always pays the wage premium to prevent its trade secrets from leaking to the competition. The intuition is that price strategies are strategic complements, so that leakage of trade secrets reduces prices by more if competition is in prices rather than quantities.

* We prove that if the MNC's technological know-how is divided evenly among its employees, that if the MNC pays a wage premium to any worker, then it pays it to all workers.

* The strategic design of jobs can make it more attractive for an MNC to pay wage premia. An MNC both engages in inefficient job splitting and offers higher wages if and only if the cost of job splitting is sufficiently small.

Other researchers have focused on the ability of the MNC to sell its superior technology to the domestic firm through a licensing contract. Ethier (1986) considers an MNC that cannot write complete licensing contracts, and Ethier and Markusen (1996) consider the possibility that the MNC cannot enforce these contracts. Researchers have also posited reasons other than the preservation of trade secrets for the wage gaps between workers at MNCs and domestic firms. Buckley and Enderwick (1983) assert that MNCs' production technologies simply require more highly skilled workers and more skilled workers must be paid more. Globerman, Ries, and Vertinsky (1994) argue that the high capital intensity of MNCs induces them to pay (higher) efficiency wages to elicit greater effort because it is more costly for capital intensive firms to suffer employee shirking or absenteeism. Finally, Head (1998) argues that the higher wages at MNCs are the outcome of a bargaining game in which the extra rents generated by the superior technologies are divided between workers and firm owners.

The next section presents our economic analysis. A conclusion follows. All proofs are in the Appendix.

II. ECONOMIC ENVIRONMENT

Consider an LDC economy with one multinational firm and one domestic firm. The firms produce solely for the domestic market of the LDC. There exists a large pool of unskilled workers who have a reservation wage [bar.w]. Workers can (a) work for the domestic firm, (b) work for the multinational firm, or (c) not work in the manufacturing sector and earn the reservation wage of [bar.w] in the agricultural sector. Agents are risk neutral.

For the sake of exposition, we consider a dress-making industry. We first suppose that each firm employs but one worker to whom they supply cloth to make the dresses. We then show how the analysis generalizes when the firms employ many workers. The MNC's technology produces dresses according to

[Q.sup.m](L, C) = C/y, L [greater than or equal to] 1

[Q.sup.m](L, C) = 0, L < 1,

while the domestic firm's technology produces dresses according to

[Q.sup.d](L, C) = C/z, L [greater than or equal to] 1

[Q.sup.d](L, C) = 0, L < 1,

where C is the quantity of cloth used and L is the number of workers employed. We capture the multinational's superior technology through its lower cloth requirement, z > y. The cost of a unit of cloth is c.

The MNC always has access to its superior technology. The domestic firm has access to its inferior technology, but it can also bid for the services of a worker who has been trained on the MNC technology. The domestic firm gains informational access to the MNC's superior technology if and only if it hires this worker. (5)

The firms sell the dresses in the LDC's domestic market. Domestic demand is given by

(1) p([Q.sup.d] + [Q.sup.m]) = A - ([Q.sup.d] + [Q.sup.m]).

The timing of moves is as follows: (i) Firms offer wages and then (ii) workers choose employers. Next, (iii) firms make output or price choices, and finally, (iv) output is sold on the market and workers receive wages.

Using the superscript d to denote the domestic firm and m to denote the multinational, let [[pi].sup.fi] denote the profits of the firm f [member of] {d, m} gross of period 2 wages if the domestic firm is informed about the MNC's production technique, and let [[pi].sup.fu] denote its profits if the domestic firm is uninformed. We begin with a preliminary lemma on the necessary and sufficient conditions for it to be profitable for the MNC to pay a wage premium to prevent the domestic firm from poaching its trained workers. The result is almost immediate and quite general.

LEMMA 1. The lowest wage that the multinational firm can pay its trained employee and still retain him is

(2) [w.sup.*] = [[pi].sup.di] - [[pi].sup.di] - [[pi].sup.du] + [bar.w].

The MNC pays [w.sup.*] [bar.w] to preserve its informational monopoly over its technology if and only if gross industry profits are larger when the information is not transmitted, that is, if and only if

[[pi].sup.mu] + [[pi].sup.du] > [[pi].sup.mi] + [[pi].sup.di] .

The wage premium equals the difference in the gross profits of the domestic firm when it is informed and when it is not, [[pi].sup.di] - [[pi].sup.du]. The MNC pays this premium if and only if the difference in the MNC's gross-of-wages profits, [[pi].sup.mu] - [[pi].sup.mi], exceeds this wage premium. Rearranging reveals that the MNC pays the higher wage if and only if gross industry profits are higher when the domestic firm is uninformed. Thus, comparisons of gross industry profits determine whether the MNC chooses to maintain its informational advantage, the same joint profit comparisons found in the industrial organization literature.

Because monopoly profits always exceed duopoly profits, it follows that if the domestic firm's technology is so bad that the multinational would be a monopolist were the domestic firm forced to use its own technology, then the multinational would always offer the higher wage, [w.sup.*], to maintain its informational advantage:

COROLLARY 1. Suppose the multinational's technological advantage, z - y, is so great that the multinational would be a monopolist if it prevents the domestic firm from acquiring information about its technology. Then, the multinational always pays the high wage, [w.sup.*], to preserve its informational monopoly over its technology.

We now characterize how the nature of competition prices or quantities--affects outcomes when both firms are active.

A. Quantity Competition

Suppose that the MNC and domestic firms compete in the local market over output. Then the MNC pays the wage premium if and only if its technological advantage is sufficiently great:

PROPOSITION 1. There exists a [z.sup.*] > y such that if z [less than or equal to] [z.sup.*] then the multinational firm pays the high wage, [w.sup.*], to preserve its informational monopoly; and if z < [z.sup.*] then the multinational firm pays w, and the domestic firm hires away the trained employee to acquire access to the multinational's technology. The value of [z.sup.*] is

(3) [z.sup.*] = (2A + 3cy)/(5c).

Quite generally, an extension of Corollary 1 holds: if the MNC's technological advantage almost makes it a monopolist, the MNC finds it profitable to pay a wage premium to maintain its technological supremacy. In contrast, the reason why an MNC with a small technological advantage will not find it profitable to pay the wage premium is more subtle--after all, the domestic firm is not willing to pay much to obtain a modest technology improvement. Suppose, initially that cloth requirements at the two firms are identical, z = y, and consider the effect on industry profits of a marginal worsening of the domestic firm's technology. The strategic effects on each firm's output are almost offsetting--because of its higher costs, the domestic firm reduces its output slightly by [DELTA][Q.sup.d], and in response, the MNC increases its output by [DELTA][Q.sup.d]/2. However, the slight worsening of the domestic firm's technology raises its production costs by an amount equal to its output level, [Q.sup.d]. Thus, if the MNC's technological advantage is small, the increase in industry profits from the reduction in total output is more than offset by the increase in the costs incurred by the domestic firm. It follows that industry profits are lower if the MNC's slight technological advantage is preserved. In turn, this implies that the MNC should not pay wage premia to preserve its technological advantage.

The qualitative prediction that the MNC should pay the high wage [w.sup.*] to preserve its informational monopoly if its advantage is substantial, but not if its cost advantage is slight is robust to functional form assumptions. The dissipation of rents if the MNC's advantage is sufficiently large makes the first half of this assertion immediate. We now prove that for slight cost advantages, the argument detailed above extends:

PROPOSITION 2. Let the MNC have convex cost function c(Q), and the domestic firm has a marginally worse cost Junction c(Q) + [epsilon]Q, where [epsilon] > 0. Suppose that demand P(Q) satisfies P"(2Q)Q + P'(Q) < 0. Then for all [epsilon] sufficiently small, the MNC should not pay the high wage [w.sup.*] to preserve its informational monopoly.

There is strong empirical evidence that MNCs pay wage premia if and only if their technological advantages are substantial. Using a cross-section of firm-level data for Uruguay, Kokko, Tansini, and Zejan (1996) find productivity spillovers to domestic firms with moderate technology gaps, (measured as the difference between the firm's labour productivity and the average labour productivity in foreign firms), but not for domestic firms which use considerably lower levels of technology. Girma, Greenaway, and Wakelin (2001) use firm-level panel data to examine productivity spillovers in UK manufacturing. They find evidence of spillovers to firms with a low difference between the firm's productivity level and the domestic industry productivity level. They refer to this difference as the technology gap and they find that domestic firms with a gap of 10% or lesser appear to increase productivity with increasing foreign presence in the industry, while firms with higher gaps seem to suffer reductions in productivity. Hale and Long (2006) use a World Bank survey of 1500 firms in five Chinese cities to study whether the presence of foreign firms produces technology spillovers to domestic firms operating in the same city and industry. They find that a smaller technology gap between foreign and local firms results in larger spillovers.

This evidence is consistent with our theory that MNCs should pay high wages only when productivity differences are significant, but not when productivity differences are small. In contrast, theories that MNC pay more because their technologies require more skilled workers (e.g., Buckley and Enderwick 1983; Globerman, Ries, and Vertinsky 1994) would suggest that wages would continue to be high when the domestic firms acquire the technology: these theories concern productivity levels not productivity differences--if anything, when domestic firms acquire the technology, wages should be bid up further due to increased demand for skilled workers.

Another distinction between our theory and the premise that the wage differential is due to the more skilled workers that MNCs hire is that when the wage differential is due to trade secrets that MNC workers acquire, then the wage gap should increase with employment tenure, that is, wage growth should be greater at MNCs than at domestic firms. In contrast, if the wage difference is due to differences in ability, then the wage gap should be independent of tenure.

Imperfect Substitutes. We have assumed that the output goods of the domestic and multinational firms are perfect substitutes. But suppose, instead, that the respective demands for their goods are given by

[P.sub.M]([Q.sup.m], [Q.sup.d]) = A - ([Q.sup.m] + b[Q.sup.d]), [P.sub.D]([Q.sup.d], [Q.sup.m])

=A - ([Q.sup.d] + b[Q.sup.m]), 0 < b [less than or equal to] 1.

Goods are perfect substitutes if b = 1. But how does b < 1 affect an MNC's willingness to pay a wage premium?

PROPOSITION 3. The greater is b, the more attractive it is for the MNC to pay the wage premium to preserve its informational advantage.

The MNC still pays the wage premium if and only if it raises total industry profits. As b rises, the output goods become closer substitutes so the output of one firm has a greater impact on the profits of the other. Consequently, the MNC's incentive to pay the wage premium to prevent leakage of its trade secrets rises.

B. Price Competition

Now suppose that the MNC and domestic firm engage in price competition rather than quantity competition. We obviously require heterogeneity in products, and to ease presentation, we model spatially the possibility that the output goods are imperfect substitutes. A measure 1 of consumers uniformly distributed on the unit interval, the MNC is located at position 0, and the domestic firm is located at position 1. Consumers must travel to the firm to purchase the consumption good. A consumer who travels a distance s to purchase the good at price p receives utility M - p - [Ts.sup.2], where M represents the consumer's utility if price and distance are zero and T > 0 is the rate at which utility falls with respect to the squared distance. The MNC and domestic firm choose prices [P.sub.m], [P.sub.d] to maximize profits.

PROPOSITION 4. With price competition, the MNC always pays the wage premium.

With price competition, industry profits are always higher if the MNC preserves its trade secrets, so the MNC always pays the wage premium. Indeed, with price competition, the requisite wage premium may be quite small. The two goods become perfect substitutes if travel costs are zero, T = 0, in which case gross industry profits are positive if the MNC retains the technology advantage, but are zero if it does not. The required wage premium is therefore zero.

More generally, what drives this result is that with price competition, the fall in the marginal cost of the domestic firm causes both firms to cut price. By contrast, with quantity competition, the increase in the output of the domestic firm is partially offset by the reduction in output of the MNC, leading to a smaller decline in price. As a result, the MNC is more willing to pay a wage premium to preserve its informational monopoly under price rather than quantity strategies.

C. Additional Domestic Firms

We have considered an industry structure with a single domestic firm. We now address how our results are affected when there is more than one domestic firm operating with the inferior technology. We distinguish between two types of domestic firms: those that can possibly implement the MNC's technology and those that cannot. Specifically, in our core model of quantity competition, we now suppose that n - 1 domestic firms operate with the inferior technology, and that j [less than or equal to] n - 1 domestic firms can band together to hire away the MNC's employee and obtain its trade secrets. Obviously, if the MNC's technological advantage would discourage domestic firms from producing with their inferior technologies, then the MNC would pay the wage premium to its trained worker to preserve its monopoly independently of n and j. The interesting scenario is when the domestic technology is not too inferior. We next characterize how the MNC's incentives to pay a wage premium to preserve its monopoly over its trade secrets vary with j and n.

PROPOSITION 5. Suppose the MNC's technological advantage is' small enough that it does not have a monopoly. Then, ceteris paribus, raising the number of domestic firms, n - 1, in the industry reduces the willingness of the MNC to pay wage premia to preserve its informational advantage. Conversely, raising the number of domestic firms j that would obtain access to the MNC's trade secrets raises the willingness of the MNC to pay wage premia to preserve its technological superiority.

If, given n and j, the MNC is indifferent between paying a wage premium to preserve its informational advantage and paying no premium and having its technology disseminate to j domestic firms, then raising the number of firms in the industry or reducing the number that would gain access to its technology would cause the MNC to cease paying the wage premium. One's intuition may be that with more domestic firms, the MNC's desire to preserve its informational advantage would rise, because the requisite wage premium falls with n. This intuition is misplaced. Indeed, if j = 1, it is never optimal for the MNC to pay the wage premium if there are more than three active domestic firms. When n is large, having the technology information leak out to a single domestic firm serves to commit the two firms to a greater market share, raising their profits. As with Stackelberg competition, total output goes up, so that industry profits fall; but, the profit share that accrues to the two firms rises, and this more than compensates. Most transparently, suppose that in the absence of information transfer, all firms would produce, but that if one domestic firm gained access to the MNC's technology, the combined output of the newly informed domestic firm and the MNC would drive the other domestic firms out of the market. Then raising n has no effect on joint profits if trade secrets are revealed, but raising n reduces their market share and price and hence profits if the MNC preserves its trade secrets. It follows immediately in this case that raising n reduces the MNC's incentive to preserve its trade secrets.

This prediction can reconcile the empirical findings of Sinani and Meyer (2004). They use a production function framework to estimate the impact of technology transfer from foreign direct investment on the growth of sales of domestic firms in Estonia during the period from 1994 to 1999. They find that increased competition among domestic firms is associated with greater increases in sales for domestic firms and with larger technology spillovers from MNCs.

Conversely, for fixed n, raising the number of firms j to which the information would leak always makes it relatively more attractive for the MNC to pay a wage premium to its workers to preserve its informational advantage. Indeed if all domestic firms would acquire the MNC's trade secrets, that is, n - 1 = j, then raising n always makes it more attractive to preserve trade secrets. This issue speaks to what effect the share of domestic firms that are able to absorb the new technology has on the incentive of the MNC to pay the higher wage and thereby restrict technology.

D. Robustness

Sale of Technology. We have assumed throughout that the MNC cannot sell its technology to the domestic firm. However, the analysis if the MNC can sell its technology is almost identical. The sole difference is that the MNC finds it more profitable to sell its technology than to have the domestic firm acquire the technology from its trained employee, as the rents of [[pi].sup.di] - [[pi].sup.du] accrue to the MNC rather than the worker. Extending Lemma 1 to this case, the MNC finds it optimal to pay the wage premium to preserve its informational advantage rather than sell its technology if and only if [[pi].sup.mu] - [[pi].sup.mi] > 2([[pi].sup.di] - [[pi].sup.du]). This condition is twice as stringent as the condition [[pi].sup.mu] - [[pi].sup.mi] > ([[pi].sup.di] - [[pi].sup.du]) that arises when the worker receives the rents.

Many Workers. A possible concern with the empirical relevance of our analysis is that MNCs employ many workers, and that if a domestic firm could acquire informational access to the MNC's technology by hiring a single worker, then the cost of preventing this information transmission would be prohibitive, even when one factored in costs of changing jobs, the time it takes to communicate the information, etc. However, just as it is a simplification to assume that firms only employ one worker, so, too, it is a simplification to assume that domestic firms can acquire access to the MNC's technology by hiring away only one worker. In a large firm with many different job assignments, a domestic firm might have to raid a significant portion of the MNC's labor force to acquire informational access to the technology, and the MNC would only have to prevent the domestic firm from hiring "key" people.

We now show that the analysis extends when the MNC employs N trained workers, each of whom has informational access to 1/N-th of its technological advantage. Thus, if the domestic firm hired j of the MNC's workers, it would lower its cloth requirements by j(z - y)/N. Index the domestic firm's gross (of wage) profits when it raids j trained workers from the MNC by j,

[[pi].sup.d](j) = (1/9) [A(c/N)[[2z(N - j) - y(N - 2j)]].sup.2].

Since the domestic firm's gross profits are a convex function of the number of trained workers that it hires, if the MNC is to keep N - j of its workers, it must offer each of those N - j workers a wage premium of

[[[pi].sup.d](N) - [[pi].sup.d](j)]/(N - j).

The multinational's problem can then be posed as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

which, since [[pi].sup.d](N) is not a function of j, again reduces to choosing j so as to maximize total industry profits,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Inspection reveals that total industry profits are a convex function of j: The multinational either pays the wage premium to all of its trained workers, or it pays the wage premium to none of its workers and permits the domestic firm to acquire informational access to its technology. Hence, it follows as a direct corollary of our existing analysis that if and only if the multinational's technological advantage is sufficiently great will it pay a wage premium (to all of its workers) in order to retain an informational monopoly over its technology.

It is also worth noting the qualitative impacts if some of the MNC's employees have access to "more" information about the MNC's technological advantage than others. It is immediate that if worker A's information would lower the domestic firm's production costs by more than B's, then the MNC would have to pay A a higher wage premium than B to keep him. This suggests that the relative wage premium the MNC must pay would be greater for workers higher in the firm hierarchy, who are likely to be privy to more information.

Information about such wage differentials in developing countries is unreliable. However, supporting evidence for our theory comes from the observation that our analysis is relevant for any firm with an informational monopoly over its superior technology that other firms can only gain access to by hiring away its trained workers. Larger firms tend to have superior technologies (and hence operate on larger scales), and our model predicts that they must pay wage premia in order to retain their trained workers and preserve their informational advantages. It has been thoroughly established empirically that larger firms in the United States pay higher wages than smaller ones, and that this wage differential is greater higher in the firm hierarchy (see, for example, Brown, Hamilton, and Medoff 1990). (6) This is consistent with the prediction that workers with "more" information must be paid more. In addition, casual empiricism suggests that the percentage of non-domestic employees is higher in MNCs in higher levels of the firm hierarchy, and one strongly believes that their relative wage premia are higher. Indeed, one reason that non-domestic workers may be placed in key positions is that they would find it less attractive to work for the domestic firm, which makes it more expensive for domestic firms to hire such workers away.

The Strategic Design of Worker's Tasks. We now look at another way in which an MNC may prevent the transfer of technology to the domestic firm--the MNC may strategically design job assignments, so that any single worker knows less of the key information that a domestic firm must acquire before it can implement the multinational's technology. Such strategic job design reduces the value of any single worker to the domestic firm, and hence the wage premium that the multinational must pay. The cost is that the multinational operates less efficiently. (7)

To capture this choice simply, suppose that the firms each employ two workers. The multinational can fully train each of its workers so that each knows all aspects of the multinational's superior technology, in which case its cloth requirements are y, but this has the associated cost that the domestic firm can learn the multinational's technology by hiring away either of its workers. Alternatively, the multinational can split the training/job assignment, which raises its cloth requirements to [bar.y], z > [bar.y] [greater than or equal to] y, but forces the domestic firm to hire both of the MNC's workers in order to acquire the MNC's technology.

We use the superscript f to index variables when the MNC fully trains its workers, and the superscript s to index variables when the MNC designs its jobs to provide distinct specialized training to each of its workers. We also index the variables by the cloth requirement at the MNC.

If the MNC fully trains each of its workers, then it must pay each of its workers the full wage premium:

(4) [w.sup.*f] (y) = [bar.w] + [[pi].sup.di](y) - [[pi].sup.du](y).

If, instead, the MNC splits the expertise that the workers acquire through training, then the wage premium it must pay each of its workers to prevent the technology's transmission is only one-half of the difference between what the domestic firm would earn if had access to the MNC's technology, and if it did not:

(5) [w.sup.*s] (bar.y) = [bar.w] + (1/2)[[pi].sup.di](bar.y) - [[pi].sup.du]([bar.y)].

The benefit to the MNC of making hiring both workers essential for information transmission to the domestic firm is that the total wage premium is reduced. The cost of the split technology is that the MNC is a less efficient producer.

The following is immediate:

PROPOSITION 6. Suppose that gross aggregate industry profits are greater if the domestic firm does not have access' to the fully trained technology, y. Then, there exists an [epsilon] > 0 such that if and only if [bar.y] - y [less than or equal to] [epsilon], the multinational inefficiently subdivides the tasks it assigns its workers.

That is, the MNC divides tasks as long as doing so is not "too" costly. The proposition presumes that the MNC retains its trained workers and thus pays the high wage. To understand the intuition, consider two extreme cases. First, suppose it is costless for the MNC to subdivide tasks, that is, [bar.y] = y. The MNC will now certainly divide task assignments as the wage premium is reduced, while the MNC's per unit costs remain unchanged. Conversely, suppose that splitting tasks is so inefficient that it raises the MNC's per unit costs to the level of the domestic firm's, that is, [bar.y] = z. In this case, it is clearly not optimal for the MNC to subdivide tasks.

Note that the MNC does not choose between job splitting and paying a wage premium. That is, it is only valuable for the MNC to divide job tasks inefficiently when it is optimal to pay a wage premium to preserve its informational advantage: dividing job tasks is not a substitute for wage premia, but rather a complement. There is no gain to splitting jobs inefficiently if it fails to prevent domestic firms from acquiring the superior technology. Job splitting simply reduces the total wage premia. Indeed, it could be that the MNC would pay wage premia only if subdividing tasks was not too inefficient.

III. CONCLUSION

This paper characterizes the conditions under which a multinational corporation in a less developed country strategically offers wage premia to employees to prevent domestic firms from raiding the MNC's trained workforce to gain informational access to the MNC's technology. We characterize how the nature of competition between firms (price vs. quantity) and how the structure of the industry affects the MNC's choice of whether to pay wage premia to prevent its trade secrets from leaking to domestic firms. We show that an MNC strategically offers wage premia if and only if its technological advantage is sufficiently great. We also provide conditions under which an MNC inefficiently divides job tasks between workers in order to raise the costs to domestic firms of acquiring informational access to the MNC's technology. (8)

To keep the analysis tractable, our model is a "snapshot" model of the costs and benefits that the MNC faces from offering high wages to preserve its technological advantage. In a stationary dynamic model, where the wage costs and the profits from having exclusive access to the superior technology would be the same each period, our qualitative predictions immediately extend. More interestingly, suppose that as an LDC's economy developed, for exogenous reasons, one domestic firm out of a few acquires the MNC's technology. How would this affect outcomes? Observe first that the required wage premium would be reduced, because the value to a second domestic firm of acquiring the technology would be reduced as it would be competing against more firms with the advanced technology. Next observe that the advanced domestic firm would also have to pay the wage premium, else the information spread to other domestic firms. Thus, a loose prediction would be that as an LDC develops, MNC wage premia should decline, and there should be growing heterogeneity in compensation by domestic firms according to the sophistication of their technologies.

APPENDIX

Proof of Lemma 1: The domestic firm will raid the MNC's workforce as long as the wage premium it pays does not exceed the difference in gross profits when informed and not. To solve for [w.sup.*], equate the domestic firm's net profits (gross profits minus labor costs) under the two scenarios:

[[pi].sup.di] - [w.sup.*] = [[pi].sup.du] - [bar.w] [??] [w.sup.*] = [[pi].sup.di] - [[pi].sup.du] + [bar.w].

The multinational firm pays [w.sup.*] if and only if its net profits when the domestic firm is uninformed are larger than when the domestic firm is informed:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Proof of Proposition 1: The profit functions of the two firms when the domestic firm is uninformed are

[[pi].sup.du] = (A - [Q.sup.mu] - [Q.sup.du] - cz)[Q.sup.du] and [[pi].sup.mu] = (A - [Q.sup.du] - [Q.sup.mu] - cy)[Q.sup.mu].

The associated profit-maximizing levels of output are

[Q.sup.du] = (1/3)[A + c(y - 2z)] and [Q.sup.mu] = (1/3)[A + c(z - 2y)],

implying firm profit levels of

[[pi].sup.du] = (1/9)[[A + c(y- 2z)].sup.2] and [[pi].sup.mu] = (I/9)[A +c(z - 2y)].sup.2].

If the domestic firm has access to the MNC's technology, the profit maximizing levels of output become

[Q.sup.di] = [Q.sup.mi] = (1/3)(A - cy).

Gross profits for each firm are:

[[pi].sup.mi] = [[pi].sup.di] = (1/9)[(A -cy).sup.2].

Substituting these profit levels we solve for

[w.sup.*] = (1/9)[(A - cy).sup.2] - [(A +c(y - 2z)).sup.2]] + [bar.w].

For the MNC to pay the higher wage, its net profits must be larger when the domestic firm is uninformed:

[[pi].sup.mu] + [[pi].sup.du] - [[pi].sup.mi] - [[pi].sup.di] > 0.

Substitution yields:

[[A - cy + c(z - y)].sup.2] + [[A - cy - 2c(z - y)].sup.2] - 2[(A cy).sup.2] > 0,

which holds if and only if z > (2A + 3cy)/(5c).

Proof of Proposition 2: It suffices to show that raising marginally reduces joint profits,

[[pi].sup.d] + [[pi].sup.m] = (P([Q.sup.d] + [Q.sup.m])([Q.sup.d] + [Q.sup.m]) - c([Q.sup.d]) - [epsilon][Q.sup.d]),

if information is not shared. At [epsilon] = 0, the first-order conditions for optimization at [Q.sup.d] = [Q.sup.m] = Q are

(6) P'(2Q)Q + P(2Q) - c' (Q) = 0.

To obtain the impact of a marginal change in [Q.sup.d] on [Q.sup.m], differentiate the first-order condition for the MNC,

[partial derivative][Q.sup.m]/[partial derivative][Q.sup.d] = - P"(2Q)Q + P' (2Q)/P"(2Q)Q + 2P' (2Q) - c' (Q).

The premises of the proposition imply that - 1 < [partial derivative][Q.sup.m]/ [partial derivative][Q.sup.d] < 0. Let a = -[partial derivative][Q.sup.m]/ [partial derivative][Q.sup.d]. Next, differentiate the first-order condition for the domestic firm with respect to [epsilon] at [epsilon] = 0, and substitute for [partial derivative][Q.sup.m]/[partial derivative][Q.sup.d] to obtain

(7) 1 = [P"(2Q)Q(1 - a)+2P'(2Q)(1 - a) - c'(Q)] ([partial derivative][Q.sup.d]/[partial derivative][Q.sub.[epsilon]).

Finally, combining these facts, differentiate joint profits with respect to [epsilon] at [epsilon] = 0 to obtain

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where the second equality follows from (6), the third equality follows from substituting for -1 using (7), and the inequality follows immediately from the premises of the proposition.

Proof of Proposition 3. The MNC pays the wage premium if and only if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Differentiating with respect to b and simplifying yields:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Proof of Proposition 4: Given prices [p.sub.d], [p.sub.m], the location, [theta], of the consumer who is indifferent between purchasing from the MNC and the domestic firm is given by

- [T[theta].sup.2] - [p.sub.m] = T[(1 - [theta]).sup.2] - [p.sub.d] [??] [theta] = (T - [p.sub.m] + [p.sub.d])/(2T).

The multinational's problem is then to maximize:

([p.sub.m] - cy)(T - [p.sub.m] + [p.sub.d])/(2T).

Assuming that T is sufficiently large, so that the equilibrium is characterized by the simultaneous solution to the associated first-order conditions:

[p.sub.m] = (T + [p.sub.d] + cy)/2; [p.sub.d] = (T + [p.sub.m] + cz)/2,

the equilibrium prices are

[p.sup.*.sub.m] = (3T + 2cy + cz)/3; [p.sup.*.sub.d] = (3T + 2cz + cy)/3.

In turn, the MNC's profits are (1/2T)[[(3T - cy + cz)/ 3].sup.2], while the domestic firm's profits are (1/2T)[[(3T - cz + cy)/3].sup.2]. After a similar computation of profits if the MNC does not pay the wage premium, we see that paying the wage premium is optimal if and only if

[[(3T - cy + cz)/3].sup.2] + [[(3 T - cz + cv)/3].sup.2] - 2[[(3 T)/3].sup.2] > 0.

For z > y, this inequality always holds.

Proof of Proposition 5: We adopt more concise notation. Let [??] = cz - cy be the difference in marginal costs between the MNC and domestic firms, and let [??] = A - cy be the difference between the demand intercept and the MNC's marginal cost.

Suppose the MNC pays the high wage to its worker to preserve its trade secrets. Routine calculations reveal that the MNC produces [Q.sup.mu] = ([??] + (n - 1)[??]]/(n + 1) and the n - 1 domestic firms each produce [Q.sup.du] = ([??] - 2[??])/(n + 1). Substituting these output levels we compute the joint profits of the MNC plus j domestic firms:

(8) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

If, instead the worker left and the trade secrets were revealed to j domestic firms, each of these firms produces [Q.sup.mi] = [Q.sup.di] = [[??] + (n - j - 1)[??]]/(n + 1) and the n - j - 1 domestic firms that did not gain access to the superior technology each produce [Q.sub.diu] = [[??] - (j + 2)[??]]/(n + 1) (the case where these firms are driven out of the market is immediate, and is discussed in the text). Substituting these output levels we obtain joint profits if information is transferred:

[[pi].sup.mi] + j[[pi].sup.di] = (j + 1)[[([??] + (n - j - 1)[??])/(n + 1)].sup.2].

Computing the difference in aggregate profits yields

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Hence [[pi].sup.mu] + j[[pi].sup.du] = [[pi].sup.mi] + j[[pi].sup.di] takes the sign of

(9) -2(n - j)[??} - [[(n - j).sup.2] - 4n + 3j - 1][??].

Differentiating (9) with respect to n yields

-2[??] - [2(n -j - 2)][??} < 0,

as [Q.sup.diu] > 0 implies [??] > (j + 2)[??]. Hence, raising n reduces the attractiveness to the MNC of preserving his informational advantage. In particular, when j = 1, we obtain

[[pi].sup.mu] + j[[pi].sup.du] - [[pi].sup.mi] + j[[pi].sup.di] = [- 2[??](n - 1) - [??][[(n - 1).sup.2] - 4n + 2)],

which is negative for all n [greater than or equal to] 5.

Differentiating (9) with respect to j yields

2. [??] + [2(n -j) - 3][??] > 0,

as [Q.sup.diu] > 0 implies [??] > (j+ 2)[??]. Hence, increasing j raises the attractiveness of preserving trade secrets.

Proof of Proposition 6: Immediate. Profits from splitting decline monotonically in [bar.y], so that for [bar.y] sufficiently great, split training is dominated by one of the other alternatives.

Since at [bar.y] = y, split training dominates full training, and gross industry profits are greater if the domestic firm does not have access to trained workers, the result follows from Lemma 1.

ABBREVIATIONS

LDC: Less Developed Countries

MNC: Multinational Corporations

R&D: Research & Development

doi: 10.1111/j.1465-7295.2008.00142.x

REFERENCES

Balsvik, R. "Is Mobility of Labour a Channel for Spillovers From Multinationals to Local Domestic Firms?" Mimeo, Norwegian School of Economics and Business Administration, 2006.

Bernhardt, D., and D. Scoones. "Promotion, Turnover and Preemptive Wage Offers." American Economic Review, 83, 1993, 771-91.

Biersteker, T. Distortion or Development?." Contending Perspectives on the Multinational Corporation. Cambridge, MA: MIT Press, 1978.

Brown, C., J. Hamilton, and J. Medoff. Employers Large and Small. Cambridge, MA: Harvard University Press, 1990.

Buckley P., and J. Enderwick. "Comparative Pay Levels in Domestically-Owned and Foreign-Owned Plants in UK: Manufacturing-Evidence From the 1980 Workplace Industrial Relations Survey." British Journal of Industrial Relations, 21, 1983, 395-400.

Ethier, W. "The Multinational Firm." Quarterly Journal of Economics, 101, 1986, 805-33.

Ethier, W., and J. Markusen. "Multinational Firms, Technology Diffusion and Trade." Journal of International Economics, 41, 1996, 1-28.

Forsyth, D., and R. Solomon. "Choice of Technology and Nationality of Ownership in Manufacturing in a Developing Country." Oxford Economic Papers, 29, 1997, 258-82.

Fosfuri, A., M. Motta, and T. Ronde. "Foreign Direct Investment and Spillovers Through Workers' Mobility." Journal of International Economics, 53, 2001, 205-22.

Gerschenberg, I. "The Training & Spread of Managerial Know-How, a Comparative Analysis of Multinationals and Other Firms in Kenya." World Development, 15, 1987, 931-9.

Girma, S., D. Greenaway, and K. Wakelin. "Who Benefits from Foreign Direct Investment in the UK." Scottish Journal of Political Economy, 48, 2001, 119-33.

Glass, A., and K. Saggi. "Multinational Firms and Technology Transfer." Policy Research Working Paper Series 2067, The World Bank, 1999.

Globerman, S., J. Ries, and I. Vertinsky. "The Economic Performance of Foreign Affiliates in Canada." Canadian Journal of Economics, 27, 1994, 143-56.

Gorg, H., and E. Strobl. "Spillovers from Foreign Firms Through Worker Mobility: An Empirical Investigation."

Scandinavian Journal of Economics, 107, 2005, 693-709.

Greenaway, D., R. Hine, and P. Wright. "Further Evidence on the Effect of Foreign Competition on Industry Level wages." Review of Worm Economics, 136, 2000, 522-38.

Hale, G., and C. Long. "What Determines Technological Spillovers of Foreign Direct Investment: Evidence from China." Mimeo, Yale University, 2006.

Head, K. "Comment," in Geography and Ownership as Bases for Economic Accounting, 59, Studies in Income and Wealth, edited by R. Baldwin, R. Lipsey, and J. Richardson, Chicago: University of Chicago Press, 1998, 255-58.

Hughes, H., and Y. Seng. Foreign Investment and Industrialization in Singapore. Canberra: Australian National University Press, 1969.

Iyanda, O., and J. Bello. "Employment Effects of Multinational Enterprises in Nigeria, Research on Employment Effects of Multinational Enterprise." Working paper 10, Geneva: International Labour Office, 1979.

Jo, S. "The Impact of Multinational Firms on Employment and Incomes: The Case of South-Korea." Worm Employment Programme Research Working Papers 12, Geneva: International Labour Office, 1976, 2-28.

Kamien, M., and Y. Tauman. "Fees Versus Royalties and the Private Value of a Patent." Quarterly Journal of Economics, 101, 1986, 471-92.

Kokko, A., R. Tansini, and M. Zejan. "Local Technological Capability and Productivity Spillovers from FDI in the Uruguayan Manufacturing Sector." Journal of Development Studies, 32, 1996, 602-11.

Lan, P., and S. Young. "Foreign Direct Investment and Technology Transfer: A Case Study of Foreign Direct Investment in North-East China." Transnational Corporations, 5, 1996, 57-83.

Landefeld, J., and R. Mataloni. "Offshore Outsourcing and Multinational Companies." Working papers, The Brookings Institution, WP 2004-06, 2004.

Langdon, S. "Multinational Corporations, Taste Transfer and Underdevelopment: A Case Study From Kenya." Review of African Political Economy, 2, 1975, 12-35.

LaPalombara, J., and S. Blank. Multinational Corporations and Developing Countries. New York: The Conference Board, 1979.

Lazear, E., and S. Rosen. "Rank-Order Tournaments as Optimum Labor Contracts." Journal of Political Economy, 89, 1981, 841-64.

Lim, D. "Do Foreign Companies Pay Higher Wages Than Their Local Counterparts in Malaysian Manufacturing?" Journal of Development Economics, 4, 1977, 55-66.

Lipsey, R., M. Blomstrom, and I. Kravis. "R&D by multinational firms and host country exports," in Science and Technology. Lessons for Development Policy, edited by R. Evenson, and G. Ranis. Boulder and San Francisco: Westview Press, 1990, 271-300.

Mason, R. "Some Observations on the Choice of Technology by Multinational Firms in Developing Countries." Review of Economics and Statistics, 55, 1973, 349-55.

Possas, M. Employment Effects of Multinational Enterprises in Brazil, Geneva: International Labour Office, 1979.

Rajan, R., and L. Zingales. "The Firm as a Dedicated Hierarchy: A Theory of the Origins and Growth of Firms." Quarterly Journal of Economics, 116, 2001, 805-51.

Ronde, T. "Trade Secrets and Information Sharing." Journal of Economics & Management Strategy, 10, 2001, 391-417.

Sepulveda, B., and A. Chumacero. La Inversion Extranjero en Mexico, Mexico, D.F.: Fondo de Cultura Economica, 1973.

Sharp, M., and M. Barz. "Multinational Companies and the Transfer and Diffusion of New Technological Capabilities in Central and Eastern Europe and the Former Soviet Union," in The Technology of Transition. Science and Technology Policies for Transition Countries, edited by D. Dyker. Budapest: Central European University Press, 1997, 95-125.

Sinani, E., and K. E. Meyer. "Spillovers of Technology Transfer from FDI: the Case of Estonia." Journal of Comparative Eeonomics, 32, 2004, 445-66.

Smarzynska, B. "Composition of Foreign Direct Investment and Protection of Intellectual Property Rights in Transition Economies," Mimeo, Yale University, 1998.

Sourrouille, J. "The Impact of Transnational Enterprises on Employment and Income: The Case of Argentina." ILO Worm Employment Program Research Working Paper 7, Geneva: International Labour Office, 1976.

Tirole, J. The Theory of Industrial Organization. Cambridege, MA: MIT Press, 1988.

United Nations. Worm Investment Report 1994. Transnational Corporations, Employment and the Workplace. United Nations, 1994.

Vaitsos, C. "The Role of Transnational Enterprise in Latin American Integration Efforts: Who Integrates and With Whom, How and for Whose Benefit?" Prepared for UNCTAD Secretariat and presented at Conference on TNCs and Economic Integration, Lima, Peru, June, 1974.

Wilkins, M. The Maturing of Multinational Enterprise: American Business Abroad from 1914 to 1970. Cambridge, MA: Harvard University Press, 1974.

Yu, C. "The Experience Effect and Foreign Direct Investment." Weltwirtsehaftliehes Archiv, 126, 1990, 561-79.

Zabojnik, J. "A Theory of Trade Secrets in Firms." International Economic Review, 43, 2002, 831-55.

Zabojnik, J., and D. Bernhardt. "Corporate Tournaments, General Human Capital Acquisition and Wage Dispersion." Review of Economic Studies, 68, 2001, 693-716.

DAN BERNHARDT and VLADIMIR DVORACEK *

* We thank Andy Williams for introducing us to the economic problem. We also thank the editor, Preston McAfee, and two anonymous referees for their helpful comments. Dan Bernhardt gratefully acknowledges financial support from NSF grant SES-0317700. Vladimir Dvoracek acknowledges funding from SSHRC. Dan Bernhardt acknowledges funding from SSHRC and the National Science Foundation grant, SES-031770. We thank George Deltas, Patrick Francois, and Andy Williams for helpful comments.

Bernhardt: Department of Economics, University of Illinois, 463 Commerce West, Champaign, IL 61820. Phone (217)244-5708, Fax (217)244-6678, E-mail danber@uiuc.edu

Dvoracek: Department of Business and Economics, University College of the Fraser Valley, 33844 King Road, Abbotsford, BC V2S 7M8. Phone (604)504-7441 (4702), Fax (604)855-7558, E-mail vlad.dvoracek@ ucfv.ca

(1.) Papers finding that multinationals generally pay higher wages than domestic firms in the same industries include Hughes and Seng (1969) for Singapore, Langdon (1975) for Kenya, Jo (1976) for South Korea, Sourrouille (1976) for Argentina, Vaitsos (1974) for Peru, Forsyth and Solomon (1977) for Ghana, Biersteker (1978) for Nigeria, LaPalombara and Blank (1979) for Malaysia and Nigeria, Sepulveda and Chumacero (1973) for Mexico, Mason (1973) for Mexico and the Philippines, Lim (1977) for Malaysia, Iyanda and Bello (1979) for Nigeria, and Possas (1979) for Brazil.

(2.) We thank Andy Williams for describing the strategies his Jamaican conglomerate employs and the reasoning behind their practices. In particular, Williams revealed both that his conglomerate paid higher compensation to reduce the dissemination of information about its superior technology (e.g., inventory and database system management) to competing domestic firms and that they divided informational access among several workers to eliminate the possibility that one or even a few informed workers could transfer the information to a competing domestic firm.

(3.) "It not only deters foreign direct investment in general, but it also discourages foreign investors from undertaking local production and encourages them to engage in non-manufacturing projects."

(4.) This result dates back to the industrial organization literature analyzing whether a monopolist will keep out potential entrants (Kamien and Tauman 1986; Tirole, 1988). Fosfuri, Motta, and Ronde (2001) and Glass and Saggi (1999) provide versions of this result in MNC contexts.

(5.) The MNC's concern is illustrated by this case study of reverse engineering, "Local staff working in the laboratories of two foreign affiliates manufacturing detergents discovered the contents of production by repeatedly trying the combinations. They then moved out to set up their own firms. In only a few years, more than ten small local firms were manufacturing detergent" (Lan and Young 1996).

(6.) Other strategic explanations for why more workers who are higher in the firm hierarchy receive wage premia include internal labor market tournaments (see, for example, Lazear and Rosen, 1981) and signaling a worker's match to competing firms. See, for example, Bernhardt and Scoones (1993). Zabojnik and Bernhardt (2001) show that internal labor market tournaments in which wages are determined endogenously by a competing firm's willingness to pay generate the firm-size/wage profile in the firm hierarchy that is found empirically.

(7.) From a case study of Shell Oil: "Shell provides know-how to its Russian partners where necessary, but does not pass on anything it regards as commercially sensitive ... (for a new plastic) Shell will supply the chemical intermediates for production, but the technology will be Russian. There is no question of the Russians either supplying the intermediates or obtaining access to the more up-to-date technology used ... in the United States." (Sharp and Barz 1997).

(8.) Zabojnik (2002) offers a related trade secrets explanation that focuses around the hierarchical structure of the firm. He supposes that middle level managers know the trade secrets pertaining to their hierarchical levels and to lower levels, but are not privy to higher level secrets and he finds that firms will pay wage premia only to those higher in the hierarchy because "higher level" secrets are more valuable. Other research on how firms can design their organization to avoid valuable knowledge spillovers to competitors via worker turnover include Rajan and Zingales (2001) and Ronde (2001).