As far as I can tell, giving a keynote speech differs from presenting a paper in two ways. First, keynote speeches have big thoughts, which is another way of saying that they do not have new results. Second, there are no discussants for keynote speeches. This means that nobody rudely points out mistakes or the lack of results in a keynote speech. These two features of keynote speeches have some obvious advantages and some inconveniences. To get full benefit of these advantages, I will focus on a topic of considerable importance about which much has been written. Since so much has been written on that topic, you should not expect me to derive new results or provide new estimates. Furthermore, since there is no discussant to point out the problems in my speech, I will be as provocative as I dare to be, given my conservative Swiss background.Lots of papers have been written about the cost of capital. You might therefore wonder why more should be said on that topic. The answer is that we know both a lot and very little about the cost of capital. Whole books have been written on how to compute the cost of capital for individual firms within the U.S. Yet, careful reading of these books reveals that most of what we know about the cost of capital is rather shaky. Here are some examples of problems with our knowledge:
- The cost of capital of an all-equity firm. Reading the textbooks, this would seem to be a straightforward exercise. Yet, all the textbooks use the static CAPM. At this point, can we really believe that the CAPM used in conventional computations of the cost of capital has any empirical relevance? Countless papers reject the static version of the CAPM used in all textbooks.
- Geometric or arithmetic risk premium. A question as simple as whether to use a geometric or an arithmetic risk premium would seem to have a straightforward answer. However, the only obvious fact in the literature is that whatever risk premium an author uses, he believes in it very strongly. A sampling of four recent valuation books shows that half of the books advocate using the arithmetic risk premium and the other books argue that the geometric risk premium should be used.
- The role of taxes. Existing books tend to assume that the tax shield of debt is equal to the tax rate times the face value of the debt. This effectively corresponds to the case where the marginal tax rate is equal to zero for individuals. If the marginal tax rate is zero for individuals, why is it that municipal bonds have a lower before-tax expected return?
- The international dimension. Most computations of the cost of capital use the CAPM applied to individual countries as if capital markets were not internationally integrated. As I show in the first issue of European Financial Management, using the one-country version of the CAPM creates a bias in estimates of the cost of capital.
For countries that are well-integrated in international financial markets, we know that using the international CAPM is better than using the domestic CAPM. We have little sense, however, how to compute the cost of capital for countries that are only partially integrated in international capital markets.
In this keynote speech, I will assume away all these uncertainties about how to measure the cost of capital and ask the question: Does the cost of capital differ for firms located in different countries? It will be clear that these uncertainties do not matter in answering this question. In the process of answering this question, I will emphasize a key difficulty in the way we think about the cost of capital which matters strongly when making cross-country comparisons of the cost of capital. The cost of capital is the hurdle rate that a project must satisfy for owners of a firm to not suffer a wealth loss if the project is taken. However, the neoclassical definition of this hurdle rate completely ignores agency costs. I will argue that projects that satisfy the neoclassical hurdle rate can destroy shareholder wealth in firms with significant agency costs of managerial discretion. Consequently, the neoclassical hurdle rate is not the appropriate hurdle rate once agency costs are taken into account.
Before focusing on agency costs, I want to answer the question of whether the cost of capital differs across countries using a traditional approach and show why this traditional approach provides an unsatisfactory solution to the question I am trying to answer.
To answer the question of whether firms located in different countries have a different cost of capital, it is useful to think about two firms competing for a specific project. Let's look at a U.S. firm and a Japanese firm thinking of building a car factory in Germany. The traditional way of evaluating a possible difference in the cost of capital for these two firms would be to assume that the stream of cash flows from the car factories are the same. The cost of capital would differ for the two firms if the present values of these cash flows differ.
In an international setting, assuming that the cash flows are the same for two projects requires a numeraire currency. In this case, we could assume that the DM cash flows are the same. Suppose first that the DM cash flows are non-stochastic. In this case, the Japanese firm could sell all the cash flows forward for a fixed Yen amount. The present value of the cash flows for the Japanese firm, ignoring taxes, would simply be the present value of riskless Yen amounts accruing at future dates. The Japanese firm could therefore use Japanese interest rates on Japanese default-free debt to compute the present value of the cash flows. The American firm could sell all the cash flows forward against U.S. dollars and discount the certain dollar proceeds at the riskless dollar rates.
If the cash flows are riskless in DMs, it is immediately clear that the U.S. and the Japanese firm value the cash flows differently only if interest rate parity does not hold. If interest rate parity holds, the DM present value of the cash flows is equal to the Yen present value of the cash flows multiplied by the price of the DM in Yen. Furthermore, the present value of the cash flows in Yen is also equal to the present value of the cash flows in dollars multiplied by the price of the dollar in Yen. Hence, with interest rate parity, the value of a project is the same for the Japanese firm and for the U.S. firm.
To understand how departures from interest rate parity create differences in the value of the project for firms located in different countries, consider the case where the Yen interest rates are lower than implied by interest rate parity with respect to American and German interest rates. In this case, the project is worth more for the Japanese firm and the Japanese firm would undertake projects that have a negative net present value for the American firm. The problem with this scenario is that interest rate parity can fail to hold only if there is some departure from the assumption that capital markets are internationally integrated, since otherwise there exist arbitrage opportunities. One way that interest rate parity could not hold between Japan and Germany so that Japanese interest rates are lower is that there is some risk that foreign funds cannot be repatriated costlessly for the Japanese firm. In this case, the interest rate parity arbitrage of borrowing in Japan, investing the funds in Germany, and selling the proceeds of the investment forward against Yen, is no longer riskless since the funds invested in Germany might not be repatriated without costs. In this case, the lower interest rate of the Japanese firm is offset by the repatriation risk. The point of this is that departures from perfect markets that make possible apparent differences in the cost of capital are accompanied by costs that may more than offset these apparent differences in the cost of capital.
American businessmen are fond of arguing that the cost of capital is lower in Japan either because nominal interest rates or real interest rates are lower in Japan. It is important to note that our simple analysis holds irrespective of whether nominal interest rates are higher or lower in Japan relative to the U.S. and that real interest rates play no role in our analysis. Keeping the cash flows constant, an increase in interest rates decreases the present value of cash flows irrespective of whether this increase in interest rates takes place because of an increase in the real rate of interest or because of an increase in the expected rate of inflation. If interest rates increase in Japan in this formulation, it must be either that the forward price of the Yen in dollars falls or that interest rates increase in the U.S. also.
So far, we have ignored risk. Suppose now that the DM cash flows are random. Could it be that American investors require a different risk premium for bearing the risk of these cash flows than Japanese investors? Not if the Japanese and American capital markets are fully integrated. Integrated markets are markets in which investors have access to the same investment opportunities. Hence, if a Japanese investor earns the same on U.S. investments in dollars as a U.S. investor and if an American investor earns the same in Yen on Japanese investments as a Japanese investor, the Japanese and American capital markets are fully integrated. With fully integrated capital markets, securities traded in different countries are priced in the same way. Stated differently, the required dollar returns of Japanese stocks satisfy the same pricing formula as required dollar returns of American stocks or, alternatively, the required Yen returns of American stocks satisfy the same pricing formula as Japanese stocks. The empirical evidence is that, although the Japanese and American markets are not fully integrated, they are sufficiently integrated that it is hard to reject the hypothesis of perfect integration. In a 1992 paper published in the Journal of Financial Economics, K. C. Chan, Andrew Karolyi and I could not reject the hypothesis that the world capital asset pricing model holds when investigating the expected returns on the S&P500 and the Nikkei 225. In general, the empirical evidence is mixed on whether the Japanese and American capital markets are integrated.
For those who are not convinced by asset pricing tests, it is worth mentioning that Japanese and American firms have raised funds side by side in the offshore markets. It is hard to believe that on these markets, Japanese and American securities denominated in the same currency would have been valued differently by investors. However, the fact that both Japanese and American firms were constantly raising funds on the Eurobond market is also evidence that the cost of capital for Japanese firms cannot have been very different from the cost of capital of American firms. If the cost of capital for Japanese firms had been lower than the cost of capital for American firms at the margin, one would not expect the Japanese firms to raise funds offshore competing with American firms.
A natural question to ask, though, is the following one: If the Japanese and American capital markets are so well-integrated, why is it that Japanese investors hold so few American stocks and American investors hold so few Japanese stocks? Much of the literature has emphasized the existence of barriers to international investment in the form of deadweight costs. If all investors face similar deadweight costs and these deadweight costs lead investors to hold portfolios that are almost autarchy portfolios, one would expect costs of capital to differ substantially across countries and to be lower in large countries because investors can diversify risks more in large than in small countries. However, it need not be the case that all investors face the same deadweight costs. At this point, the literature has not explored extensively the implications of deadweight costs that are heterogeneous within countries. In a recent paper with Walter Wasserfallen, we showed that heterogeneous deadweight costs of a particular type could help to understand the existence of shares restricted to domestic investors in Switzerland.
The conclusion of this analysis is that, ignoring taxes, it seems unlikely that a proper comparison of hurdle rates between Japanese and U.S. firms would lead to the conclusion that Japanese firms have a significant pre-tax cost of capital advantage at the margin. Could the introduction of taxes affect this conclusion? To consider the effect of taxes, let's start with an extreme case. Suppose that there are only corporate taxes and no personal taxes. There are two countries. One country is the U.S. with a marginal corporate rate of 36% and the other is Liechtenstein with a marginal tax rate of 0%. One would think that in this case taxes would matter a great deal. Suppose that a U.S. firm and a firm from Liechtenstein are considering building a car factory in Germany. In this case, cash flows repatriated by the American firm would be taxed at 36%, so that one would think that that firm would reject projects that the firm located in Liechtenstein would undertake. This need not be the case, though. It is easiest to see this for the case where the cash flows are riskless in DM. In this case, the American firm could completely finance the project with debt so that it would not pay corporate taxes on the project. It would immediately follow that the American firm and the firm in Liechtenstein would have very different capital structures but would face the same cost of capital. Obviously, if the cash flows are risky, the American firm cannot have an all-debt capital structure, but it may also have non-debt tax shields. The point of this is that apparently large differences in tax rates may have little impact on the cost of capital because firms behave differently under different tax codes.
So far, we have compared the cost of capital between Japan and the U.S. in a traditional way using the tools of neoclassical economics. Based on this analysis, there is no good reason to believe that there should be important differences in the hurdle rate used by Japanese and U.S. firms for comparable projects. Is this, however, the correct way to assess differences between firms? This traditional approach assumes that firms only take positive NPV projects and that if no such projects are available, the firm returns the cash flows of the project to the shareholders. If managers always maximize shareholder wealth, the net present value of a project corresponds exactly to the increase in the value of the firm resulting from the adoption of the project. Hence, under this scenario, the cost of capital is computed correctly.
Suppose, however, that managers do not act to maximize shareholder wealth but instead maximize their own utility. In this case, their actions will depend on their compensation contract, but also on other factors. For instance, they might care about the size of the firm and might prefer to invest cash flow in bad projects rather than give it back to the shareholders. In this case, the computation of the net present value of a project must be modified. This modification can be made in two different ways. First, the numerator of the NPV computation can be changed to reflect the use of the cash flows from the project. Second, the numerator of the NPV calculation can be left unchanged but the denominator can be changed to reflect the agency costs of managerial discretion. The first approach is similar to the APV approach in valuation. It amounts to adding to the usual NPV calculation an additional negative term which reflects the loss in value resulting from agency costs. The second approach is similar to the WACC approach. It amounts to starting from the value of the project gross of agency costs, so that the project is the same across firms with different agency costs, and then incorporating into the discount rate the impact of agency costs in the same way that the WACC approach incorporates in the cost of capital the tax shield of debt. We will focus on the second approach and call the resulting cost of capital the agency cost adjusted cost of capital.
It is important to understand that the agency cost adjusted cost of capital is relevant for existing shareholders but not for buyers of new securities issued by the firm. When the firm raises funds in the capital markets, the price that security buyers are willing to pay is a function of the cash flows they expect to receive from the new securities. For a given distribution of cash flows, new investors do not care whether these cash flows come from the profits of new projects or from the existing shareholders who lose money because management invested in poor projects. Hence, the cost of capital computed in the traditional way exactly reflects the rate of return required by new investors. However, a firm could invest in a project that has a positive net present value if the cost of capital is computed in the traditional way and yet existing shareholders could suffer a large loss. The cash flows of a project discounted at the agency cost adjusted cost of capital exactly correspond to the wealth effect of the project for the existing shareholders.
At this point, it is useful to look at an example. Consider a firm where management is intent on increasing firm size even if doing so is not advantageous to shareholders. This firm has a project that has a positive NPV with a neo-classical cost of capital. However, suppose that for this firm it seems unlikely that it will again have a positive NPV project. Hence, as the cash flows from the existing project accrue, the managers will invest them in marginal projects which will not benefit shareholders. This means that even though the project in which management is investing now has a positive NPV, shareholders capture at most a fraction of that NPV. In this case, shareholders would want managers to reject projects that would decrease the value of the existing shares if undertaken. The existing cost of capital calculation, because it ignores agency costs, does not provide the decision rule for new projects that maximizes firm value.
The traditional NPV calculation assumes that managers maximize shareholder wealth. We know that management maximizes its own interests and that its interests do not always coincide with the objective of maximizing shareholder wealth. For management to be effective, it has to have some freedom in choosing investments. However, management is concerned about perpetuating and growing the firm under its control. For firms with good investment opportunities, this is not a problem: investing benefits management and benefits shareholders. For firms with poor investment opportunities, though, management wants to reinvest cash flows even when shareholders would prefer to receive dividends instead. The traditional NPV analysis ignores these costs of managerial discretion and views each project separately. With the traditional NPV analysis, management can always raise funds to invest in good projects, so that future investment does not depend on current cash flows, and management never wants to undertake poor projects.
Consider a firm where these agency costs are important and where management wants to undertake a new project and finance it with equity. Let's assume that the project has a positive NPV. If there are no agency costs of managerial discretion, management could raise equity to finance the project and the value of the firm would increase. With agency costs of managerial discretion, management could raise equity and see the value of the existing shares fall because the market anticipates that there is some chance that management will invest the cash flows from the investment poorly. In this case, raising equity will be expensive for the firm and, because of this, management may give up the positive NPV project when another firm with less significant agency agency costs of managerial discretion might take the same project.
With this perspective, American managers could find that outside funds are expensive when Japanese investors do not find these funds expensive. Japanese investors would find their cost of capital lower not because their neoclassical cost of capital is lower but because they have lower agency costs of managerial discretion due to differences in the organization of firms and/or investment opportunities. Let's go back to our example of a car factory in Germany. If the car factory belongs to an American firm and generates hefty cash flows, this firm is likely to invest the cash flows in a diversifying acquisition if it does not have good investment opportunities. Although frequent, such actions are likely to reduce shareholder wealth. We know that diversified firms are valued less than firms that focus on single activities. For instance, Larry Lang and I showed in a recent paper that the Tobin's q of diversified firms is significantly lower than the Tobin's q of portfolios of specialized firms whose activities match those of the segments of the diversified firms.
Anecdotal evidence of the reluctance of management to pay out cash flows in the form of dividends is probably best provided by the example of Unocal. The firm in the early 1980s had large cash flows that it was wasting on poor projects. When a director of Unocal told the CEO Hartley "Fred, we are making a lot of money, why don't we raise the dividend?", Hartley replied "You're crazy. Why would we give a bunch of money to people we don't even know?"
Let's look now at the case of the Japanese firm. There are some good reasons to suspect that, at least for a good part of the post-World War II period, a Japanese firm would have been less likely to waste funds. Part of the reason is simply that Japanese firms had spectacular growth opportunities for a long period of time. However, there were also important organizational reasons. To understand these reasons, we discuss successively how capital structure, large shareholders and the market for corporate control affect the agency costs of managerial discretion. In each case, we show how Japan and the U.S. differ and the impact of these differences on the agency costs of managerial discretion.
First, the agency costs of managerial discretion depend on a firm's capital structure. In a paper published in the Journal of Financial Economics, I show that there is an optimal capital structure that involves debt when there are agency costs of managerial discretion. The argument is straightforward.
When there are agency costs of managerial discretion, there is a benefit to debt, but there is also a cost. Suppose that a firm has an unexpectedly low cash flow so that after servicing the debt, there are no funds left over for investment. Management always wants funds to invest, even when it does not have good projects. Hence management will have a hard time convincing outsiders to provide funds if outsiders are less well-informed than management. With too much debt, therefore, some good projects may not be undertaken because outsiders do not believe that management cannot finance them from internal funds.
In this model, debt reduces investment in all states of the world: this is good when cash flow is large and good investment opportunities are limited, but bad when cash flow is small compared to good investment opportunities. Hence, with this model, there is an optimal amount of debt which increases with expected cash flow and falls with the size of the good investment opportunities.
For much of the post-World War II period, Japanese firms have been more highly levered than American firms. Perhaps more importantly, though, in the U.S., debt is more likely to be public debt which is extremely difficult to renegotiate. In Japan, debt tends to be bank debt which can be renegotiated. Further, whereas a firm can convey private information to banks, it cannot do so to public bondholders. Hence, in Japan, a firm can have more debt and yet bear less of a cost from being highly levered. In addition, Japanese banks also hold equity, so that their incentives are more in line with those of shareholders. Since they acquire a lot of information about the firm, they are also in a position to affect managerial actions and to make it difficult for management to undertake poor projects. All this means that the capital structure of firms in Japan is more effective at regulating managerial discretion: more debt limits managerial discretion more, but the fact that there is more bank debt means that the adverse effects of debt are mitigated.
Next, we consider monitoring by shareholders. Shareholders can investigate the actions of managers and prevent bad investments from taking place. However, for shareholders to be effective in this role they have to be large and/or organized. In the U.S., shareholders are typically small and because of laws and regulations, they cannot organize or collude.
In Japan, firms are controlled by large shareholders who act as a group. These shareholders have incentives to monitor: They can share the monitoring costs and have large enough holdings that they benefit from monitoring. In addition, the large shareholders in Japan typically have significant business ties with the firm, so that their cost of monitoring is cheap.
Hence, when we look at the capital structure, the debtholders or the shareholders, it is clear that Japanese managers are monitored more closely than American shareholders. Knowledgeable observers will immediately argue that Japanese managers are not restrained by the market for corporate control whereas American managers are. This is correct and it is the case that, in the U.S., managerial discretion is limited by the market for corporate control. Management will be careful to not misbehave in such a way that outsiders would make money by acquiring the firm. In Japan, such hostile takeovers don't take place because the firm is controlled by large shareholders who would not sell their shares. The problem is that hostile takeovers are extremely costly and difficult. The extremely large premia paid in the case of hostile tender offers shows that the value increase brought about by a takeover must be extremely large for it to take place. Hence, there can be a substantial loss in firm value resulting from agency costs of managerial discretion before management is threatened and, if the firm is sufficiently large, management may never feel threatened.
In summary, therefore, despite the lack of an active corporate control market, our analysis indicates that the agency costs of managerial discretion should be lower in Japan, at least when all the mechanisms we just discussed were fully in place. In particular, the role of banks has become less important in Japan because of deregulation.
Where could we find empirical evidence that makes it possible to assess the analysis of the agency costs of managerial discretion in the U.S. and Japan just discussed? There are lots of directions one could go to generate empirical evidence. Unfortunately, most of the existing empirical evidence on corporate finance in Japan has little to say on this topic. Some of that evidence confirms our interpretation of the role of banks and large shareholders, but it fails in showing that as a result agency costs of managerial discretion are less in Japan. In the spirit of being provocative, I will focus on evidence in two recent papers that I co-authored.
When a firm sells equity, managerial discretion is increased. Management gets money that it does not have to pay back. If management chooses to waste the funds raised, this will affect the value of the existing shares - the new shareholders will buy shares so that their investment is a fair investment. Hence, if agency costs of managerial discretion are large, one expects a fall in the value of existing equity if a firm issues new shares. Let's look at the evidence. In the U.S., if a firm issues equity, on average, the value of the existing equity falls by 2.75%.
Let's now turn to the evidence for Japan. One piece of evidence is provided in a paper by Shalheim and Kato (1991). They show that the announcement of an equity issue for a Japanese firm increases the value of its equity by 0.7%. In two recent papers, various co-authors and I provide further evidence on the case of Japan. In one paper, we look at offshore warrant bond issues.
I realize that interpreting the abnormal returns associated with equity and convertible issues as evidence of the magnitude of the agency costs of managerial discretion brings me far from the mainstream of finance research. The most popular model for understanding the abnormal returns associated with equity and convertible issues is the model presented by Myers and Majluf (1984). In that model, equity and convertible issues are interpreted by the markets as evidence that the issuing firm is overvalued. In the view that I expressed, it is also the case that the firm is overvalued when it issues equity or convertible debt. However, the source of overvaluation is that the value of the firm with the equity issue is not necessarily the value of the firm before the issue announcement plus the value of the equity issue. This is because an equity issue provides management with free cash flow which it can misuse. The greater the potential of misuse, the more adverse the effect of the equity issue on firm value since the buyers of the new equity will only buy it at a price that reflects the potential for misuse. This view of the adverse stock-price reactions to equity issues and convertible issues is consistent with the empirical evidence.
When comparing the cost of capital in Japan and the U.S., one reaches different conclusions if one looks at the required rate of return on securities issued to finance a new project or at the required rate of return on a new project that insures that shareholders would not lose from the acceptance of the project. The rate that insures that shareholders do not lose is typically higher than the neoclassical hurdle rate because the neoclassical hurdle rate presumes that the cash flows will not be used by management to further its own objectives. Yet, when a firm faces the prospect of issuing securities, its management must contemplate the effect on existing shareholders of the security issue. Hence, management may find capital expensive if its existing shares fall substantially if it tries to issue new shares and may therefore decide to abstain from issuing new shares. In this sense, management will find capital expensive and in this sense, it may well be that capital was more expensive in the U.S. than Japan.
To conclude, therefore, the neoclassical cost of capital and the agency-adjusted cost of capital give different answers to the question of whether the cost of capital differs between Japan and the U.S. However, whereas the traditional debate on the cost of capital suggests that only macroeconomic policies can be used to equalize the cost of capital across countries, any discrepancy between the neoclassical cost of capital and the agency-adjusted cost of capital can be decreased by actions of management to commit themselves to a course of action that maximizes shareholder wealth.
