- there are many Minsters' nephews, competing to buy risky assets.

In such a worst-case scenario for moral hazard, the owner of an intermediary will view investing in an asset as profitable if there is any state of nature in which that asset yields a return greater than the safe interest rate. At the same time, competition among intermediaries will eliminate any economic profits. The result must therefore be that the prices of assets are driven to their "Pangloss values": what they would be worth based, not on the expected outcome, but what would happen if we lived in the best of all possible worlds.

To see the implications of this setup, consider first a one-stage game, in which intermediaries initially compete to buy an asset with uncertain future payoff - call it land - and then learn what that payoff is. In particular, consider land that may yield a present value of future rent of either 100 (with probability 1/3) or 25 (with probability 2/3). In the absence of moral hazard, risk-neutral investors would be willing to pay a price of 50, the expected value of the land. In the extreme moral hazard regime we have described, however, each Minister's nephew will realize a profit in the favorable state of nature as long as the price is less than 100, and will simply walk away from the intermediary if the state of nature is unfavorable. So competition among the nephews will drive the price to its Pangloss value of 100.

Next consider a two-stage game. In period 1 land is bought. In period 2 initial rents are revealed, and land may be resold. Finally, in period 3 a second round of rents are revealed. It simplifies matters, without changing the substance, if we suppose both that rents are iid (specifically 25 with probability 2/3, 100 with probability 1/3) and that the safe interest rate is zero.

In an undistorted economy we can solve backwards for the price. The expected rent in period 3, and therefore the price of land purchased at the end of period 2, is 50. The expected return on land purchased in period 1 is therefore the expected rent in period 2 (50) plus the expected price at which it can be sold (also 50), for a first-period price of 100. This is also, of course, the total expected rent over the two periods. (In this example, the price of land declines over time, from 100 to 50, even in the undistorted case. This is merely an artifact of the finite horizon and should simply be regarded as a baseline).

Now suppose that intermediaries are in a position to borrow with guarantees. Again working backward, at the end of period 2 they will be willing to pay the Pangloss value of third-period rent, 100. In period 1 they will be willing to pay the most they could hope to realize off a piece of land: the Pangloss rent in period 2, plus the Pangloss price of land at the end of that period. So the price of land with intermediation will be 200 in period 1 - again, twice the undistorted price.

Our next step is to allow for the possibility of changes in the financial regime.

Let us continue to focus on our three-period economy, with random rents on land in periods 2 and 3. And let us also continue to assume that in the first period competition among intermediaries with guaranteed liabilities causes asset prices to be determined by Pangloss rather than expected returns. However, let us now introduce the possibility that this regime may not last - that liabilities carried over from period 2 to period 3 might not be guaranteed.

As a first step, let us simply posit that the regime change is exogenous - that from the point of view of investors there is simply some probability p that the government will credibly announce during period 2 that henceforth creditors of intermediaries are on their own. (Perhaps this reflects the election of a reformist government that is no longer prepared to tolerate "crony capitalism"; or perhaps the end of moral hazard is imposed by the International Monetary Fund).

Again, we work backward, and consider the price of land in the second period. If liabilities of intermediaries are not guaranteed, then nobody will lend to them (the moral hazard will remain, but its burden would now fall on investors rather than on the government). So intermediation will collapse, and the price of land will reflect only its expected return of 50. On the other hand, if intermediaries are guaranteed, the price will still be 100.

What about the price of land in the first period? Investors now face two sources of uncertainty: they do not know whether the rent in the second period will be high or low, and they do not know whether the price of land in the second period will reflected expected values or Pangloss values. However, as long as there is competition among intermediaries in the first period, the price of land will once again be driven to a level that reflects the most favorable possible outcome: rents of 100 and a price of 100. So even though this is now a multi-period world in which everyone knows that disintermediation and a decline in asset prices is possible, current asset prices are still set as if that possibility does not exist.

Finally, let us ask what happens when the change in regime is endogenous.

In reality, of course, throughout Asia's arc of crisis there has indeed been a major change in financial regime. Finance companies have been closed, banks forced to curtail risky lending at best and close their doors at worst; even if the IMF were not insisting on financial housecleaning as a condition for aid, the days of cheerful implicit guarantees and easy lending for risky investment are clearly over for some time to come. But what provoked this change of regime? Not an exogenous change in economic philosophy: financial intermediaries have been curtailed precisely because they were seen to have lost a lot of money.

This suggests that a more or less realistic way to model the determination of implicit guarantees is to suppose that they are available only until they have had to be honored (or more generally until honoring them has turned out to be sufficiently expensive - the criterion used in Krugman 1998b). In the context of our three-period example, this criterion can be stated alternatively as the proposition that creditors of financial intermediaries will be bailed out precisely once.

To see what this means, first suppose that in period 2 rents are disappointing - 25, not 100. Given the structure of our model, in the absence of intermediaries this should have no effect on the price of land at the end of the second period, since it does not change the probability distribution of future rents. But a less-than-Panglossian rent in period 2 means that creditors of intermediaries need to be bailed out in that period, and therefore that future creditors can no longer expect the same. So the intermediaries collapse, and the price of land drops from 100 to the expected rent 50.

Notice that this means that there is a magnification effect on the losses of the intermediaries established in the first period. The "real" news about the economy is that rents in period 2 were 25, not the hoped-for 100. But land bought for 200 will now yield only 25 in rents plus 50 in resale value, a loss of 125 rather than merely 75. The magnification effect is caused, of course, by the circular logic of disintermediation: the prospective end to intermediation, driven by the losses of the existing institutions, reduces asset prices and therefore magnifies those losses.

And now we come to the possibility of multiple equilibria. Suppose that in fact intermediaries have been lucky, and that second-period rents do turn out to be 100. Now if everyone then expects that the government will continue to guarantee intermediaries in the future, the land price at the end of the second period will also be 100. In that case no bailout will be needed; and so the government guarantee for intermediation will in fact continue.

But on the other hand, suppose that despite the high rents in the second period potential creditors become convinced that there will be no guarantee on newly incurred liabilities of intermediaries. Then they will not be able to attract funds, and the price of land in the second period will be only 50. That means, however, that intermediaries that borrowed money in the first period based on Pangloss values, including the Pangloss value of 100 for land sales, will require a bailout - and since the government's willingness to provide for bailouts is now exhausted, investors' pessimism is justified.

In short, our stylized little model appears to generate a story about self-fulfilling financial crises, in which plunging asset prices undermine banks, and the collapse of the banks in turn ratifies the drop in asset prices. But it is not the only such story.

3. Modeling the crisis II: Disintermediation and liquidiation

Even as the conventional wisdom has appeared to crystallize around the view that moral hazard and the resulting asset price inflation created the preconditions for the Asian crisis, some observers have violently disagreed. Recently Radelet and Sachs (1998) have argued that "The East Asian crisis resulted from vulnerability to financial panic, combined with a series of accidents that triggered the panic. Since we view the crisis as a case of multiple equilibrium, our hypothesis is that the crisis could have been avoided without any dramatic change in fundamentals." In effect, they argue that the pre-crisis asset values were more or less reasonable, and that it is the current deflated values that are an aberration - obviously an important point for assessing fire-sale FDI.

What kind of model could make sense of this view? The main contender is a "bank run" model along the lines of the classic paper by Diamond and Dybvig (1983). Such models, like the moral hazard model, attribute crisis to the collapse of financial intermediaries. However, financial intermediaries are now seen as essentially benign institutions, which reconcile the need for long-term commitment of capital to projects without short-term payoffs with the desire of individual investors to be able to withdraw funds on demand. The problem with such intermediaries, according to the model, is that they are vulnerable to self-fulfilling investor panics: if investors believe that sufficiently many other investors will try to cash in early, they will follow suit - and in so doing force destructive early liquidation of real investments.

In this section I offer a simplified exposition of a Diamond-Dybvig-type model. It is a highly abstract example, substantially harder to map into real-world developments than the simple Pangloss-collapse model of the previous section, but it does give us at least a first pass at the alternative view.

As in the crisis model above, we consider a three-period world. (Three periods is the minimum for financial crises, which must involve an initial investment and then something going wrong with expectations rather than or as well as actual earnings. While three-period models may seem artificial - why not an infinite horizon? - my own experience, in which the infinite-horizon Krugman 1998b actually preceded the finite-horizon Krugman 1998a, suggests that for exploratory theorizing simplicity wins out over the marginal gain in realism). In this case, however, there are real investment opportunities, of two kinds. Investors can put their wealth into a short-term asset - say, dollar treasury bills - that yields a known rate of return r. Or they can back investment projects that yield a higher rate of return, say , but that take two periods to mature. That is, one of these projects takes one unit of initial capital and transforms it into (1+h)² units of output in period 3, where h>r.

Crucially, we assume that for some reason it is not possible to sell a halfway-completed project to some other investor who will finish it. One can imagine a variety of reasons for this - perhaps some kind of lemons problem - but for the purposes of this model we simply take the nonmarketability as a given. Thus an investor who decides to liquidate a long-term asset in period 2 must actually scrap the real investment, realizing only a liquidation value v that we assume less than 1+r.

The need for financial intermediaries is created, following Diamond and Dybvig, by the need of individuals for liquidity. Each individual starts with one unit of capital, but does not know when he will want to consume: only after investing does he discover whether he wants to consume in period 2 or in period 3. This creates a dilemma: an individual who invests in a long-term project, then discovers a need for short-term consumption, is stuck with only the liquidation value. On the other hand, an individual who invests in the safe asset, then discovers that his consumption will take place in period 3, has foregone an opportunity to achieve a higher standard of living.

Figure 1 illustrates the dilemma of an individual investor in state space, with consumption in period 2 (if he turns out to be a period-2 consumer) on the horizontal axis, consumption in period 3 (if he turns out to be a period-3 consumer) on the vertical. If he invests only in the short-term asset, he will have consumption of 1+r if he turns out to be a period-2 consumer, (1+r)² (because he must then reinvest his capital short-term) if he turns out to be a period-3 consumer. On the other hand, if he invests only in the long-term asset, he will receive only v if he must consume in period 2, but (1+h)² if he consumes in period 3. And he can, of course, choose any convex combination of the two; say point A.

But now suppose that there is a financial intermediary which pools the capital of a large number of individuals, investing some in the short-term and some in the long-term asset. Ignoring for a moment the possibility of a bank run, such an intermediary can in effect exploit the law of large numbers to allow each investor to withdraw money at will, while still having a predictable aggregate withdrawal in period 2. To see the advantage of this pooling, suppose that the intermediary were to allow each contributor of capital to withdraw 1+r in period 2, and suppose that the intermediary knows that a fraction p of the population will turn out to be period-2 consumers. Then all the intermediary needs to do is put a fraction p of the funds it receives into the short-term asset, 1-p into the long-term asset; then each investor will expect to receive 1+r if he consumes in period 2, (1+h)² if he consumes in 3 - dominating the range of possibilities available without the intermediary. In general, of course, investors will choose some other point, such as B, on the budget constraint passing through that point, so that they will do even better.

So far so good. But such an intermediary is, as Diamond and Dybvig pointed out, potentially subject to a bank run. In our case this possibility arises because the liquidation value v is less than the promised payout to early withdrawers.

The point is straightforward. Suppose that for some reason - it does not matter what that reason is - investors who would ordinarily not have withdrawn their funds become convinced that many other such investors will attempt to withdraw their funds. Should investors who plan to consume in period 3 nonetheless withdraw funds in period 2, the intermediary will not have enough of the safe asset, and will therefore have to liquidate projects in mid-stream; and since the liquidation value is less than the promised payout, not all investors will in fact be able to withdraw their funds. And therefore the rumor of such a run will lead to a rush to withdraw funds by investors anxious not to be last in line. (Of course, in principle the possibility of a run should be taken into account in the initial investments and offers by the intermediary; one can justify the approach here by supposing that such a run is perceived as a very unlikely event).

As in the previous model, this gives us a story about a crisis that can be set off merely by self-fulfilling expectations. However, in the moral hazard model the pre-crisis state of affairs is fundamentally unsustainable; in effect, the asset market is in a "metastable" state, like a sandpile with a "supercritical" slope, and any small shock causes an avalanche - a slump in asset values toward their appropriate level. In the bank-run model, the state pre-crisis was reasonable, and could have been sustained, but was undermined by an unnecessary panic - which produces real costs due to the premature interruption of productive activities.

Both views can be given some support from anecdotal evidence, as argued below. But let us turn next to the implications of the two views for foreign direct investment.

4. The role of FDI

As Kindleberger (1969) pointed out long ago, foreign direct investment is essentially about transfer of control rather than movement of capital per se. Indeed, a quick look even at balance-of-payments measures of FDI for emerging market economies reveals that there is very little relationship between overall capital flows and FDI. Tables 1 and 2 show overall capital inflows and inward FDI for Argentina and Mexico from 1990-96; even though such balance of payments numbers tend to confuse internal capital transfers within firms (which behave like portfolio capital) with true changes in control, there is still a striking lack of correlation - or perhaps even an inverse correlation - between overall capital inflow and FDI.

Kindleberger's discussion suggested that in order to think about FDI we must therefore ask not why capital might flow into a country, but rather why some particular asset would be worth more under foreign than under domestic control. This in turn could reflect either higher expected earnings under foreign control, or a lower foreign cost of capital and hence a higher valuation on given earnings.

The interesting point is then that this dichotomy between two possible reasons for foreign ownership neatly matches our two different stories about financial crisis.

Consider first the moral-hazard view. The only reason for foreign ownership, in a pure model of that sort, would be that foreign firms could manage the assets better than domestic rivals, and therefore extract higher rents. Suppose, for example, that under foreign owners a given piece of land would yield 20 percent more than under domestic management. Then land would yield 120 in the good state, 30 in the bad, for an expected value of 60. Absent moral hazard, foreigners would outbid domestic investors for the physical assets.

But if domestic firms can borrow with implicit guarantees, they will be willing to pay higher prices than foreign owners despite their lower expected returns. As a result, foreign firms will be crowded out of the domestic market. (In terms of the balance of payments, this might well mean that domestic firms raise capital directly or indirectly by borrowing abroad).

The fire-sale FDI story is now clear. Suppose that in period 2 there is a collapse of the Pangloss regime, either because of actual bad news or because of self-fulfilling expectations. Then the prices that domestic firms are willing to pay for assets will drop - in our case from 100 to 50 - while foreign firms will still be willing to pay 60. So the result will be a transfer of ownership to the more efficient foreign firms.

In a pure moral-hazard version of the crisis, then, the drop in asset values is basically appropriate, and the transfer of ownership is an efficient move from the world's point of view: assets are being placed under the control of those who can use them best.

If we take a financial-panic point of view, matters look quite different. Suppose that foreign firms, unlike domestic investors during a panic, are not liquidity-constrained; they can borrow and lend at the safe rate r throughout. But they are less efficient at running domestic investment projects than domestic firms (which must be the case here, otherwise they would have made the investments in the first place). In the absence of a crisis the foreign firms will not get involved. But once there is a crisis, any foreign firm that can take over a project in midstream and do sufficiently well to earn a final return greater than v(1+r) - that is, any firm that is not liquidity-constrained and can earn more than the liquidation value by keeping the project in existence - will be in a position to buy the project from the crisis-stricken domestic intermediary. In this case there will truly be a fire sale. And such fire sales will typically transfer ownership to a foreign firm that is less efficient than the domestic firm, but now able to outbid domestic residents because of its superior cash position.

Thus our two alternative crisis stories seem to have opposite implications for the efficiency consequences of fire-sale FDI. If the drop in asset values really reflects the collapse of a moral-hazard-driven bubble, the reallocation of control is putting assets into the "right" hands; if it reflects an essentially arbitrary run on domestic intermediaries, it puts assets into the "wrong" hands.

Before we make too much of this distinction, however, we should notice that in either case the presence of foreign buyers will limit the actual fall in asset prices. In the moral hazard case, land falls from 100 to 60, not 50 - which means that the losses to domestic investors (and taxpayers) are less than they would have been if foreign acquisitions had been blocked. In the financial panic case, the willingness of foreign investors to buy half-completed projects means that the costs of liquidation are avoided, which is necessarily a gain that more than offsets the loss from the transfer into less efficient hands.

Finally, we should note a final point: the availability of potential foreign buyers may in itself be a stabilizing factor. Suppose that we take the pure financial panic model, but add a large number of potential foreign buyers who would be willing to pay a price high enough to pay off all investors, even if everyone decided to withdraw funds early. In that case investors, knowing that they had nothing to lose by failing to join in a run on the intermediary, would not in fact withdraw their funds unless they needed to consume in period 2 - which means that the possibility of a bank run would be eliminated. Or to put it a bit differently, the presence of potential foreign buyers would provide sufficient liquidity to make a liquidity crisis impossible. This suggests an unconventional additional payoff to opening one's economy to foreign direct investment: quite aside from any transfers of technology, managerial skills and so on, the mere potential for FDI may act as a stabilizer against the risk of domestic financial panics.

In any case, our analysis of both models seems to indicate that whether or not foreign investors are getting bargains - whether asset prices have fallen because they were initially overpriced or because they are now underpriced - given that a crisis has occurred, the "fire sale" of domestic companies is currently in the interest of the afflicted countries.

It remains interesting, however, to ask which of these stories we believe to be closer to the truth.

5. What kind of fire sale?

As long as we view the Asian crisis as a matter of collapsing financial intermediaries, it is easy to explain why that crisis should be accompanied by the sale of domestic assets to foreign firms. However, we have also seen that the efficiency implications of those sales - whether assets are being sold into or out of the "right" hands - depends on whether asset values are slumping toward or away from their appropriate levels.

What evidence do we have on the nature of the crisis? It seems hard to deny that there was a very significant moral hazard issue on the eve of the crisis. The role of "finance companies" in Thailand fits the "minister's nephew" story almost perfectly; in Indonesia many dubious investments (including the ambitious plans of a taxi company, which caused the spectacular failure of Hong Kong's Peregrine) involved members of the Presdident's family. In South Korea, all accounts suggest that the chaebol were engaged in reckless, ill-conceived expansion plans - with the industrial groups moving into businesses far from their core competences, and in many cases overseas ventures that seemed foolhardy even at the time - that would surely have come to grief even without the speculative attack. Indeed, a series of chaebol banruptcies took place even before the onset of speculation against the currency.

To some extent the "overborrowing syndrome" (as McKinnon calls it) shows up in balance of payments statistics. Figure 4 shows total capital flows into South Korea, inward direct investment, and outward direct investment. The striking points are both the very low level of inward investment given the size of overall inflow, and the remarkable position of Korea as a net direct investor abroad. Whatever the strengths of Korean management, this seems a peculiar position for a middle-income country; a parsimonious explanation of the pattern is that moral-hazard-driven lending both allowed Korean firms to crowd out potential inward investors and to pursue grandiose schemes abroad.

And yet while asset prices were surely overheated on the eve of the crisis, it is also easy to make the case that the crisis itself has overshot. The market values of Asian firms do seem extraordinarily low, even given their debt burdens. Moreover, while much of the real slump in Asia may be due to demand-side effects of plunging asset values, and to the effects of the high interest rates being used to defend currencies against hyper-devaluation, there is also considerable anecdotal evidence of a supply-side disruption of activity due to a breakdown of the credit system; this may be viewed as a version of the liquidation costs in our financial panic model.

So does the foreign purchase of Asian assets represents the transfer of control to efficient owners who were previously unable to buy at a reasonable price; or does it represent sales to inefficient owners who happen to have cash? Alas, probably some of both. What we need -surprise - is more research. Luckily, the issue of fire-sale FDI is not likely to go away anytime soon; even if the Asian crisis eases, its legacy of foreign ownership will be a contentious legacy for years to come.

REFERENCES

Flood, R. and Garber, P. (1984) "Collapsing exchange rate regimes: some linear examples", Journal of International Economics 17:1-13

Krugman, P. (1979) "A model of balance of payments crises", Journal of Money, Credit, and Banking 11: 311-325

Krugman, P. (1998a) "Bubble, boom, crash: theoretical notes on Asia's crisis", mimeo

Krugman, P. (1998b) "What happened to Asia?", mimeo

McKinnon, R. and Pill, H. (1996), "Credible liberalizations and international capital flows: the overborrowing syndrome", in T.Ito and A.O. Krueger, eds., Financial Deregulation and Integration in East Asia, Chicago: Chicago University Press.

Obstfeld, M. (1994) "The logic of currency crises", Cahiers Economiques et Monetaires 43:189-213.

Table 1: Capital inflows to Mexico, $ million

Table 2: Capital inflows to Argentina, $ million

Table 3: Capital flows, South Korea