Introduction

The world has been plagued by episodic deep downturns. The crisis that began in 2008 in the United States was the most recent, the deepest and longest in three quarters of a century. It came in spite of alleged “better” knowledge of how our economic system works, and belief among many that we had put economic fluctuations behind us. Our economic leaders touted the achievement of the Great Moderation.² As it turned out, belief in those models actually contributed to the crisis. It was the assumption that markets were efficient and self-regulating and that economic actors had the ability and incentives to manage their own risks that had led to the belief that self-regulation was all that was required to ensure that the financial system worked well, and that there was no need to worry about a bubble. The idea that the economy could, through diversification, effectively eliminate risk contributed to complacency—even after it was evident that there had been a bubble. Indeed, even after the bubble broke, Bernanke could boast that the risks were contained.³ These beliefs were supported by (pre-crisis) DSGE models—models which may have done well in more normal times, but had little to say about crises. Of course, almost any “decent” model would do reasonably well in normal times. And it mattered little if, in normal times, one model did a slightly better job in predicting next quarter’s growth. What matters is predicting—and preventing—crises, episodes in which there is an enormous loss in well-being. These models did not see the crisis coming, and they had given confidence to our policy makers that, so long as inflation was contained—and monetary authorities boasted that they had done this—the economy would perform well. At best, they can be thought of as (borrowing the term from Guzman (2014) “models of the Great Moderation,” predicting “well” so long as nothing unusual happens. More generally, the DSGE models have done a poor job explaining the actual frequency of crises.⁴

Of course, deep downturns have marked capitalist economies since the beginning. It took enormous hubris to believe that the economic forces which had given rise to crises in the past were either not present, or had been tamed, through sound monetary and fiscal policy.5 It took even greater hubris given that in many countries conservatives had succeeded in dismantling the regulatory regimes and automatic stabilizers that had helped prevent crises since the Great Depression. It is noteworthy that my teacher, Charles Kindleberger, in his great study of the booms and panics that afflicted market economies over the past several hundred years had noted similar hubris exhibited in earlier crises. (Kindleberger, 1978)

Those who attempted to defend the failed economic models and the policies which were derived from them suggested that no model could (or should) predict well a “once in a hundred year flood.” But it was not just a hundred year flood—crises have become common. It was not just something that had happened to the economy. The crisis was man-made—created by the economic system. Clearly, something is wrong with the models.

Studying crises is important, not just to prevent these calamities and to understand how to respond to them—though I do believe that the same inadequate models that failed to predict the crisis also failed in providing adequate responses. (Although those in the US Administration boast about having prevented another Great Depression, I believe the downturn was certainly far longer, and probably far deeper, than it need to have been.) I also believe understanding the dynamics of crises can provide us insight into the behavior of our economic system in less extreme times.

This lecture consists of three parts. In the first, I will outline the three basic questions posed by deep downturns. In the second, I will sketch the three alternative approaches that have competed with each other over the past three decades, suggesting that one is a far better basis for future research than the other two. The final section will center on one aspect of that third approach that I believe is crucial—credit. I focus on the capitalist economy as a credit economy, and how viewing it in this way changes our understanding of the financial system and monetary policy.

I. Three fundamental questions

Economic systems have always exhibited volatility, and economic theorists have long sought to describe and explain these fluctuations. In the United States, the National Bureau of Economic Research began as an attempt to characterize and date business cycles (for instance, with the work of Wesley Mitchell and his student, Simon Kuznets). In the middle of the twentieth century, a number of theories explaining why markets endogenously give rise to cyclical fluctuations were developed. Particular attention was paid to inventory cycles. But much of this literature was concerned with small oscillations, not the deep downturns that are the focus of this lecture.

A. What is the source of large perturbations?

The key questions are: Is the source of those major perturbations exogenous or endogenous? And how do economic structures and policies affect the depth and duration of these perturbations?

The standard models (referred to earlier in the introduction, and discussed at greater length below) assume that the source of the disturbances to the economy is exogenous. To me, it is clear that that is simply wrong: the credit and housing bubble that were at the center of the 2008 crisis was created by the market. The bubble was a market phenomenon. To assume that it was exogenous is to assume that there is nothing that we can do either to prevent the creation of a bubble or to curtail its size.

Equally troublesome, much of the literature referred to below assumes that the shocks to the economic system are technology shocks. As I explain below, when we look at many of the key economic fluctuations, it is hard to see any changes to technology that could have generated such changes.

By the same token, one of the reasons that the crisis spread so quickly and virulently beyond the borders of the United States was globalization⁶—the result of policies that had been pursued over the previous quarter of a century that resulted, for instance, in much greater financial integration. Again, it is simply wrong to assume that the shocks confronting a country are exogenous: even if the origin of shocks were exogenous, the extent to which a country is exposed to shocks is endogenous. (Indeed, in the case of developing countries, there is a wealth of evidence that most of the shocks to their economies come from outside their borders.)⁷

B. How can we explain magnitude of volatility?

The change in the physical state variables is typically small, and yet, in deep downturns, the change in output and employment are typically large. Consider the 2008 crisis. There was no destruction of physical or human capital, as often happens in war or in a natural disaster. Yet there were huge changes in the level of macro-economic activity, corresponding to large changes in the behavior of households and firms.

In many cases, shocks seem to have been amplified, rather than “buffered,” as suggested by traditional economic models, where price adjustments and inventories help absorb shocks and dampen the size of fluctuations.

Earlier literature (the Samuelson accelerator-multiplier model) tried to explain amplification through an investment accelerator: an increase in (expected) growth leads to an increase in the demand for investment goods (to produce the larger output), which then amplifies growth itself. But in the aftermath of the rational expectations revolution, it appeared hard to reconcile such behavior and the cycles to which the accelerator-multiplier model gave rise with rational expectations. For instance, in one variant of these models, at some point the economy reached full employment. This constrained growth. But prior to that constraint binding, investors should have realized that this would occur; and this would have constrained investment in these earlier periods.

But Greenwald and Stiglitz, in a series of papers beginning in the 1980s⁸, and Bernanke and Gertler (1990) showed that imperfect information in capital markets gave rise to a financial accelerator.⁹ In  particular, Greenwald and Stiglitz (1993a) explained how shocks to a firm’s equity base (resulting from a shock to the demand it faced, arising from any source) lead to reductions in both demand and supply, its demand for investment goods and its willingness to produce and its employment. There was an accelerator effect. The effects of shocks were amplified, not dampened. Greenwald and Stiglitz (1991, 2003a) show further amplification effects through similar disturbances to banks’ balance sheets, resulting in potentially large changes in the supply of credit and the terms at which it is made available through the banking system.

C. How do we explain persistence?

The third question is, how do we explain the persistence of the effects of a shock? How do we explain that after the onset of a deep downturn, output and employment remain low? The losses in GDP after the 2008 crisis have been far greater than those associated with the misallocation of resources before crisis, as the economy remained markedly below the crisis level and its pre-crisis trend growth for a long period.¹⁰ Yet, there are the same real assets (physical, human, natural capital) after the crisis as before.

More recently, in the aftermath of the 2008 crisis, many have referred to the excess leverage in the economy. But even debt shouldn’t matter: standard General Equilibrium theory says that there is a market clearing competitive equilibrium—with full employment. Debt simply is a matter of who has claims on society’s resources. There is nothing in standard equilibrium theory that says that there are patterns of ownership claims that are inconsistent with full employment.

It should be clear that the source of persistence is not in the capital stock or labor supply (though eventually, extended periods of unemployment do have effects on human capital and extended recessions and depressions do have effects on the stock of physical capital). There is persistence in unemployment—and explaining why the economy does not quickly return to full employment is key.

The Zero Lower Found is Not the Problem

Recent discussions have focused on the zero lower bound—the inability to lower nominal interest rates below zero. I believe the explanation does not lie (or does not lie just) in the zero lower bound (ZLB). The 2008 crisis was, in this respect, markedly different from the Great Depression, when prices were falling at an annual rate of 10%. Then a zero nominal interest rate meant a real interest rate of ten percent, enough to discourage investment.11 In the recent crisis, prices were increasing at around 2 percent, so the zero lower bound meant a negative real interest rate of 2 percent. Some argued that if one could increase inflationary expectations—say by committing to an inflation target of 4%—we could lower the real interest rate, say to minus 4 percent. Neither theory nor evidence suggests that such a change in the real interest rate would have restored the economy to full employment—even if one could have credibly committed to maintaining inflation at 4%. (Moreover, if the underlying problem was the inability to change intertemporal prices through monetary prices, there was an alternative: changing them through taxes, e.g. through a schedule of changing investment tax credits and consumption tax rates.)

But even if part of the failure of the economy to be restored to full employment did lie in the ZLB, it is critical to understand what gives rise to the ZLB problem. Clearly, the underlying problem is the lack of aggregate demand, the result of liquidity—the ability to spend income—being in the hands of those who don’t want to spend (or consume). Before the crisis, the bottom 80% of Americans were spending 110% of their income (on average). This was possible because banks were reckless in their lending; they believed too that the market value of houses represented their true value, so that lending against this collateral was safe. With the breaking of the bubble, the willingness of banks to lend decreased. So too did their ability, with the marked decrease in the value of their net worth. So too did the willingness of households and firms to borrow, as they saw their balance sheets contract. (All of these effects had been earlier analyzed in Greenwald and Stiglitz, 1993a, 2003.) As we shall see in the subsequent discussion, if models are constructed in which the economy will always be at full employment in the future and there are no distributional effects, then the possibilities for stimulating the economy today is constrained: more spending today can only be induced by changing intertemporal prices.¹²

Pseudo-wealth

Guzman and Stiglitz (2014, 2015a, 2015b) have provided another explanation: before the crisis, differences in views, e.g. about the likelihood of the housing bubble breaking, gave rise to bets (speculation), which led to the creation of pseudo-wealth—with both sides to the bet believing that they were going to win, both believed that they were wealthy, or more precisely, the aggregate perception of wealth was greater than true wealth. After the crisis, pseudo-wealth got destroyed. Indeed, there was even negative pseudo-wealth: borrowers may have believed that what they would pay, in expected value terms, to the lenders was greater than the leaders believed that they would receive.

Dynamics of adjustment

If the real wealth of the economy after the crisis is not much different from that before the crisis, if the market equilibrium depended only on critical aggregate state variables like the state of technology, capital stock, natural resources, and the amount of human capital13, then the full employment equilibrium after the crisis would not be much different from that before. Thus, the magnitude of adjustments in wages and prices that would be required to attain full employment would not be that different; and with some degree of wage and price flexibility, full employment should quickly be restored. But if the economy before the crisis was supported by a bubble or pseudo-wealth or if there are large distributional effects¹⁴ then there can be large changes in aggregate demand at the previously prevailing wages and prices; and to restore the economy to full employment would require large changes in wages and price.

But at least in the short run, these wage and price adjustments in a decentralized market economy can, be destabilizing. Unemployment, an excess supply of labor, leads to lower wages, which reduces aggregate demand. Such changes are redistributive, but the increase in spending out of the increase in profits is less than the decrease in spending by workers; and this is especially so if they worry about their future income and face borrowing constraints. Defaults and expected defaults associated with deflation—because debt contracts are not indexed—worsen banks’ balance sheets, and lead to a contraction in lending.¹⁵

Finally, an analysis of deep downturns has to be based on an understanding of why there is such suffering associated with them. If the decrease in hours worked were evenly shared, if there were full intertemporal and interstate smoothing, and if crises were not so persistent, then the social cost of economic fluctuations would be much less. Even in the deepest of downturns, say in Greece in the aftermath of the Euro crisis, output fell only around 25%. If the downturn lasted only a couple of years, a reduction of 25% for two years out of a working life of 40 years, at a low discount rate would amount to a reduction in lifetime income of only around 1.25%—and most individuals could easily absorb such a change. But, of course, none of these assumptions are accurate: the reduction in hours worked is not equally shared—some individuals become unemployed, while others experience only a slight reduction in hours worked; capital markets do not allow easy intertemporal smoothing; there is no private market for interstate smoothing—and unemployment insurance provides only limited insurance.¹⁶ And, if there is mean reversion, it is limited: with deep downturns, the losses in output often appear permanent, with growth in the future occurring off of a diminished basis, and with hysteresis effects even affecting growth rates (e.g. because of the diminution of human capital, a result of the lower job experiences of young people). Thus, any meaningful theory of deep downturns must come to terms with these critical aspects of markets and market dynamics; Lucas’ (1987) failure to do so explains why he has grossly underestimated the social costs of economic fluctuations.

II. Three strands of theory

In this part of the lecture, I will look at the three major strands of modern macro-economics—real business cycles (and related work); new Keynesian theories with rigid wages and prices; and alternative strands of New Keynesian economics, based on the work of Irving Fisher (1933) and Greenwald-Stiglitz. While each may have worked to help explain different historical episodes (oil price shocks, the Great Moderation and the early 90s), I will argue that only the third provides a convincing interpretation of deep downturns, such as the Great Depression and the Great Recession. After presenting these three strands, I will illustrate the differences by contrasting some key policy recommendations.

A. Real business cycles (and related work) (1st generation DSGE models) (RBC)

The assumptions underlying this work, which dominated macro-economic analysis in much of the latter part of the twentieth century, are sufficiently well known that I do not have to recount them here. I will focus my attention on those which I believe are central to these models’ failure to provide insights into deep downturns. Among the critical assumptions are: (a) exogenous shocks; (b) perfectly flexible wages and prices, which means that all markets clear—there is, in particular, always full employment (though there may be variations in the hours worked); (c) the perfectly flexible price system, together with inventories, dampens shocks (so that inventories are counter-cyclical); (d) all market participants have rational expectations; and there is common knowledge. There is, of course, still uncertainty; but there is no learning—in the way that learning is usually defined—and individuals have nothing to learn from each other.¹⁷ There are, of course, no betting markets—since individuals all agree about the probabilities of all events.

Full equilibrium, with no unemployment

Under the assumptions of the model, the economy is always in equilibrium—market acts as if there were futures markets going out infinitely far into the future and as if there were a complete set of Arrow-Debreu securities. Of course, neither of these assumptions are valid, and they make sense only in the context of one more very strong assumption: all individuals are identical. The representative agent model is, in this sense, more than just a simplifying assumption. With a representative agent model, it is conceivable that the individual could solve the infinite life-time model—could understand the transversality conditions and figure out the dynamic path consistent with them. With, however, a society with heterogeneous agents, each individual would have to solve a complicated infinite period general equilibrium problem—knowing the preferences of all other individuals and the production technology of all firms—obviating one of the basic presumed advantages of competitive markets, the ability to fully decentralize decision making (so firms only have to know prices and their own technology to know what to do).¹⁸

The Problematic Nature of the Representative Agent

But the representative agent model, as unrealistic as it is, brings with it three further problems which make it particularly unsuitable for analyses of deep downturns. First, there can’t be any interesting problems of information asymmetries, that is, unless individuals suffer from acute schizophrenia; and it is hard to reconcile that assumption with the previous assumption of full rationality.

Secondly, financial markets are largely irrelevant. After all, who is lending to whom? And by the same token, the nature of the financial instruments is irrelevant. There is no one with whom one can share risk. There is no “matching” of risks with risk preferences. Thus, by construction, these theories have little to say about financial crises. (It is possible, of course, to “solve” this conundrum by assuming a small open economy, borrowing from abroad.)

And thirdly, distribution is not important; distributive consequences of shocks and policies do not matter.

Policy implications

The model has strong policy implications. Because, under the highly stylized assumptions, markets respond efficiently to exogenous shocks, there are no market failures, and hence no role for government. There is no unemployment: variations in employment are just variations in the amount of time that individuals spend enjoying leisure. There is something patently unrealistic about this interpretation of unemployment: normally individuals seem happy when they are enjoying leisure; but there is ample evidence (both from surveys and behavioral symptoms, such as suicides) that individuals are not happy in economic downturns, and especially so in deep downturns.

Technology Shocks and Collective Amnesia

These are but a few of the large number of unconvincing aspects of these models. Earlier, I noted others (e.g. the assumption that the shocks to the economy are exogenous, not endogenous). I note just one more: The shocks to the economy that these focus on are technology (or supply) shocks. Employment decreases when labor becomes less productive. Of course, while it is easy to explain positive technology shocks (we discover a new way of producing something), it is much more difficult to explain negative shocks. Has there been a bout of collective amnesia, such that we forgot what we knew? Was there an epidemic of amnesia in 1929 and the years following, so far undetected in medical records?

In spite of the implausibility of such negative shocks, such models continue to be used to explain economic downturns. Thus Kydland and Zarazaga (2002) purport to explain Argentina’s 2001 crisis as the consequence of a negative technology shock within a RBC framework—as if there had been an epidemic killing off a large fractions of their engineers.

There is a further problem: typically, the timing is off. There was no major technology shock coinciding with any of the major economic downturns that could explain a perturbation of the magnitude observed. Of course, in models with forward looking agents with rational agents, as important as the shock itself is the news of the shock: when agents learn that technology in the future is going to be different from what they had expected. But this doesn’t help; in some ways it makes explaining large perturbations even more difficult: there was no major news, no sudden change in expectations about future technology surrounding the onset of any of the major economic downturns.

But there is a still further problem: typically, a shift in the supply curve to the left would lead to higher prices. (The oil price shock has this effect.) But deep downturns are marked by deflation, not inflation. Something else must be going on.

B. New Keynesian theories with rigid wages/prices (DSGE Generation II)

These theories keep most of the assumptions of the Real Business Cycle models, and thus most of their flaws. They represent the minimal change in the conventional competitive equilibrium model required to get sustained unemployment.

Thus, like the real business cycle theories, they assume that the shocks to the economy are exogenous (and still mostly supply side shocks). And they continue to assume rational expectations. As we noted earlier, there was no exogenous event, or even news about previously unanticipated exogenous events that would be occurring in the future, that could explain the sudden decrease in economic output in 2008 or the marked shift in the economic “equilibrium”.

Nominal rigidities

The main difference with real business cycles is the assumption of rigid wages and prices (and typically, it is nominal, not real rigidities, that are assumed). The consequence, of course, is that markets may not clear—in particular, the market for labor, so that there can be persistent unemployment.

But analyzing intertemporal models without market clearing is very difficult: in such models, the short side of the market dominates (that is, if demand is less than supply, equilibrium output is determined by demand). The regime that the economy is in at each date can affect the regime it is in other dates: they can be a large multiplicity of equilibria. Moreover, there are important income effects: an increase in aggregate demand at any date changes budget constraints, and has income effects at other dates, possibly causing a change in regime at those later dates. Beginning in the 1960s, a variety of versions of such models were studied by Gandmont, Barro and Grossman, and Stiglitz and Solow, amongst many others. But these models were not well suited to be a simple generalization of the RBC models.

Thus the new models typically assumed enough flexibility that markets cleared. The labor supply was assumed variable, so there were variations in the employment rate, though not in the unemployment rate.¹⁹ This by itself is a major limitation of these models in explaining deep downturns: a key issue, we suggested earlier, is the depth and duration of unemployment.

The major explanation of nominal rigidities was menu costs—a name which perhaps appropriately trivialized the explanation itself. With modern technology, reprinting the menu, to reflect changes in prices, is almost costless. More fundamentally, a shift in, say, the demand (or supply) curve for a perishable product necessitates either a change in output or a change in prices. A firm needs to look at these adjustments costs together, not separately. Typically, the costs of adjusting prices are an order of magnitude smaller than those associated with adjusting quantities. Thus, for perishables, menu costs simply cannot explain price rigidities.

For non-perishables, matters are different, for there is a third alternative: goods can be put into or taken out of inventories. One then has to compare those costs with the costs of adjusting prices or quantities. But this is a quite different problem than that posed in the menu cost literature. Greenwald and Stiglitz (1989) have argued that central to the analysis is uncertainty, including uncertainty about how other market participants will respond to a change in price or quantity. They explain how such uncertainty can explain nominal price rigidities.

Another strand of literature explaining nominal rigidities focuses on contract rigidities, e.g. in models with staggered contracts. Contracts may affect infra-marginal adjustments, but there is normally ample scope for marginal adjustments, and in “standard theories” (e.g. ignoring efficiency wage effects, which we discuss in the next subsection) those marginal adjustments should suffice to restore full employment, or at least to greatly reduce unemployment.²⁰ In short, contract theory on its own provides an incomplete basis for explaining the nominal rigidities that could give rise to a sustained high level of unemployment; but contracting rigidities can play an important role in a more general theory.

Still another market imperfection commonly introduced (in combination with the previous) is monopolistic competition, through the use of Dixit-Stiglitz preferences. There is an art form in introducing distortions; they have to be simple enough to be tractable, plausible, and yet not so simple that there would be a simple policy by which government could undo the distortion and return the economy to the RBC “ideal.” Because of the distortions, of course, the market equilibrium is not in general efficient, and so in principle there is a role for government policy.

But the problem in deep downturns is not price rigidity but deflation!

The focus on price rigidities seems strangely out of keeping with the major pre-occupation in deep downturns with deflation. Today, central bankers in many countries are worried about deflation. In Japan, the objective of inflation targeting is to convince the market that there will be inflation in the future—and to bring an end to the era of deflation, which it is believed has contributed to Japan’s weakness over the past two decades. In the Great Depression, prices fell at 10% a year—it is hard to call that “nominal rigidity.”

In early versions of New Keynesian models, financial markets were assumed to work efficiently; in later versions, financial frictions were introduced (see, e.g. Bernanke and Gertler (1990)), but these never came to play the central role that they played in the alternative strand of New Keynesian models discussed in the next sub-section—not surprisingly, because it is hard to have an interesting financial market in a representative agent model; and a major source of financial frictions are information asymmetries, which again can’t arise easily within a representative agent model. Of course, as we noted in the case of RBC models, one can introduce financial markets in a small open economy model; but even then, such models have to be carefully crafted to avoid distributional effects. If there is an open economy with equity markets, then changes within the economy can redistribute income to foreign owners, with macro-economic consequences.

Price and wage rigidities, as we have noted, were at the center of the model, so that as a result financial markets and frictions within those markets were largely a side-show, and not elaborated upon. There were not, for instance, banks, and there was not as a result an analysis of how monetary policy affects banks, and through the banks the supply of credit. Additional complications were introduced as Ptolemaic exercises, as it became increasingly evident that these models did not provide a good description of what was going on. In particular, financial markets and their imperfections do not, in this strand, play the central role that they do the models to which we now turn.

C. Alternative strands of New Keynesian

Even this brief description of the two dominant strands of macro-economic literature that I have just described should make it clear that these models do not provide a sound basis for understanding deep downturns. They are not constructed to provide answers to the three central questions posed in the first part of this lecture; and they do not do that.

There are several alternative (and in some ways complementary) strands to a literature which I should, for lack of a better term, refer to as finance based New Keynesian models. Much of my discussion will focus on the Fisher debt deflation model (revived by Greenwald-Stiglitz in 80’s, early 90’s)²¹, but I will have a few words to say about the important work of Minsky (1992).

I note that the DSGE models with fixed money wages and prices seem to have successfully appropriated for themselves the title “New Keynesian.” There is a natural sense in which Hicks’ price and wage interpretation of Keynes was extended to these models. But Hicks modelled only some aspects of Keynes’ General Theory. I believe that the models discussed here, where wages and prices are flexible, are as or more consistent with the central insights of the General Theory, and thus, while they have little to do with the New Keynesian model with fixed wages and price, they fully deserve the epithet New Keynesian.

Financial frictions

In this work, the financial sector is central: it is the source of “rigidities” that help explain the amplification and persistence of shocks, and it is the source of the perturbations to the economy.

The models focus on several financial market imperfections, several deviations from the perfect markets paradigm. There can be credit and equity rationing. Firms may not be able to borrow; the amount they can borrow may be limited by their collateral or their net worth; they may not be able to raise new equity—or more precisely, there may be a high cost from the dilution of existing shareholders from doing so. Much of the lending in the economy is mediated by institutions (banks), which themselves face credit and equity constraints, as well as regulatory constraints.

Moreover, financial (debt) contracts are typically not indexed to inflation (or other relevant shocks.) That means that when a firm experiences a negative shock to its demand, leading to lower prices, its revenues decrease, but what it is obligated to pay remains unchanged.²² Capital market imperfections explain why a firm cannot easily raise additional equity to offset the reduced equity resulting from the reduced revenues.

Moreover, risk markets—insuring firms against the important risks which they face—simply don’t exist.

Each of these market imperfections, in turn, can and has been explained in terms of imperfect, costly, and asymmetric information, transactions costs, the costs of verifying states of nature, costs of bankruptcy, and the difficulties and costs of writing and enforcing contracts.²³

Macro-economic externalities are pervasive

Whenever there is an incomplete set of risk markets and imperfections/asymmetries of information—that is always—markets are not (constrained) Pareto efficient. (Greenwald and Stiglitz, 1986.) As we explain, pecuniary externalities matter. Systematic micro-economic externalities add up to macro-economic externalities. These are pervasive: As Korinek (2011) and Jeane and Korinek (2010) point out, there can be, for instance, too much indebtedness or too much borrowing from abroad.

Balance sheets matter

In the standard neoclassical theory (which underlies both Real Business Cycle Theory and the New Keynesian rigid wage and price models), balance sheets don’t matter; neither does a firm’s cash flow. Investments are based on expectations—if the expected present discounted value of the profits associated with a project is greater than the cost, the project will be undertaken. There are perfect capital markets—there is no credit rationing—and perfect risk markets. Bankruptcy is simply not a matter of concern.

But if firms cannot insure themselves against risks that matter (for instance, that the price of their product will fall) and they cannot (easily) raise new equity, then they will act in a risk averse manner, and this affects every aspect of their behavior.²⁴ In particular, as a firm expands, it must borrow more; and that means that if the firm expands, there is a higher probability of bankruptcy. Now, part of the marginal cost of producing more is the marginal bankruptcy cost. So too, if the balance sheet of the firm shrinks, it must borrow more to produce the same amount that it otherwise would have produced; but, that means the bankruptcy probability is higher.

Hence, a shrinkage of the firm’s balance sheet means that firms will not wish to produce as much, employ as many workers, or engage in as much investment. (Greenwald and Stiglitz, 1993a.) There are effects both on demand and supply.

Redistributions matter

A corollary of these capital market imperfections is that redistributions have real effects—and this would be true even if all individuals have the same preferences, but is even more so, given that they don’t. Changes in prices that on the face of it might look simply redistributive have real aggregative effects. In a closed economy, for instance, the increase in the price of oil improves the balance sheet of oil producers and hurts those of users, and these two effects might appear to cancel—and they do in the standard perfect markets models. But if there are capital market imperfections, then the changes in, say, demand for investment and employment by those who gain may be far less than the reductions on the part of those who lose.

By the same token, one might have thought that not indexing debt contracts has only a distributive impact: the lenders gain (in the face of a negative shock) and the borrowers lose, and these effects too would just cancel. But they do not.

Banks as firms

In these models (unlike the two other strands just discussed) the financial sector is important, and financial institutions—banks in particular—matter. Banks are specialized institutions that gather and process information about borrowers’ credit worthiness and monitor debtors to make sure that they can pay back what is owed. There is a system of accountability, far better than with the shadow banking/securitization model: if a bank makes a loan and it turns bad, the bank bears the consequence. Hence the bank has strong incentives to do a good job on due diligence.

Banks borrow money (accept deposits) and, after ascertaining the credit worthiness of different potential borrowers, lend it out. The more they lend, the more they have to borrow, and this increases the probability that they go bankrupt (unless they were to increase their capitalization proportionately). The theory of the risk averse firm described in previous paragraphs applies just as well to the theory of the risk averse bank.

Accordingly, changes in bank balance sheets have first order effects. Changes in bank balance sheets affect the ability and willingness to lend. This affects credit availability and the terms at which credit is available. As Stiglitz and Weiss (1981) pointed out, credit rationing is not a phantasm. Indeed, everyone in the financial crisis talked about a liquidity crisis—firms and banks couldn’t get access to credit at any interest rate. Remarkably, in this, the onset of the deep downturn that was to become the Great Recession, even those who had seemingly believed in perfect markets were willing to talk about the lack of credit availability—something that was fundamentally inconsistent with the standard paradigm. The evidence of credit rationing was overwhelming.

Why deflation—or disinflation—a problem

The Fisher/Greenwald/Stiglitz theory explains why deflation can be a source of problems. The real value of what firms owe and have to pay on their debts increases. This diminishes their net worth from what it otherwise would have been. In effect, deflation increases a firm’s leverage. And the theory explains why leverage matters. At the same time, as we have noted, the gains to the creditors do not offset the losses to the debtors. In the representative agent model without financial frictions, the increase in leverage would not matter.

It should be noted that this argument even applies to disinflation—lower rates of wage and price inflation than were anticipated mean that the balance sheet is weaker than anticipated and, therefore, that output, employment, and the demand for investment goods are all lower.

Why the absence of indexing matters

This discussion should have made clear why the absence of indexing of debt contracts matters so much. The absence of indexing in itself can lead to lower employment and investment, in the face of bankruptcy costs; and this can help explain the amplification of shocks as well as persistence. For as firms borrow more, the probability of bankruptcy increases, and they take that into account in making production and investment decisions. An adverse shock which lowers net worth induces, as we have noted, less investment and production on the part of each firm—and when such individual decisions are aggregated throughout the economic system, they lead to overall lower output and employment. A small shock to net worth can have a macro-economic effect which is a multiple of the original shock. And because balance sheets cannot be restored instantaneously (indeed, it typically takes a considerable period of time, especially as production is scaled back), and because firms will be reluctant to raise new equity (even if they can), the effects of a shock may last a long time.