| English
"Куда идет мир? Каково будущее науки? Как "объять необъятное", получая образование - высшее, среднее, начальное? Как преодолеть "пропасть двух культур" - естественнонаучной и гуманитарной? Как создать и вырастить научную школу? Какова структура нашего познания? Как управлять риском? Можно ли с единой точки зрения взглянуть на проблемы математики и экономики, физики и психологии, компьютерных наук и географии, техники и философии?"

«Towards a General Theory of Deep Downturns» 
Joseph E. Stiglitz

Balance sheet recessions

Thus, the Greenwald-Stiglitz (1993a) model provides an explanation of balance sheet recessions. (Many have suggested that the 2008 recession is a balance sheet recession.25) There are three critical aspects of a balance sheet recession: (a) A shock which leads to an unanticipated worsening of the balance sheets of many firms in the economy. (b) The response to the worsening of the balance sheet: a contraction in economic activity; and (c) A slow “healing” of balance sheet—necessary and sufficient for the restoration of the economy to full employment.

The shock can be a price level shock (contractionary monetary policy leads to overall lower prices, meaning that the real value of what the firm owes has increased); a demand shock to a particular category of goods (individuals shift away, at least temporarily, from buying large cars); or a supply shock (the price of oil increases.)

Greenwald and Stiglitz then describe the response to the worsening of the balance sheet. There are both demand and supply side effects. Firms will not wish to produce as much, employ as many workers, or engage in as much investment. As we have noted, all of this is quite different from the standard neoclassical world, where there are neither the demand nor the supply side effects; and the difference is even greater with the representative agent models, where there aren’t any redistributive effects. Here, a relative price shock which strengths the balance sheet of some firm and weakens that of another by exactly the same amount still has real effects on the aggregate equilibrium.

Central to the notion of a balance sheet recession is the idea that the adverse impacts on the balance sheet cannot be undone overnight simply by raising more equity. The equity in the firm has to be rebuilt, through retained earnings, and this is a slow process. And that is a key reason for the persistence of downturns.

But, as I discuss below, truly deep downturns are more than balance sheet recessions: they can persist even after balance sheets are largely restored.

Risk perceptions and confidence

In the standard model, confidence and the extent of uncertainty matters little: because risk is diversified throughout the economy (through perfect insurance markets), what matters is the mean return. But in these New Keynesian models, risk matters a great deal. A mean preserving increase in risk increases the probability of bankruptcy, leading to results similar to those that arise from a balance sheet shock.

This provides another channel through which amplification and persistence occurs: after the initial adverse shock, unless active government policies prevent a deep downturn, uncertainties increase. No firm can be sure whether its customers or suppliers will survive. A bankruptcy of any of its trading partners poses the risk of another negative balance sheet shock.

Indeed, in an interlinked economy, the probability of the bankruptcy of one firm leads to an increase probability of the bankruptcy of its trading partners, but this leads to an increased probability of the bankruptcy of those firms trading partners. The disruption emanates through the economic system.

(With the bankruptcy of one firm depending on firms with which it is interconnected, we have a complex general equilibrium system, which, with bankruptcy costs, has only been solved recently. (Roukny, Battiston, and Stiglitz, 2015).26)

Systemic Bankruptcy

Bankruptcy itself has real societal costs. Much of the information in our society is embedded within corporations. The bankruptcy of an enterprise leads to the loss of organizational and informational capital—a negative shock to the economy, with effects similar to those associated with the negative productivity shocks assumed in real business cycle theory, but which are far more plausible. And of course, the magnitude of these and their consequences depends on the organizational structure of the economy and on economic policies.

Hysteresis effects and persistence

These real consequences of bankruptcy are also part of the explanation of persistence, part of the reason that there are significant hysteresis effects.

When I was chief economist of the World Bank, during the East Asia crisis, we had many disputes with the IMF. As the crisis first broke out in Thailand in July 1997, the IMF wanted to respond with its usual recipe, jacking up interest rates, intended to prevent the collapse of the currency. We were worried that with evidence that the Thai economy already weakening, the increase in the interest rate would exacerbate the economic downturn; and we feared that the hoped for positive effects would not be felt—the increased uncertainty about the weakening economy to which the higher interest rate would contribute could even exacerbate capital outflows, worsening the fall of the Thai Baht. We were afraid too that the higher interest rates would lead to the bankruptcy of a large number of firms. The IMF’s response was that we shouldn’t worry: if our analysis was correct (and evidence that that was the case mounted), they would reconsider their position, reversing course and lowering the interest rate. But I argued that there are important hysteresis effects: the firms that are forced into bankruptcy at the high interest rates do not suddenly become unbankrupt when interest rates are lowered. Though the IMF prevailed at the moment—imposing their usual combination of high interest rates and austerity, our analysis proved totally correct: the policies brought on a deep recession, with a large fraction of the loans becoming non-performing and a large fraction of firms not being able to pay what was owed.

Just as it takes time to restore firm and bank balance sheets, it is even harder to create new firms, to replace the firms that have been lost. There are analogous effects on human capital: those whose education is interrupted by a downturn typically do not resume their education when growth resumes; the loss in human capital associated with the on the job learning that did not occur simply because large numbers were not on-the-job is never recouped.27

SME Lending

These models also help us understand better some of the details of the current and similar downturns. First, investment and employment in small and medium sized enterprises was particularly hard hit, both in the United States and Europe. SMEs are particularly dependent on banking—information imperfections preclude access to capital markets for most SMEs. SME lending in turn is linked to regional banks, not a surprise given the importance of local information.28 Thus, it made a difference to aggregate lending where money was pumped into the system—whether the big banks had their balance sheets restored, or the smaller and regional banks were helped. (In the 2008 bailouts, neither the Administration nor the Fed grasped this. They worried about systemically significant banks. But collectively, the smaller banks are systemically significant. Too little money went to help the community and regional banks, part of the reason that SME lending remained depressed for so long.)

Of course, to understand how policies will affect lending, especially to small and medium sized enterprises, a theory of banking is required.29 The link between what monetary authorities do with the changes in the lending behavior of banks has to be understood. It is remarkable that most central banks used models with no banks: under these models, if they were correct, central banks might not even have come into existence.


But while the standard models didn’t explain lending by banks well, neither did they do a good job explaining what happened to non-bank lending, the vast securitization market that was at the center of the crisis. They simply assumed that markets worked well, that risks had been diversified, and these improvements in risk management were given much of the credit for the “great moderation.” In fact, markets were not working well, and diversification may have made matters worse, rather than better.

There were, in fact, fundamental flaws in the model of securitization (which helps explain why the US government still has a central role in the American mortgage market, underwriting more than 90% of mortgages, more than eight years after the breaking of the housing bubble.) Information is a public good, and, as Sandy Grossman and I long ago pointed out (Grossman-Stiglitz, 1976, 1980), if financial markets were informationally efficient—in the sense that information was efficiently conveyed through prices, there would be no incentive to gather information. Markets must necessarily be informationally inefficient. Some thought one could “solve” this problem with credit rating agencies. But the CRA were paid by those that they were rating: there was a race to the bottom. Their performance was dismal.

There is, in fact, no easy solution—there is no viable private alternative. Banking provides an answer—those who make a loan bear the consequences of any mistakes.30

The structure of credit networks and optimal diversification

Before the crisis, the standard models paid no attention to what happened within the financial sector. These were details of no moment, or so they argued. One could describe the financial sector by a representative firm, just as one could describe the household sector by a representative household. But, of course, what happens inside the financial sector matters a great deal. Financial interlinkages—financial networks, and their structure—matter. A collapse of one firm (Lehman Brothers) might have led to the collapse of the entire financial sector, had the Federal government not intervened.

There were two critical analytic mistakes. The first is related to the macro-economic externalities discussed earlier: if there are externalities exerted by one financial agent on others which they don’t take into account, then the behavior of the financial sector cannot be described well by a model with a single (rational) financial firm.

The second is related to the hypothesis that diversification unambiguously improves matters. That is the case in models where all relevant functions are convex. But non-convexities are pervasive: there are, for instance, bankruptcy costs. Moreover, a firm which faces a negative shock may find itself paying higher interest rates—leading to a process of trend reinforcement. In such a situation, diversification may make matters worse. (No one would suggest that the response to a contagious disease is diversification—spreading those who have been exposed to the disease around the world; but that, in effect, was what diversification implied.)

Ironically, the advocates of globalization and financial integration argued for the virtues of diversification before crises; but afterwards, they shifted their attention towards contagion—emphasizing in effect how interlinkages could lead to the spread of a problem in one country to others. Intellectual consistency should have forced them to take into account these risks ex ante, before the crisis. In fact, one can show that there is an optimal degree of interlinkage (diversification); and that after a crisis, it may be desirable to sever linkages—imposing capital controls.31

Financial sector imperfections and the creation of volatility

Financial market imperfections are important not only because they help explain amplification and persistence. They are also an important part of the explanation of the source of the disturbance: the breaking of the housing bubble and the consumer spending bubble to which it had given rise.

Kindleberger (1978) and Minsky (1992) both stressed the role of bubbles as a source of economic fluctuations. While there is some research suggesting that bubbles can exist even in the presence of rational expectations32, a closer look at this bubble, as well as earlier bubbles, suggests that irrational expectations as well as massive fraud33 played a central role. It seems a misuse of intellectual resources to dwell too long on the counterfactual: could there have been a crisis if all individuals had acted rationally and had rational expectations.

But as modern behavioral economics has emphasized, even if individuals are not fully rational, their behavior can be predictable.34 Thus, Minsky emphasized that as markets became better at managing risk, individuals and firms might undertake more risk, so that the stability of the system might not improve much.35 If they became overconfident in their ability to manage risk (as seemed to be the case with some of our policy and financial leaders), then the system might become even more unstable.

Thus, we can and should study not just how markets might behave if everyone had rational expectations and common knowledge, but how they actually do behave—and how policies (regulations) can affect that behavior. It might not have been rational for investors, seeing returns on safe assets fall with loose monetary policies, to “grab for yield.” (Standard theory would have predicted that the lower returns would have led to more, not less, risk averse behavior.) But if investors systematically do so, then we need to design countervailing policies.

By the same token, it might not have made sense for markets to shift so much risk associated with variations in interest rates to homeowners36, but given these failings in mortgage markets, it makes sense for government to impose regulations to prevent the adverse macro-economic consequences.

Whether rational or not, there are large macro-economic consequences to these behaviors, and thus, there is a need for government regulation.

Going forward, I believe an important part of the research agenda should be trying to understand better the way our economic system creates volatility. Understanding the creation and breaking of credit and asset bubbles is part of the story—a part in which we already have a good sense of how deregulation and loose monetary policy can facilitate the creation of such bubbles. But there is more: Earlier, I described ongoing research on fluctuations in “aggregate” expectations of future wealth—the process of pseudo-wealth creation and destruction, which can give rise to fluctuations in aggregate demand. Creating new opportunities for gambling (trading on the basis of differences in probability judgments)—while defended on the grounds that doing so was “completing markets,” making the economy closer to the Arrow-Debreu ideal—actually can increase volatility. (Guzman and Stiglitz, 2014, 2015a, b.)

Before turning to the contrasting policy positions of these various strands of macro-economics, I want to address three other critical differences between this line of work, and the two described earlier.

Disequilibrating dynamics

There is another reason suggested by these theories for the persistence of downturns: Short run adjustments may be disequilibrating. Lowering real wages—which typically is the decentralized market’s response to unemployment and is normally thought to be required if the gap between the demand and supply of labor is to be reduced—lowers real aggregate demand, exacerbating problems of unemployment.

But there are other reasons that even a lowering of nominal wages and prices may exacerbate the economic downturn, linked to the financial market failures already noted: Lowering nominal wages and prices increases leverage of households and firms, lowering aggregate demand.

The increased leverage can, in turn, increase bankruptcy probabilities, leading to the destruction of information and organizational capital, and increasing uncertainty, with both supply and demand side effects. This, in turn, leads to weaker banks, decreasing lending and increasing interest rates charged by banks (increasing the spread between the T bill rate and the lending rate.)

Real” rigidities matter

In real business cycle theory, there are no rigidities, so that markets always clear. By contrast, we have argued that especially in the context of deep downturns, it is obvious that neither the labor nor capital markets “clear”: there is both unemployment and credit rationing (liquidity constraints bind). In the New Keynesian analysis, attention is centered around nominal rigidities. In our earlier discussion, we explained why some of the traditional explanations of these nominal rigidities are not convincing. But more importantly, it is real rigidities that prevent markets from clearing. Even if there were nominal wage rigidities, if prices were flexible, real wages could adjust to clear markets.

Imperfections of information can give rise to real rigidities, both in labor and capital markets.37

Real rigidities and decentralized adjustment processes

Slow processes of adjustment38 too can lead effectively to real rigidities. In a decentralized economy, wages fall in response to unemployment (due to lack of clearing in the labor market), and then prices fall in response to the gap between aggregate demand and supply (due to lack of clearing in the product market). Then net effect is that real wages change little. (Solow and Stiglitz, 1968.) 39

Labor market frictions

There is one more important source of real rigidity in the economy—one about which the standard excessively aggregated macro-models can have little to say: individuals can be “trapped” in one sector, unable to make a transition to a higher wage job elsewhere, either because they cannot afford to obtain the requisite skill or because they cannot afford the moving costs. Capital markets are imperfect, and those needing to move are often in a poor position either to afford these expenditures on their own or to borrow. In some cases, the sectors out of which individuals need to transition are, in a sense, successful: productivity has increased so much that the demand for labor has decreased. That was the case in agriculture in the first part of the twentieth century, and is the case in manufacturing today. Markets on their own do not manage such structural transformations well.40

Is the current recession more than a balance sheet recession?

In some ways, we are in a similar situation today. The enormous increase in productivity in manufacturing—beyond the increase in demand—means that global employment will be going down; and changing comparative advantage means that an increasing fraction of that employment will be in developing countries and emerging markets. In the United States and Europe, there needs to be a structural transformation from manufacturing to the service sector, and again, markets do not manage such transformations smoothly on their own. Making matters worse is the fact that two of the service sectors into which resources will be moving, education and health, are sectors where government, understandably, plays a big role. But with cutbacks in government spending, the public sector will not be able to play the role that it should in moving the economy from a manufacturing to a service sector economy.

That is why I believe that the current downturn is more than a balance sheet recession. Even when balance sheets are restored say to pre-crisis levels, the economy will not be back to health: the economy was not in fact health in 2007 and 2008 before the crisis. What sustained the economy then was an unsustainable bubble. In the absence of the bubble, aggregate demand would have been insufficient to maintain the economy at full employment.

Rational Expectations

We noted the central role that rational individuals with rational expectations played in the first two strands of economic research. I believe that rational expectations provides a poor guide to understanding macro-behavior.


There are some contexts in which rational expectations might make sense: a static society, in which the only source of variability is weather. In such a world, individuals might eventually learn the full consequences of each possible observable state of nature (though given the number of such states, it is plausible that they would in fact use simpler heuristics), and base their behavior rationally on such expectations.

But today, even in this simplistic context, rational expectations would make no sense: climate change means that we are experiencing weather patterns not seen for perhaps millions of years, and not only are memories limited, but man was not even around the last time weather was similar.

The world is always changing, so that it is not even clear what is entailed by rational expectations. The structure of the economy today is markedly different from what is was not that long ago, and economic policies that are being tried today are different from those that were even tried in the past. How could a rational expectations model make sense in such a situation?

In particular, there hasn’t been a downturn as deep as this one for 80 years. The world 80 years ago was markedly different—with different politics, different beliefs about the economy, a different economic and financial structure.

In rational expectations models, everyone has same beliefs. But as we have argued, divergences in beliefs are of first order importance for understanding markets and macroeconomic behavior. This helps explain the large diversity of interpretations of events and policies. Even now, there are disagreements about magnitudes of multipliers. Some of these disagreements are because some economists are working with the wrong model. It should be obvious that any model that assumes that labor and capital markets clear is not going to provide insights into a deep downturn. Inferences based on models estimated in “normal” times are of little relevance in deep downturns. In normal times, the economy is close to full employment, so that an expansion of government spending crowds out private spending. But this is not so when there are large amounts of underutilized resources, as in a deep downturn.

In RE models, there is no learning, no problem of assessing whether we are experiencing an extreme outcome in an old regime, or whether we have moved into a new regime. But such learning is, I believe, central to behavior of economic agents. (Guzman, 2013.)

Many critical aspects of what went on in the economy in the run up to crisis cannot be reconciled with rational expectations behavior on the part of large fraction of economic actors—although there were often a few who made some money by exploiting seeming irrationality on the part of others. But these did not suffice to prevent the creation of a major bubble.41 (Thus, the argument that the economy acts as if everyone were rational—because those that are can “arbitrage out” any inefficiencies created by those who are not, was patently wrong.)

This is more than just a statement that the crisis was not “expected”.42 As I noted earlier, the design of mortgages did not represent a “rational” and efficient system of risk sharing.43 Moreover, the dominant market ethos seemed predicated on the belief that housing prices could increase, seemingly forever, unabated. But it should have been obvious that it was virtually inconceivable that housing prices/real estate prices could continue to grow: there were limits on spending on housing (especially given the stagnation in incomes of so many Americans) and there is a virtually unlimited supply of land in Nevada desert, limiting the extent to which land prices there could have increased. It was thus remarkable that so many of our economic leaders, including the chair of the Federal Reserve, seemed unconcerned that there might be a bubble.

D. Contrasting Elements

While the three strands of macro-economics that we have described share much in common, I hope that the above discussion has highlighted some of the important distinctions. There are several in particular to which I wish to draw attention, and which will play an important role in the policy discussion to follow.

Deep downturns are a manifestation of deep and pervasive market failures—deviations from the standard competitive equilibrium model which underlies real business cycle theory. Markets fail in many ways, and the strand of New Keynesian model emphasizing nominal wage and price rigidity probably has identified one such market failure, but perhaps not the most important one. I have suggested that the alternative strand of New Keynesian models emphasizing financial sector imperfections may be more relevant, at least for deep downturns. Indeed, the third strand emphasizes that the central problem confronting economies facing deep slumps may not be price rigidities, but price flexibilities. The current worries about deflation are justified. Economies in which wages and prices are less flexible perform better.

RBC models emphasize technology shocks. In the New Keynesian models of either variety, monetary policy shocks can also have important consequences.44 But the first two strands, with their emphasis on rational expectations, rule out the most important source of fluctuations: Deep downturns are caused by shocks to the economy created by the economy itself. The behavior giving rise to the shocks typically is associated with irrational behavior, and even when such irrationality is exploited by rational actors, they cannot fully undo the effects.

Finally, New Keynesian models with some price rigidities but with sufficient wage flexibility such that there is full employment, even if they might be able to explain movements in output and employment, have little to say about one the central variables of concern in deep downturns, unemployment.45

One consequence of the pervasive market failures is that there are large macroeconomic externalities—the application at the aggregate level of the pervasive market externalities arising in economies with imperfect information and imperfect risk markets discovered by Greenwald and Stiglitz (1986). Macroeconomic externalities help explain both amplification and persistence. While these effects and their implications (e.g. for multipliers) was originally studied in the context of New Keynesian models with financial frictions, similar results obtain in models with price rigidities. (Farrhi and Werning (2015).)

Deep downturns are often related to problems in the financial sector. (One has to be careful, however, to avoid the superficial reading of data which suggests that financial crises are different—they last long and are deeper. For an example, see Rogoff and Reinhart, 2009. A deep downturn, whatever its origins46, will eventually affect the financial sector as borrowers will be unable to pay; and when the downturn is deep enough, there will be a financial crisis in the absence of government intervention. In such a situation, the financial sector may be an important part of the process of amplification and persistence; but the statement that downturns involving a financial crisis are deeper and last longer may be nothing more than a tautology, saying nothing more than that deep and prolonged downturns are deep and prolonged.

The financial sector in turn cannot be understood within a representative agent model; information asymmetries and incomplete markets are central. Macro-economic externalities arise especially in the financial sector and highlight why one needs to study carefully the structure within the financial sector (including the interlinkages among financial institutions). The financial sector cannot be adequately summarized in a money demand equation, as the first two strands of macro-economic analysis attempted to do.

E. Contrasting Policy Implications

The alternative theories described in the previous subsections differ not only in their interpretation of downturns, but also in the advice they give concerning appropriate policy responses. In this section, we look at three examples.

Monetary policy

Traditionally, monetary policy has been the instrument of choice in responding to economic fluctuations. But it is clear, in deep downturns, that it may be ineffective, or at least, insufficient. Keynes explained this in term of the liquidity trap, the inability to lower interest rates enough to stimulate investment (or consumption) enough to restore the economy to full employment. This theme has been picked up in some of the recent literature under the concept of the zero lower bound—nominal interest rates could not be lowered below zero. (Keynes also explained the ineffectiveness of monetary policy in such a situation by using an analogy: it was like pushing on a string.)

We have already suggested why matters today are different than they were in the Great Depression, when real interest rates were 10%. Obviously, if we lowered real interest rates enough, we could presumably restore the economy to full employment—an interest rate of minus 100% would change matters. (This is equivalent to just giving money away; later in this lecture, I discuss “money rain.”) But that is not what those who are focusing on the ZLB mean. They argue that if only we could lower real interest rates, say to -4% (from the minus 2 percent that it was in the aftermath of the onset of the Great Recession) we would have recovered. I have never seen any convincing evidence in support of that contention, and as I noted earlier, if the reason for the continued poor performance of the economy were the zero lower bound (ZLB—the fact that nominal interest rates could not be lowered below zero), intertemporal prices could be changed through tax policies.

The central flaws in the analysis of real business cycles and the New Keynesian models with wage and price rigidities are that (a) they see monetary policy working through interest rates; (b) in these models, there is only one relevant interest rate, the T-bill rate; (c) the institutional arrangements through which credit is made available (banks) makes no difference; and (d) distributional effects are assumed away.

We argued before, though, that there can be credit rationing; and liquidity constraints are especially relevant in deep downturns. If there is credit rationing, then monetary policy affects aggregate behavior through credit availability; and credit availability is affected not just by conventional monetary instruments (open market operations), but also by the whole gamut of micro-and macro-prudential regulations (such as capital requirements).

But even if there is no credit rationing, what matters for firm behavior is the lending rate, not the T-bill rate (few firms can borrow at the T-bill rate—evidence itself of the importance of bankruptcy and insolvency). But the spread between the two is endogenous; one of the main objectives of Greenwald and Stiglitz (2003) was to explain that spread.47

The literature in the third strand of New Keynesian economics described above explains the ineffectiveness of monetary policy in a way quite different from that focusing on the zero lower bound: Banks are unable or unwilling to lend, both because of shocks to their net worth (as a result of higher than expected defaults) and increases in risk perceptions. In such situations, lowering the T-bill rate (or even a capital injection through preferred shares) has little effect: it may neither lead to a substantial increase in credit availability nor lower interest rates. Banks may, moreover, not pass on lower interest rates to customers (and this may be especially so if markets are not highly competitive).

Moreover, monetary policy can have large distributive effects, and these can undermine its effectiveness. Quantitative easing may have led to a stock market bubble, which may have led to (a slightly) increased consumption by the very rich, though given that it was announced that the intervention was only temporary, it was hard to understand why there would be significant consumption impacts in the absence of significant impacts on intertemporal prices—and these effects were minimal. But the lowering of the income of retired people dependent on the return to government bonds would have been expected to have a significant adverse effect on the consumption of these individuals.

There may be even more adverse effects on the medium term: the lower cost of capital may induce firms to use more capital intensive technology, leading to a jobless recovery, exerting downward pressure on wages.48

Fiscal policy49

Real business cycle theory argued that fiscal policy was not only unnecessary but ineffective: with the economy always at full employment, more government spending simply displaced private spending. Tax cuts financed by debt simply led individuals to save more.

By contrast, in the more general New Keynesian theories, fiscal policy can be very effective; multipliers can be large—and indeed, they can be even larger with rational expectations (or more generally, with future-oriented individuals). Public investment can, in fact, lead to crowding in of investment, if there is complementarity between public and private investment.50 There can be crowding in of consumption, if there are expectations of future higher incomes. (Neary and Stiglitz, 1983.)

A large balanced budget multiplier means that expansionary fiscal policy can work even with budget constraints.

Still, in the NK models with price rigidities, there was a strong preference for the use of monetary policy, except in the limiting case of the ZLB, and it is important to know why. It is largely is because the fashionable versions of these models are designed to exhibit Ricardian equivalence.

Ricardian equivalence

For instance, a tax cut financed by government borrowing induces individuals to save more, to repay the liability which they will have to pay in the future. Consumption is unchanged. The conditions under which Ricardian equivalence holds are, of course, very restrictive, but the NK models with price rigidities typically embrace all of the special conditions required. While Stiglitz (1988) showed that Ricardian equivalence holds under much more general conditions than those assumed by Barro—and for a much wider range of public financial operations—it did not hold when there was credit rationing or other forms of financial restraints. Nor did it hold in general in models with finite lived individuals.

But even within the infinite life-time dynastic models with perfect markets, Ricardian equivalence doesn’t hold, as we noted above, if public spending goes for public consumption that are not perfect substitutes for private investment goods or for public investment goods that are not a perfect substitute for private investment goods, and this is true whether there is market clearing or not.

There is a second reason that the rigid price NK models with full employment focus on monetary policy—seen as changing intertemporal choices. Consider the limiting case where there is no labor supply elasticity. Then in such models, incomes are fixed. That means that if we are to achieve full employment today, we simply have to shift enough of lifetime demand into the current period; with some intertemporal substitution we can always do that. If we can just adjust real intertemporal prices correctly, we can easily achieve full employment every period. The price rigidities mean that welfare may not be as high as it would be in the first best world of RBC; but at least, we have “solved” the macro-economic problem. (With variable employment, matters are slightly more complicated, because there can be income effects from the changes in intertemporal prices, as individuals decide to supply more or less labor.)

Reduced payroll taxes

The consequences of a temporary reduction in the payroll tax illustrate the sensitivity of results to particular assumptions. Assume that there is a temporary shift in demand away from the present, that leaves aggregate demand today weak. In models with wage flexibility (even NK models with rigid prices) wages will fall until markets clear, and employment decreases. In models with rigid (real) wages, there will be unemployment. Assume the government wishes to smooth out the fluctuation by reducing the payroll tax. Standard theory says it doesn’t make any difference to the competitive equilibrium on whom the tax is levied, whether the worker or the firm; but in the discussion below, we focus on how in the short run, adjustments differ and that matters for the short run equilibrium. Consider an institutional arrangement where the payroll tax is paid by workers, and the reduced tax increases the paycheck received by the worker. With Ricardian equivalence, the worker saves the entire amount. In reality, he doesn’t—partly because he may be financially constrained, partly because he assumes the debts will be paid by someone else’s great grandchildren. If workers spend more, aggregate demand increases and unemployment falls. In NK models with full employment, it is the supply side effect that dominates: at the higher real wage, workers want to work more. There is a shift in the demand and supply curve for labor. If the payroll tax cut is temporary, little of the extra income will be spent this period, so the demand effect is limited, and the supply effect may be larger—this period of high after tax wages is a good time to work. To accommodate the greater labor supply, the wage paid by the firm must fall. There is thus some redistribution effect; but in representative agent models, where the worker also owns the firm, these redistributive effects can be ignored.

There is a problem: if there is a large labor supply elasticity and little extra demand, there will be an excess supply of labor. Normally, monetary policy can accommodate this, shifting demand forward. But with the ZLB, this is not possible. Hence, the main effect of the payroll tax cut is to lower before tax wages, leaving after tax wages little changed. The payroll tax cut is not very effective.

This assumes, of course, that prices don’t adjust. But if there is some price flexibility, the lower wage that results induces a slightly lower price (those firms that can lower prices do so), and this induces even more first period consumption. (In principle, we need to consider further reactions on labor supply and consumption in later periods, and how these reverberate back to the first. We assume these second round assumptions are of second order importance.)

The alternative strand of NK models argues that this analysis underestimates the effects on employment by (a) underestimating the current effect on aggregate demand by assuming Ricardian equivalence and ignoring the role of finance constraints for most individuals; and (b) overestimates the labor supply side effect—.51 Output increases, not because workers are willing to work more, but simply because there is more demand. These effects are especially marked in a world with wage and price rigidities, so that in the initial situation, there is unemployment. The labor supply effect is then largely irrelevant.

In the case where there is some price flexibility, the alternative strand of NK models would suggest too that the intertemporal substitution effect—shifting consumption forward in time—is overestimated. At the same time, so are the implications of the ZLB.

In the Fisher/Greenwald/Stiglitz models with price flexibility, there is, however, a slightly offsetting effect: a transfer of wealth as a result of the lower prices from debtors to creditors, and this diminishes the magnitude of the multiplier.52

But now assume that the tax is imposed on the employer, and thus he sees the cost of labor as lowered. If prices remain unchanged, as in the NK fixed price models, his profits increase, and the owners of the firm are wealthier; but they see that as being offset by the same amount as their future tax liability. Labor is cheaper, so they increase their demand for labor, and this drives up the wage. There are redistributive consequences, but in a representative agent model, these do not matter. But again, with the after tax wage lower, if there is some flexibility in prices, prices today will fall, and this will shift consumption forward in time, with all the consequences that we previously described.

Again, compared to the other NK models, the demand effects may be underestimated, because of the assumption of Ricardian equivalence, or overestimated, because of the absence of redistribution effects. So too, the analysis underestimates the benefits from the strengthening of firms’ balance sheets, while the intertemporal substitution effects are overestimated.

The point of this lengthy exercise is to emphasize that the results of any policy change can depend critically on the assumptions, about what constraints are binding, about labor supply elasticities and intertemporal substitutability, and about the magnitude of redistributive effects.53

The achievement of modern macro-economics in recent years is to construct models within which a wide range of policies can be debated. The question then becomes, what are the most plausible models? Which ones best capture what is going on? In constructing good models of deep downturns, what matters is not so much how well the model “works” in good times, but how well it explains the details of what we observe in these rare but highly important situations.

Open Economies

The analysis of small open economies offers many advantages: prices can be taken as given, countries can borrow and lend, etc. But there are some difficulties: a small adjustment of exchange rates would normally confront the economy with unlimited demands. How, then, can aggregate demand be a problem?

Greenwald and I in our joint work took one approach: we focus just on domestic supply. With imperfections in financial markets, the amount that firms are willing to produce, the amounts that they are willing to spend on investment, and the amount of labor they are willing to hire will be limited. We focused on how these variables were interrelated and would be affected by various shocks and policies.

Guzman and I in our joint work have taken another approach: there is a large non-tradeable sector, and resources cannot move perfectly between the tradeable and non-tradeable sector. Their models help us understand better the failure of programs of austerity in Greece and many of the other countries in the euro-crisis. It was hoped that austerity would lead to wage and price decreases; lower costs, it was thought, were necessary to increase competitiveness, in the presence of exchange rate rigidities.

But there are adverse effects on non-traded goods’ demand, especially since some of the distributive benefits of lower wages are garnered by the foreign owners of the firms within the country54; and their gains do not lead to increases in the demand for non-traded goods. These adverse effects on the non-traded goods sector can more than offset any benefits from international competitiveness. Even worse, the increased bankruptcy among producers of non-traded goods has spillover effects on producers of traded goods (e.g. through the banking system), and these effects can be so large that there may be no increase in exports. This happened during the East Asia crisis, and, more recently, in Greece.