Successful marketing typically makes a product new and old at the same

Transcription

1 Journal of Economic Perspectives Volume 7, Number 1 Winter 1993 Pages Will the New Keynesian Macroeconomics Resurrect the IS-LM Model? Robert G. King Successful marketing typically makes a product new and old at the same time. For example, many new automobile models are presented to consumers every year. With each new model, the manufacturer seeks to hold on to the existing market and to attract new buyers. A typical advertising campaign thus stresses that traditional features have been kept, with exciting new ones added. But the stress varies across market groups: traditional buyers get a heavy emphasis on maintained features, while the prospective new purchasers get a picture of a virtually new product. So too it is with economics. Much recent attention is attached to the development of a "New Keynesian Macroeconomics." 1 In the twin volumes of this title edited by N. Gregory Mankiw and David Romer (1991), the product is a seductive one. The reader is presented with a wide array of short papers developing particular aspects of aggregate supply; for example, on the pricing of final products and on the operation of labor markets. Microeconomics is emphasized heavily, in contrast to traditional 1960s Keynesian work, although supporting empirical evidence receives much less stress than it did in the earlier vintage. Taken together, these papers are viewed as refuting "the new Classical argument that the Keynesian assumption of nominal rigidities was 1 In addition to the papers in this symposium, there are a number of useful recent surveys. A prominent new Classical economist, Robert Barro (1989), is skeptical that the new Keynesian macroeconomics has much new to contribute about wage or price stickiness. A traditional Keynesian, Robert Gordon (1990), argues that the new Keynesian macroeconomics provides a useful general framework for the supply side, but needs to make more contact with specific empirical puzzles. Other noteworthy descriptions of the new Keynesian program are contained in the popular writings of Mankiw (for example, 1990). Robert G. King is Professor of Economics, University of Rochester, Rochester, New York.

2 68 Journal of Economic Perspectives incapable of being given theoretical foundations" (Mankiw and Romer, 1991, p. 15). Along with this advertising of new developments, a great effort is made to reassure the old-time consumer of the Keynesian product. Popular articles and research conferences often feature a great deal of discussion about how the new Keynesian macroeconomics has rationalized most of the old practices of Keynesian macroeconomics. The veteran Keynesian is told that the IS-LM model remains the best way to understand the determinants of aggregate demand and to conduct contemporary policy analysis. 2 It is argued that a return to the research agenda of the 1960s is appropriate; for example, Olivier Blanchard (1991) argues that research since 1975 represents a "long pause" in the accumulation of macroeconomic knowledge. Overall, the picture is that Keynesian macroeconomics stumbled on the stagflationary episodes of the 1970s and that a major professional overreaction occurred. But economists should not be led astray by this advertising campaign, sophisticated and recurrent though it is. The IS-LM model has no greater prospect of being a viable analytical vehicle for macroeconomics in the 1990s than the Ford Pinto has of being a sporty, reliable car for the 1990s. Because of its treatment of expectations, the IS-LM model, as traditionally constructed and currently used, is a hazardous base on which to build positive theories of business fluctuations and to undertake policy analysis. Of course, every macroeconomic model contains some set of equations that can be labeled as its IS and LM components, since these are just conditions of equilibrium in the goods markets and the monetary sector markets. But the operational question is whether these are simple functions of a sufficiently small set of variables to be a useful "back of the envelope" tool. The lesson of the past decade of research is that the answer is no. While some of us may choose to use the IS-LM framework to express results that have been discovered in richer models, it is not a vehicle for deriving those results. To simplify economic reality sufficiently to use the IS-LM model as an analytical tool, economists must essentially ignore expectations: we now know that this simplification eliminates key determinants of aggregate demand. In particular, the last two decades of research have taught economists that the assumption of rational expectations is a powerful part of economic explanations of individual and market behavior, ranging from consumption and investment dynamics to pricing of stocks and bonds. The emphasis on expectations in the macro-model is the end result of a process of building microeconomic underpinnings that was initiated in the 1950s and 1960s, when the goal was to develop dynamic theoretical foundations for the IS and LM schedules: 2 For example, N. Gregory Mankiw (1990) notes that "the textbook IS-LM model, augmented by the Phillips curve, continues to provide the best way to interpret discussions of economic policy in the press and among policy makers." He rationalizes the continued use of this model by arguing that "the observation that recent developments have had little impact on applied macroeconomics creates at least the presumption that these developments are of little use to applied macroeconomists."

3 Robert G. King 69 Figure 1 IS-LM and Expectations inevitably, consideration of dynamic choice pushed the question of expectations to the forefront. As a result, most of the equations of the IS-LM model are now viewed as summarizing purposeful economic behavior in which choices over time play a central role. However, as we will see, this finding means there is no way to maintain traditional uses of the IS-LM model. Most readers of this article are familiar with the IS-LM model as a classroom device, so it is useful to start there. Figure 1 draws the familiar IS and LM schedules as functions of the nominal interest rate (R) and real income (y). The IS specification is typically portrayed as downward sloping, which results from assuming that saving is a positive function of income and investment is a negative function of the interest rate (as the cost of capital). The LM schedule is typically portrayed as upward sloping since money demand depends positively on real income (as a measure of the scale of transactions) and negatively on the nominal interest rate (as an opportunity cost of cash balances). From the perspective of a teacher discussing the consequences of an increase in the money stock for aggregate demand, the well-known result is that money raises demand and lowers both real and nominal interest rates: this is the standard result illustrated in the first panel of Figure 1. But the influence of expectations on investment can easily make the IS curve shift enough so that real and nominal rates both rise. Real rates may rise because investment is sensitive to expected future demand and changes in money may signal sustained increases in demand. Nominal rates can rise because increases in money signal higher future price levels generating expected inflation and lowering the cost of investing at any given nominal interest rate. And it turns out that these channels of influence are very important both in theoretical macroeconomic models and in actual U.S. history. Whether for the academic researcher or the policy-maker, rational expectations introduces a powerful new set of determinants of aggregate demand that shift through time in response to policy actions and with other information

4 70 Journal of Economic Perspectives about the future: the IS-LM model provides little guidance about how to think about these events. To emphasize the importance I attach to the influence of expectations on aggregate demand, the current essay does not concern the role of expectations on the aggregate supply side. Despite their importance, my sense is that economic theories of aggregate supply of either a new Keynesian or new Classical sort are simply too underdeveloped for economists to evaluate the relative importance of expectations and other mechanisms in explaining aggregate supply behavior. But on the demand side, a range of empirical and modeling perspectives has provided more lasting contributions. In drawing these connections, my essay draws upon a wide range of research: some assuming continuously clearing labor and product markets, some assuming nominal price or wage rigidities, and other work agnostic on aggregate supply. In this essay, the focus is on aggregate demand, as in fact was standard in the old Keynesian tradition. By contrast, the other contributions to this symposium focus on aggregate supply. In reviewing these essays, it strikes me that all of the authors are asking the wrong question. Both the new and old Keynesians are trying to answer a question of the 1950s and 1960s: why are prices sticky? But the question that needs to be asked is: why are prices sticky in certain historical episodes and rapidly adjusting in others? In answering this question and in unraveling the related issue of the consequences of nominal disturbances for real economic activity, static models that ignore expectations will be as useless for aggregate supply theory as for aggregate demand theory. 3 The Keynesian Revolution and the IS-LM Model The central questions of macroeconomics today are essentially the same as when John Maynard Keynes published The General Theory of Employment, Interest and Money in 1936: Why do economies grow and fluctuate at different rates? Which public policies yield desirable patterns of growth and fluctuations? To provide answers to these questions, Keynes made three parallel and bold assumptions in the General Theory. First, he represented the private sector by simple behavioral rules. For example, on the real side, consumption was linked to income via the "propensity to consume." On the monetary side, demand for cash balances was determined via "liquidity preference." Nominal prices were treated as largely predetermined when analyzing how real activity evolved during specific historical episodes, at most responding only gradually and mechanically through time. Second, Keynes viewed key economic activities as dependent on expectations, as in his idea that investment demand was 3 There are models of price setting behavior that stress dynamics and expectations in the Mankiw and Romer volume, including basic work by Stanley Fischer (1977) and John Taylor (1980). More promising recent contributions along these lines are by Guillermo Calvo (1983) and Julio Rotemberg (1982). But by and large, as is clear from the symposium contributions, many new Keynesian authors believe that these issues are very secondary.

5 Will the New Keynesian Macroeconomics Resurrect the IS-LM Model? 71 determined by the future yield to capital assets. But, critically, he saw these beliefs as largely exogenous and invariant to the policy actions he proposed. At the same time, he stressed that they would also be a source of shocks with a powerful influence on economic activity particularly as they affected investment. Third, Keynes believed that desirable public policy could be found by asking how the monetary and fiscal authorities should respond on a case-by-case basis, treating the simple behavioral rules and expectations as fixed while the authorities responded to particular economic disturbances. Of course, a large literature has sought to explain the nature of the contributions that Keynes made in the General Theory. Without denigrating that work, it is important to look at Keynes' (1936, pp ) own summary of his theoretical structure: Thus we can sometimes regard our ultimate independent variables as consisting of (1) the three fundamental psychological factors, namely the psychological propensity to consume, the psychological attitude to liquidity, and the psychological expectation of the future yield to capital assets, (2) the wage unit as determined by the bargains reached between employers and employed, and (3) the quantity of money as determined by the action of the central bank; so that, if we take as given the factors specified above, these variables determined the national income... and the quantity of employment. Working from this base, John Hicks (1937) and Franco Modigliani (1944) built essentially static models that operationalized the Keynesian vision in now familiar IS-LM schedules, with appendages for product and labor markets. In summarizing his approach in the sentences above, Keynes clearly adopted the short-run focus that A. C. Pigou (1951) identifies as the fundamental conception in the General Theory. There was a good reason for this focus. In modern terms, Keynes' view was that business fluctuations resulted mostly from shocks to aggregate demand; he suggested that expectationsinduced shifts in investment demand were the principal business cycle impulse. 4 His three assumptions led naturally to a largely static, short-run model that could be used to think through the consequences of these shocks. Keynes saw large shocks to private behavior buffeting the economic system; he used his theory to argue for the prompt, frequent policy responses to counteract to this instability. Clearly, this modeling procedure was a short-cut. To produce an operational policy, Keynes was willing to put aside issues about how private sector behavior arose from the choices of households and firms. Sometimes, as in his 4 The strength of his convictions on the variability of investment and the conclusions to which it led him can seem surprising to modern readers in the light of subsequent decades of interpretation. For example, Keynes (1936, pp. 378) wrote: [I] seems unlikely that the influence of banking policy will be sufficient by itself to determine an optimum rate of investment. I conceive, therefore, that a somewhat comprehensive socialization of investment will provide the only means of securing an approximation to full employment."

6 72 Journal of Economic Perspectives stress on the interest sensitivity of money demand, he identified margins of substitution that had been largely neglected in prior work in monetary economics. But more generally, he put aside a range of thorny questions. Most important, he abstracted from asking how factors relevant to long-run determination of prices and quantities influenced economic activity over the course of the historical episodes he sought to explain. 5 For Keynes, the long run was very far away; the main purpose of the General Theory was to generate an operational model of short-run macroeconomic activity with which to think through policy actions. Of course, the General Theory offers many other interesting conjectures and ideas; it is a lengthy volume written by a wide-ranging intellect. But the central element of Keynes' work is a perspective on the determinants of economic activity, captured by Hicks and Modigliani in the IS-LM model and then later built into macroeconometric models by Lawrence Klein, Modigliani, and their co-workers. As codified by Hicks and Modigliani, the Keynesian framework has certain key elements. There was a neat separation of the saving-investment sector from the monetary sector. On the IS side of the model, mechanical behavioral rules governed consumption and investment. These rules specified a small set of determinants of each (output, disposable income, and interest rates), as well as the level of government demand. On the LM side of the model, the money demand function (depending on interest rates and income) was equated to the money stock. This made policy analysis neat and operational: shifts in government purchases and taxes affected only the IS curve, monetary policy actions influenced only the LM curve. The Refinement of the IS Specification During the 1950s and 1960s, microeconomic models of consumption and investment were developed to aid in understanding the determinants of these variables. On the consumption side, the permanent income theory of Milton Friedman (1957) and the life-cycle theory of Modigliani and Richard Brumberg (1954) were each developed within frameworks that emphasized the intertemporal nature of a household's consumption decision. On the investment side, the neoclassical theory of Dale Jorgenson (1963) emphasized the importance of expected future production levels as well as the expected costs of capital. 5 His skepticism about the relevance of the long run was deeply held and long standing. Perhaps its most famous statement appears in The Tract on Monetary Reform (1923), where Keynes belittled the relevance of the quantity theory of money as follows: "Now in the long run this is probably true. If, after the American Civil War, the American dollar had been stabilized and defined by law at 10 percent below its present value, it would be safe to assume that n (the money stock) and p (the price level) would now be just 10 percent greater than they actually are... But this long run is a misleading guide to current affairs. In the long run we are all dead. Economists set themselves too easy, too useless a task if in tempestuous seasons they can only tell us that when the storm is passed the ocean is flat again."

7 Robert G. King 73 These investigations were not just "theory for theory's sake." The permanent income theory of consumption was explicitly designed to reconcile shortrun and long-run evidence on the behavior of aggregate consumption. The theory also had implications for individual-level consumption data that Friedman used to buttress his case for the permanent income hypothesis. The neoclassical investment theory was designed to aid in more accurate measurement of the cost of capital its implicit rental price as well as providing a restriction on how this cost variable should enter into the investment demand function. In particular, the neoclassical model of investment developed as part of a wide-ranging investigation into empirical investment demand functions: starting from the work of Jan Tinbergen (1939) and Klein (1950), econometric model-builders had found that their investment equations displayed large residuals that were linked to cyclical movements in income and product. On one level, this result was consistent with Keynes' suggested business cycle impulse; however, it also left modelers with the uncomfortable sense that critical determinants of investment had been left out. A common element in these consumption and investment theories were expectations about the future: future income levels, future production levels, and future costs. But as long as expectations were treated as exogenous, as in Keynes' work, then little change was necessary in the operation of the IS-LM framework. The new theories provided a useful base for researchers. For those involved in empirical work, they provided an econometric specification, including a list of variables to be used in regression equations and detailed hypotheses about magnitudes of effects to be tested. For others, they provided a way of deciding on the importance of various influences in the Keynesian IS-LM framework. For example, Friedman's permanent income theory was widely viewed as implying a low marginal propensity to consume and, hence, a low value of the multiplier. As the neoclassical investment theory developed, it became clear that the investment process contained important dynamic features beyond the basic dimension stressed by Jorgenson. Versions of the neoclassical investment theory that incorporated additional features like the costs of rapidly changing the capital stock or delivery lags tended to heighten the emphasis on expectations, since optimal intertemporal adjustment in these frameworks involved forecasting many periods worth of output and rental prices. Refinement of the LM Specification The 1950s and 1960s also brought a comparable effort to articulate microeconomic models of money demand. The most fruitful of these, work on the transactions demand for money by William Baumol (1952) and James Tobin (1956), formalized Keynes' view that money demand should be sensitive to interest rates. At the same time, though, these and other microeconomic analyses of money demand sharply limited the variables that should enter in the demand for money. If holding money imposed an opportunity cost, then

8 74 Journal of Economic Perspectives speculative activities based on anticipations of changes in return and risk would not result in variations in transactions money, but rather in relative stocks of other assets (exchanges of Treasury bills for longer-term bonds or stocks, for example). Similarly, according to the theory, changes in expected inflation should influence the holdings of cash balances only via their influence on the nominal interest rate. On the other side of the money market, an interest sensitivity of money supply was typically derived by considering aspects of how banks managed their reserves. In this way, a largely static view of money supply and demand behavior was incorporated into the IS-LM model. However, three exceptions to this perspective did appear. First, Friedman (1959) put forward a view of money demand in which wealth, in addition to income or expenditure, entered into the demand for money. Second, a range of empirical researchers reported that measures of expected inflation entered in the demand for money (see, for example, the summary discussion in David Laidler, 1970, pp ). Third, empirical researchers found it necessary to introduce "partial adjustment" specifications into the demand for money and sought to investigate the importance of expectations to this adjustment process, paralleling developments in investment analysis (Feige, 1967). 6 Rational Expectations and the IS-LM Model The critique of econometric policy analysis by Robert E. Lucas, Jr., (1976) set in motion a program of investigating the role of rational expectations in macroeconomic models. As an economist who joined the ranks of our profession when this program was just beginning, I am acutely aware that it seemed a risky venture. It was not at all clear that the effect of introducing rational expectations into macroeconomics was going to be important, either in the quantitative performance of models or in the empirical explanation of economic fluctuations. But strong evidence now supports both of those linkages. 7 6 It was the fact that the real balance effect emphasized by Pigou and Don Patinkin (1965) was so inconsistent with this separation that led to the profession's interest in it, despite a widespread view that real balance effects were empirically unimportant for most countries and time periods. One additional, important reason was that the Pigou effect played a role in a major theoretical puzzle of the time: whether price level declines would restore neoclassical equilibrium, if there were a liquidity trap. 7 The evidence discussed below is drawn from (i) aggregate time series and panel data studies of consumption and investment; and (ii) the operation of small-scale dynamic models. But rational expectations have also been shown to be quantitatively important in the operation of medium-scale Keynesian macroeconometric models. Fair (1979, p. 551) summarized his incorporation of rational expectations into just the stock and bond markets of his large-scale model as indicating that "unanticipated policy actions are about half as effective (with respect to real output changes) when there are rational expectations...than when there are not." Since there was no important alteration of the role of expectations in labor or product markets, my sense is that this experiment is an additional indication of the importance of expectations for aggregate demand, but in a substantially different framework from those that I concentrate on below.

9 Will the New Keynesian Macroeconomics Resurrect the IS-LM Model? 75 Rational expectations challenges the logic of the IS-LM model in a fundamental way. In his own thinking, we have seen that Keynes drew two sharp distinctions: (i) between a long run that was very far away and a short run that was very important; and (ii) between shifts in the savings-investment area and shifts in the monetary sector. Each of these separations were built into the IS-LM model by Hicks and Modigliani. But in models with rational expectations, these neat separations are destroyed. First, expectations about the future require that the long run and the short run are treated jointly. Second, expectations link together the conditions in the saving and monetary sectors. And rational expectations has proved a powerful tool for understanding economic phenomena. To take just one example, Mankiw, Miron and Weil (1987) show that the pricing of medium-term government bonds changes systematically with the introduction of a U.S. central bank in exactly the way predicted by a rational expectations theory of the term structure. When there are large, systematic changes in policy, then the tool of rational expectations gives a good account of how private responses will be altered. We next explore some specific examples of how incorporating rational expectations affects the IS-LM analysis. But before we do, it is useful to dispose of one potential source of confusion. It is sometimes assumed that discussing the implications of rational expectations for a macroeconomic model is essentially related to investigating whether monetary and fiscal policies have real effects. There are examples, such as the influential model of Thomas Sargent and Neil Wallace (1975), where this is precisely the issue. But, more generally, incorporating rational expectations simply alters the operation of a macroeconomic model: it can act to either raise or lower the potency of macroeconomic policy. Expectations, Consumption and the IS Curve The rational expectations version of the permanent income hypothesis has proven remarkably powerful in three distinct ways. First, in the hands of Robert Hall (1978) and Marjorie Flavin (1981), it has shown a consistency with the aggregate time series, providing a theoretical link between changes in consumption and movements in income. Second, in studies of panel data following Hall and Frederick Mishkin (1982), it has given a good account of the joint behavior of individual income and consumption. Each of these lines of research suggests that some small fraction of "liquidity constrained" consumers (typically one-tenth or less) follow the rule of equating consumption to current income. 8 But the dominant variation in consumption is that predicted by the permanent income theory. 8 John Campbell and N. Gregory Mankiw (1989) present the only estimate that is well outside this range, based on aggregate data, which suggests that about one-half of consumption expenditure is governed by the simple rule of setting consumption equal to disposable income. Christiano (1989) provides a careful critique of this work, which stresses two of its inconsistencies with other evidence on consumption. First, the overwhelming evidence from panel studies is that this fraction is small. Second, if the Campbell and Mankiw estimate is included in a standard business cycle model, it moves the implications for relative volatility of consumption well outside the observed range.

10 76 Journal of Economic Perspectives Third, in small scale dynamic general equilibrium models, the rational expectations permanent income model has demonstrated the ability to explain the relative volatility of a consumption and income. Some might be suspicious of any evidence developed using a standard "real business cycle model," but that would be a mistake. 9 It is well-established that productivity is, in an accounting sense, the predominant source of fluctuations in output. One interpretation of the question addressed by a real business cycle model is thus: given the observed variation in output over the business cycle, what implications does the model have for the variability of the components of output? From this perspective, real business cycle models give a remarkably uniform finding: the permanent income theory of consumption is a central mechanism in generating relative volatility of consumption and output which resembles actual macroeconomic time series. With the permanent income theory governing saving, expectations then play a major role in the IS sector. Under the modern interpretation, it is not correct to claim that the permanent income theory just means that the marginal propensity to consume is low. Rather, the marginal propensity to consume depends in an essential way on the nature of the income profile under consideration: it is low for changes in income that are temporary; high for those that are permanent; and depends in a delicate way on the details of the stochastic process for situations in between. More generally, rational expectations models of consumption imply that all variables useful for forecasting income and interest rates should enter into the consumption function. Thus, a reduced form IS curve would contain virtually no useful exclusion restrictions; it will depend on everything. Expectations, Investment and the IS Curve The program of developing rational expectations models of investment has made important progress in the last decade or so. When these findings are incorporated into dynamic models, their quantitative implications are dramatic, often overwhelming the typical considerations focused on by textbook IS-LM analysis. We will provide examples of how rational expectations investment decisions reverse standard predictions about the effects of both monetary and fiscal actions. Studies of the neoclassical determinants of investment stressing costs of changing the capital stock, beginning with those of Andrew Abel (1982) and Fumio Hayashi (1982), have provided an important update of the Jorgenson model. Taken together with follow-up studies, they affirm the connection between the determinants stressed by Jorgenson's neoclassical theory changes 9 Examples of the "standard real business cycle model" are given in Edward Prescott (1986) and Plosser (1989).

11 Robert G. King 77 in expectations of future production levels and implicit rental prices and the behavior of aggregate investment. However, the dynamics of the real world investment process also display a richness with costs of rapid adjustment, and lengthy intervals of production for capital goods which means we do not currently have a single model of investment that is as simple and empirically effective as the permanent income theory of consumption. Recent dynamic models of investment have illustrated the power of rational expectations on investment demand. One good example is afforded by work on the effects of persistent shifts in government purchases on real economic activity, within models that feature market-clearing processes for goods and labor (Ayigari, Christiano and Eichenbaum, 1990; Baxter and King, 1990). A standard IS-LM analysis would focus on the consumption decision to analyze how this fiscal action shifts the IS curve. Under the assumption that an increase in these purchases is financed by lump sum taxation (perhaps a cut in transfer payments), several levels of analysis arise. A first cut would use an estimate of the marginal propensity to consume out of current disposable income; a more sophisticated analysis would use the permanent income theory applied to consumption; a final and more sophisticated analysis would apply the permanent income perspective to choice of both consumption and work effort. These analyses will offer a range of consequences for the IS schedule: on net, private consumption would decline somewhat, government would increase, and some "crowding out" of private investment would appear likely. But this analysis would entirely miss the quantitatively largest influence identified in the two papers discussed above: higher expected future government demand induces a major upward shift in current private investment demand, of a sufficient scale that investment rises in equilibrium. 10 These investment demand effects are not limited to models with a marketclearing structure; they are also centrally important to the consequences of monetary policy in models with sticky wages or prices (Koenig, 1989; King, 1990). If changes in the money stock are persistent, then they lead to persistent changes in aggregate demand. With a rational expectations investment function of the neoclassical form, persistent changes in demand for final output lead to quantitatively major shifts in the investment demand schedule at given real interest rates. These effects are generally sufficiently important that real interest rates actually rise with a monetary expansion rather than fall: the IS curve effect outweighs the direct LM curve effect. Further, if persistent changes in the money stock are only gradually translated into price or wage increases, then 10 In the market clearing framework, the channels of influence are roughly as follows. First, individuals respond to the wealth loss occasioned by higher government demand by working more permanently at unchanged prices. Second, in labor market equilibrium, this increased supply is used by firms to produce goods. Third, with higher labor input, a constant returns to scale production function implies that there is a higher marginal product of capital at each capital stock. Thus, with a persistent disturbance, the higher future marginal product of capital schedule shifts up today's investment demand function.

12 78 Journal of Economic Perspectives there are very large additional expected inflation effects on nominal interest rates. Expectations and the Monetary Sector Economists' understanding of monetary dynamics is remarkably rudimentary, relative to our understanding of consumption and investment. At this stage, it is not possible to determine whether the LM curve displays expectations-induced shifts of quantitative importance. This is a critical area for future research. Put in slightly different terms, we need to know how to separate observed money demand residuals into (i) exogenous shifts in determinants of transactions demand patterns and transactions costs; and (ii) responses to omitted expectational elements. The IS-LM Model and Practical Macroeconomics The traditional logic of the IS-LM model shows up in a range of our professional activities, leading economists generally downplay the role of expectations and focus on direct effects of policy and other disturbances. This section discusses three interlocking areas of application: education, policy analysis, and econometrics. In each of these, recognizing the power of expectations means abandoning the mode of thought based on the standard IS-LM model. When the traditional IS-LM model is used as a teaching device about the consequences of policy disturbances, the teacher is making a decision about major and minor channels of influence. The two alternatives are perhaps best expressed, at present, by the difference between recent undergraduate texts: Robert Barro's (1984) text teaches the student about the permanent income theory of consumption before introducing the IS curve. Major alternatives, like the recent text by Mankiw (1992), do the opposite. Barro's approach stresses expectations as first-order determinants of shifts in aggregate demand, although from my perspective too little emphasis is given to the interactions of investment and expectations. The conventional view treats the direct effects of government purchase, monetary and other determinants as dominant. Accordingly, the student is led to view events that occur without directly observable sources as evidence of instability in private behavior, although the rational expectations research program has provided good reason to doubt this conventional view. Without systematic incorporation of expectations, the IS-LM model inevitably must treat private behavior as subject to large and frequent behavioral disturbances. This focus directs policy attention on the aggregate demand front towards actions aimed at response to current shocks. Correspondingly, it directs attention away from policies that stabilize aggregate demand through their influence on market expectations. Since expectations about future economic developments are a powerful determinant of aggregate demand, government policies can also have important consequences via these channels.

13 Will the New Keynesian Macroeconomics Resurrect the IS-LM Model? 79 The models that are used to develop the intuition of students and policy makers are ultimately based on econometric evidence. We now turn to consider a recent attempt to explore the consistency of the IS-LM model with postwar U.S. evidence. A good example of the recent intellectual straightjacket of the IS-LM framework is provided by Jordi Gali's recent exploration (1992) of how well the IS-LM model fits postwar U.S. data. To begin, let me say that Gali is pursuing exactly the right research strategy given the state of knowledge about macroeconomic models: he is comparing the implications of a specific simple theory to the dynamic paths for economic activity estimated to arise from disturbances identified by the researcher according to transparent simple methods. In fact, in recent work, King and Watson (1992) have developed formal econometric methods for evaluating individual economic models and choosing between models exactly on the basis of such comparisons. The substantive conclusions of Gali's investigations are as follows. First, he provides an empirical account of postwar U.S. business cycles that decomposes economic fluctuations into four basic sources of shocks: aggregate supply, money supply, money demand and IS shocks. In this decomposition, aggregate supply shocks account for 70 percent of fluctuations in gross national product at business cycle horizons (that is, forecasting one to ten quarters). Money supply and IS shocks account for most of the rest. For an adherent of real business cycles, it is tempting to dwell on the estimated role of supply shocks: even an empirical investigation of the IS-LM framework now suggests a dominant business cycle role of permanent supply shocks that have the effects stressed in real theories. In particular, a positive supply shock raises output permanently, raises real interest rates temporarily, lowers the price level permanently, and produces a temporary decline in the inflation rate. But we must move on. What are the IS shocks? Following the standard IS-LM perspective, Gali views these as behavioral shocks. If Keynes could have overcome his hostility to econometrics, he would have likely seen them as the outcome of "animal spirits" on investment demand: a positive IS shock has the effect of raising output and the nominal interest rate in the short run. But that identification strikes me as preposterous when one looks at the full dynamic path of macroeconomic variables laid out in Gali's analysis. The "IS shock" has the effect of raising the inflation rate: there is a surge in inflation within the first few quarters, but inflation rises permanently at all horizons, including the very long run. And the real interest rate declines with a positive IS shock. All in all, the dynamic responses look much like a monetarist treatment of the effects of a permanent change in the inflation rate. Higher expected inflation at every horizon lower the real cost of investing at an initial nominal interest rate: macroeconomic equilibrium thus requires higher nominal interest rates. The initial inflation surge is a natural consequence of the inverse effect of higher nominal interest rates on the demand for real balances. And there is a

14 80 Journal of Economic Perspectives modest short-run effect of unexpected inflation on output lasting roughly a year that is mirrored in a temporary decline in the real interest rate. The considerations arising in interpreting this set of econometric estimates is thus a very good example of exactly the problem that constantly arises when the IS-LM framework is used to discuss monetary policy issues. Because it is standard to regard expectations as empirically unimportant, its users are forced to view private sector behavior as subject to large shocks. If the IS shock discussed above was truly an outcome of animal spirits, then many would see a role for activist monetary policy to offset its influence on output. But more likely, monetary policy is the cause of not the cure for the IS shock. Concluding Comments New Keynesian macroeconomics has pushed the idea that traditional IS-LM analysis remains the best way to think about the determination of aggregate demand. This essay has challenged that view, arguing that expectations are typically omitted from that analysis, and that they are quantitatively important determinants of aggregate demand. But will the attempt to resurrect the IS-LM model divert the profession from this essential line of research? After two decades of rational expectations research, the agenda for academic researchers and policy economists should hold two main issues. First, economists should be as many are constructing new small-scale dynamic theoretical models to take the place of the IS-LM apparatus. And the synthesis that is envisioned by John Campbell (1991) is not very different from the one envisioned by King, Plosser and Rebelo (1988a, b). The framework will be the basic neoclassical model; rational expectations will play a central role; and the effects of various frictions will be systematically explored. Second, economists should be although few currently are constructing medium-size dynamic models that can serve as laboratories for thinking about the details of policy choice. Ultimately, that activity is best done as a joint product by researchers working in academic and policy institutions. The attempt to resurrect the IS-LM model is unlikely to have much direct effect on the first activity: it is simply too major a component of ongoing academic research. That fact lends an inevitability to the ultimate demise of the IS-LM model, but it leaves open the issue of how far away is the long run. But the insistence on the continued utility of the IS-LM framework will have a negative, direct effect on the second task: it will contribute to postponing the development of a new generation of macroeconometric policy models in which rational expectations play a central role. As a result of the intellectual capital accumulated under the rational expectations program, we are now at a point where it is feasible, though hardly costless, to execute this development. The danger is that macroeconomists and policy-makers will pay too much attention

15 Robert G. King 81 to the new Keynesian advertising, and assume for too long that the old product is a sound one. The author would like to thank without implicating Marianne Baxter, N. Gregory Mankiw, Charlie Plosser, and Timothy Taylor for useful comments at various stages of the preparation of this paper. References Abel, Andrew, "Dynamic Effects of Permanent and Temporary Tax Policies in a q Model metric Model with Rational Expectations in Fair, Ray, "An Analysis of a Macro-Econo- of Investment," Journal of Monetary Economics, the Stock and Bond Markets," American Economic Review, September 1979, 69:4, May 1982, 9:3, Ayigari, Rao, Martin Eichenbaum and Feige, Edgar, "Expectations and Adjustment in the Monetary Sector, American Eco- Lawrence Christiano, "The Output, Employment, and Interest Rate Effects of Government Consumption," Federal Reserve Bank of nomic Review, May 1967, 57:2, Fischer, Stanley, "Long Term Contracts, Minneapolis working paper, March Rational Expectations, and the Optimal Money Barro, Robert J., Macroeconomics, New York: Supply Rule," Journal of Political Economy, John Wiley and Sons, February 1977, 85:1, Barro, Robert J., " New Classicals and Keynesians, or the Good Guys and the Bad Guys," Flavin, Marjorie, "The Adjustment of Consumption to Changing Expectations about Fu- Schwiez Zeitschrift fur Volkswirtschaft und Statistik, 3/1989, ture Income." Journal of Political Economy, Oct. Baumol, William, "The Transactions Demand for Cash," Quarterly Journal of Economics, Friedman, Milton, "The Demand for 1981, 89:5, November 1952, 67:4, Money Some Theoretical and Empirical Results," Journal of Political Economy, August Baxter, Marianne and Robert G. King, "Fiscal Policy in General Equilibrium," University of Rochester working paper, Friedman, Milton, A Theory of the Consump- 1959, 67:4, Blanchard, Olivier, "New Classicals and tion Function. Princeton: Princeton University New Keynesians: The Long Pause," Federal Press, Reserve Bank of St. Louis conference paper, Gali, Jordi, "How well does the IS-LM October Model Fit the Post-War U.S. Data?" Quarterly Calvo, Guillermo, "Staggered Prices in a Journal of Economics, May 1992, 107:2, Utility-Maximizing Framework," Journal of Monetary Economics, September 1983, 12:3, Gordon, Robert J., "What is New Keynesian Economics?" Journal of Economic Literature, Campbell, John Y., "Inspecting the Mechanism: An Analytical Approach to the Stochas- Hall, Robert, "Stochastic Implications of the September 1990, 28, tic Growth Model," working paper, Princeton Life Cycle Permanent Income Hypothesis: University, Theory and Evidence," Journal of Political Campbell, John Y. and N. Gregory Mankiw, Economy, October 1978, 86:5, "Consumption, Income and Interest Rates: Hall, Robert E. and Frederick Mishkin, Reinterpreting the Time Series Evidence," "The Sensitivity of Consumption to Transitory Income: Estimates from Panel Data on NBER Macroeconomic Annual, 1989, Christiano, Lawrence J., "Comment," Households," Econometrica, March 1982, 50:2, NBER Macroeconomic Annual, 1989,

16 82 Journal of Economic Perspectives Hayashi, Fumio, "Tobin's Marginal q and Average q: A Neoclassical Interpretation," Econometrica, January 1982, 50:1, Hicks, John R., "Mr. Keynes and the 'Classics': A Suggested Interpretation," Econometrica, April 1937, 5, Jorgenson, Dale, "Capital Theory and Investment Behavior," American Economic Review, May 1963, 53:2, Keynes, John M., The General Theory of Employment, Interest and Money. London: McMillan Press, Keynes, John M., Tract on Monetary Reform. London: McMillan Press, King, Robert G., "Money and Business Cycles," working paper, University of Rochester, 1990, forthcoming Journal of Monetary Economics, Fall King, Robert G., Charles I. Plosser and Sergio T. Rebelo, "Production, Growth and Business Cycles, I: The Basic Neoclassical Model," Journal of Monetary Economics, 1990a, 21:2/3, King, Robert G., Charles I. Plosser and Sergio T. Rebelo, "Production, Growth and Business Cycles, II: New Directions," Journal of Monetary Economics, 1990b, 21:2/3, King, Robert G., and Mark W. Watson, "On the Econometrics of Comparative Dynamics," working paper, University of Rochester, August Klein, Lawrence, Economic Fluctuations in the United States. New York: John Wiley and Sons, Koenig, Evan, "Are the Permanent-Income Model of Consumption and the Accelerator Model of Investment Compatible?" working paper, Federal Reserve Bank of Dallas, Laidler, David E. W., The Demand for Money: Theories and Evidence. Scranton: International Textbook Company, Lucas, Robert E., "Econometric Policy Evaluation: A Critique." In Brunner, Karl Brunner and Allan Meltzer, eds., The Phillips Curve and Labor Markets. Carnegie-Rochester Conferences Series, Vol. 1, Amsterdam: North-Holland, Mankiw, N. Gregory, Macroeconomics. New York: Worth Publishers, Mankiw, N. Gregory, "A Quick Refresher Course in Macroeconomics," Journal of Economic Literature, December 1990, 28, Mankiw, N. Gregory, Jeffrey A. Miron and David N. Weil, "The Adjustment of Expectations to a Change in Regime: A Study of the Founding of the Federal Reserve," American Economic Review, June 1987, 77:3, Mankiw, N. Gregory and D. Romer, New Keynesian Macroeconomics. Cambridge: M.I.T. Press, Modigliani, Franco, "Liquidity Preference and the Behavior of Interest and Money," Econometrica, January 1944, 12:1, Modigliani, Franco and Richard E. Brumberg, "Utility Analysis and the Consumption Function: An Interpretation of Cross-Section Data." In Kurihara, K., ed., Post-Keynesian Economics. New Brunswick: Rutgers University Press, 1954, Patinkin, Don, Money, Interest and Prices. New York: Harper and Row, 1965, 2d edition. Pigou, Arthur C., Keynes's 'General Theory': A Retrospective View. London: McMillan and Company, Plosser, Charles I., "Understanding Real Business Cycles," Journal of Economic Perspectives, Summer 1989, 3:3, Prescott, Edward C., "Theory Ahead of Business Cycle Measurement." In Brunner, Karl and Allan Meltzer, eds., Real Business Cycles, Real Exchange Rates and Actual Policies. Carnegie-Rochester Conferences Series, Fall 1986, 25. Amsterdam: North-Holland, Rotemberg, Julio J., "Sticky Prices in the United States," Journal of Political Economy, December 1982, 90:6, Sargent, J. Thomas and Neil Wallace, "Rational Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule," Journal of Political Economy, March/April 1975, 83:2, Taylor, John, "Aggregate Dynamics and Staggered Contracts," Journal of Political Economy, February 1980, 85:1, Tinbergen, Jan, Business Cycles in the United States of America. Geneva: League of Nations, Tobin, James, "The Interest Elasticity of the Transactions Demand for Cash," Review of Economics and Statistics, August 1956, 38:3,