NBER WORKING PAPER SERIES NEW CLASSICALS AND KEYNESIANS, OR THE GOOD GUYS AND THE BAD GUYS Robert J. Barro Working Paper No.2982 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 may,1989 Prepared for presentation at the 125th anniversary meeting of the Swiss Economic Association, April 1989. This paper is part of NBER's research program in Economic Fluctuations. Any opinions expressed are those of the author not those of the National Bureau of Economic Research
NBER Working Paper #2982 May 1989 NEW CLASSICALS AND KEYNESIANS, OR THE GOOD GUYS AND THE BAD GUYS Old-style Keynesian models relied on sticky prices or wages to explain unemployment and to argue for demand-side macroeconomic policies. This approach relied increasingly on a Phillips-curve view of the world, and therefore lost considerable prestige with the events of the 1970s. The new classical macroeconomics began at about that time, and focused initially on the apparent real effects of monetary disturbances. Despite initial successes, this analysis ultimately was unsatisfactory as an explanation for an important role of money in business fluctuations. Nevertheless, the approach achieved important methodological advances, such as rational expectations and new methods of policy evaluation. Subsequent research by new classicals has deemphasized monetary shocks, and focused instead on real usiness cycle models and theories of endogenous economic growth. These eas appear promising at this time. another development is the so-called new Keynesian economics, which includes long-term contracts, menu costs, efficiency wages and insider-outsider theories, and macroeconomic models with imperfect competition. Although some of these ideas may prove helpful as elements in real business cycle models, my main conclusion is that the new Keynesian economics has not been successful in rehabilitating the Keynesian pproach Robert J. Barro Harvard University Department of Economics Littauer Center Cambridge, MA 02138
Keynesian Model When I was a graduate student at Harvard in the late 1960s, the Keynesian model was the only game in town as far as macroeconomics was concerned Therefore, while I had doubts about the underpinnings of this analysis, it seemed worthwhile to work within the established framework to develop a model that was logically more consistent and hopefully empirically more useful Collaborating with Herschel Grossman(Barro and Grossman, 1971), we made some progress in clarifying and extending the Keynesian model. But that research also made obvious the dependence of the central results on fragile underlying assumptions. The model stressed the failure of private enterprise economies to ensure full employment and production, and the consequent role for active macro policies as instruments to improve outcomes. Shocks to aggregate demand--but not aggregate supply--were the key to business fluctuations, and mere changes in optimism or pessimism turned out to be self fulfilling These properties, which seem odd to economists who think in terms of price heory and well-functioning private markets, suggest coordination problems on a grand scale. But this perspective hardly accords with the basic source of market failure that characterizes the standard Keynesian model. It is the mere stickiness of prices or wages, primarily in the downward direction, that accounts for the principal results. Of course, many macroeconomists think of price stickiness as an as if device--a problem that is not to be viewed literally, but instead as a proxy for serious matters, such as incomplete information,adjustment costs, and other problems of coordination among economi ic agents. But this viewpoint has not been borne out by subsequent research. For example, the incorporation of these serious matters does not
support the Keynesian stress on aggregate demand, and also does not provide normative basis for activist government policies of the usual Keynesian type One important function of a macroeconomic model is to isolate the sources of disturbances that cause aggregate business fluctuations. Keynesi analyses focus on shocks to aggregate demand, and typically attribute these shocks either to governmental actions(disruptive or corrective fiscal and monetary policies, or to shifts in private preferences that influence consumption or investment demand. Keynes's own discussion(Keynes, 1935 hapter 12)referred to the I spirits"of businessmen, and the consequent volatility of investment demand due to shifting moods of optimism or pessimism. Thus, aside from governmental actions, the Keynesian model is not strong at pinpointing observable, objective events that cause recessions or boom One reason that Keynes may not have been troubled by this " deficiency" is that he te economy as inherently unstable. It did not take large(and presumably objectively observable)shocks to trigger a recession because even a small shock--when interacting with the multiplier(and, in some models, also the investment accelerator )--could generate a significant and sustained drop in output and employment. Curiously, however, later Keynesian developments deemphasized the multiplier. For example, in the well-known IS/LM model (in which interest rates ad just and matter for aggregate demand)or in Keynesian analyses that incorporate some version of the permanent- income hypothesis, multipliers need not exist. These extensions do improve the model's fit with some facts about business cycles ch as the apparent absence of a multiplie e of output changes in government purchases and the relative stability of consumption
3 over the business cycle. But the elimination of the multiplier means also that large responses of output, as in a substantial recession, require large impulses; hence, it again becomes important to identify the kinds of shocks that typically matter for aggregate fluctuations I think that the desire to find observable, aggregate shocks motivated many Keynesians--although not Keynes nor many of his immediate followers--to assign a substantial weight to monetary disturbances as a source of the business cycle. Within a framework where prices adjust slowly and output is determined by aggregate demand, it is easy to conclude that an increase in money raises output and also leads gradually to a higher price level Moreover, the positive correlation between money and output--and perhaps between the price level and output--showed up in some data During the 1960s and early 1970s, Keynesian analysis became increasingly identified with this Phillips curve- view of the world. Thus, this analysis also lost considerable prestige when the Phillips curve disappeared in the mid 1970s; the rise in unemployment along with the increasing rate of inflation was difficult to explain in this kind of model. New Keynesians have,however,demonstrated their flexibility by arguing that the old ynesian model merely need to be patched up to incorporate the supply side Bu argument does not work. In a single market, one can think of qua as determined by demand with the excess supply rationed--as in the Keynesian model--so that changes in quantity depend only Then if this situation applies to the majority of markets, one can generate orthodox Keynesian prescriptions for demand-oriented governmental policies Alternatively, quantity in a typical market could be determined by supply with the excess demand rationed--as in markets subject to effective price
controls--so that movements in quantity depend only on shocks to supply. If this situation holds for the majority of markets, one again gets prescriptions for the government's macro policies, but they are basically opposite to the ose from the K odel. The serious alternative to either of these two polar cases is a framework where demand and supply are somehow balanced or equilibrated on the various markets. Although I regard this equilibrium approach as the logical way to think about macroeconomics, this approach - pursued by new clas macroeconomists-- turns out to be inconsistent with basic Keynesian themes assia Approach The new classical macroeconomics, sometimes referred to as rationa expectations macroeconomics or as the equilibrium approach to macroeconomics, egan with Bob Lucas's research(Lucas, 1972, 1976) in the early 1970s. A guiding discipline of this work was that economic agents acted rationally in the context of their environment; notably that people assembled and used nformation in an efficient manner. Although the approach stressed fully worked out equilibrium theories, the analysis was directed at explaining eal-world business fluctuations. The basic viewpoint implied that it would be unsatisfactory to explain"these fluctuations by easily correctable market failures, such as those present in Keynesian models. Hence fluctuations had to reflect real or monetary disturbances, whose dynamic economic effects depended on costs of obtaining information, costs of ad justment, and so on The biggest challenge to the new classical approach was to explain why money was non-neutral, and, in particular, why monetary disturbances played a
major role in business cycles. This area was a significant challenge because first, it seemed to be empirically important, and second, the equilibriu framework with flexible prices tends to generate a close approximation to monetary neutrality Initially, the approach seemed to achieve notable successes. On theoretical level, short-term real effects of monetary disturbances could arise from imperfect information about money and the general price level Monetary shocks, which affected the general price level in the same direction, could be temporarily misperceived as shifts in relative prices which led to ad justments in the supply of labor and other quantities. These real effects vanished in the long run, but could persist for awhile because of information lags and costs of adjusting the quantities of factor inputs On the other hand, anticipated monetary changes--which include systematic monetary policies--would not matter because they did not lead to informational confusions(Sargent and Wallace, 1975) On an empirical level, there was also evidence that appeared to support the approach. Monetary disturbances seemed to be important sources of business fluctuations, and there was some indication that it was mainly the unanticipated or surprise part of monetary movements that mattered for real variables(Barro,1981).Some cross-country evidence supported the theoretical predictions concerning the relation between the volatility of money and the slopes of estimated Phillips curves( Kormendi and Meguire The theory was al sistent with the observed absence of a substantial long-term relationship between real economic performance and the growth rates of money and prices; that is, with the absence of a long-run Phillips curve
Further investigations cast doubt on these successes. First, the informational lag in observing money and the general price level did not seem to be very important. If incomplete information about money and the general price level mattered a lot for economic decisions, people could expend relatively little effort to find out quickly about these variables. Second the theory did not do so well in terms of its predictions about monetary effects on real interest rates, real wage rates, and consumption. Third, the predicted Phillips curve-type relation between price surprises and real economic activity basically disappeared after the early 1970s. Fourth, the sitive relation between monetary shocks and output shows up most clearly with broad monetary aggregates. The relation with narrow aggregates, such as the monetary base, is much weaker The upshot of these arguments is that the new classical approach does not do very well in accounting for an important role of money in business fluctuations. However, this failing may not be so serious because the empirical evidence on the causal role of money for real variables seems also to have been overstated. In other words, the accounting for major short-run non-neutralities of money was a misplaced priority for the new classical approach. Some empirical evidence supports this conclusion; for example, the observation that the correlation of real economic activity with broad monetary aggregates is greater than that with the monetary base or the price level, or the finding that real effects from the quantity of money are weak once the behavior of nominal interest rates is held constant. These results suggest that endogenous responses of money--partly from the behavior of policymakers and partly from the workings of the financial system--may
account for most of the correlations between money and real economic activity This verdict does not invalidate some of the major successes of the new lassical approach. In terms of methodology, these successes include the application of equilibrium modeling to macroeconomic analysis, the use of rational expectations as part of this modeling, and the revolution ir pproaches licy evaluation, One specific application in which the equilibrium approach has achieved some success is in analyses of fiscal policy(see Barro, 1989b, for a survey). Some of this research revolves around the Ricardian equivalence theorem, which provides conditions under which substitutions of budget deficits for taxes are of no consequence. I But further developments have brought out the real effects from government purchases and public services, the composition and timing of distorting taxes, and so on. Another interesting off-shoot from the new classical approach is the application of game theory to the interaction between government policymakers and the private sector. The results here involve the distinction between rules and discretion, and the related roles of commitment, credibility, and reputation(see, Rogoff, 1989, for a survey). Some of the early analyses in this area dealt with monetary models; specifically, with the Phillips curve and the tradeoff between unemployment and inflation. But subsequent applications, such as to tax and regulatory policies and to international debt, do not rely on an important role for money in business fluctuations 1 Bartley(1989) claims more than I would by describing Ricardian uivalence as "an Exocet aimed at the heart of the Keynesian notion that deficits stimulate the economy
8 Real Business Cycle Theory With the deemphasis on monetary models of the business cycle proponents of the new classical approach have moved over the last to ten years to analyses that rely on real disturbances as sources of business fluctuations(see Kydland and Prescott, 1982, and the survey by McCallum 1989). These models stress technology shocks or other disturbances to the supply side as central driving forces, but allow an important role for the dynamic elements that influence the ways that shocks propagate are equilibrium in style, featuring cleared, competitive markets; optimizing agents who are typically modeled as representative households with infinit and neoclassical production functions that are sub ject stochastic disturbances. Although the models deemphasize monetary shocks the analysis of propagation mechanisms would apply as much to monetary models as to real models. In the real business cycle (or rBC)framework, any positive correlation between output and money reflects the endogenous response of monetary aggregates(see King and Plosser, 1984) A number of authors have simulated versions of RBC models on u.s. data he underlying parameters of preferences and technology are calibrated to be consistent with findings from cross-sectional studies. In many respects the results accord with observed characteristics of business cycle For example, RBC models can get right the relative variances of consumption, nvestment, capital stocks, and worker hours; and also account for the procyclical behavior of these variables. However, the models tend to overstate the procyclical patterns of hours, productivity, real interest rates, and real wage rates. In addition, to explain the standard deviation output growth, the models require a standard deviation for technological