An Exploration of Optimal Stabilization Policy

This paper examines the optimal response of monetary and fiscal policy to a decline in aggregate demand. The theoretical framework is a two-period general equilibrium model in which prices are sticky in the short run and flexible in the long run. Policy is evaluated by how well it raises the welfare of the representative household. Although the model has Keynesian features, its policy prescriptions differ significantly from those of textbook Keynesian analysis. Moreover, the model suggests that the commonly used “bang for the buck” calculations are potentially misleading guides for the welfare effects of alternative fiscal policies. w is the optimal response of monetary and fiscal policy to an economy-wide decline in aggregate demand? This question has been at the forefront of many economists’ minds for decades, but especially over the past few years. In the aftermath of the housing bust, financial crisis, and stock market decline of the late 2000s, households and firms were less eager to spend. The decline in aggregate demand for goods and services led to the most severe recession in a generation or more. The textbook answer to such a situation is for policymakers to use the tools of monetary and fiscal policy to prop up aggregate demand. And, indeed, during this recent episode the Federal Reserve reduced the federal funds rate, its primary policy instrument, almost all the way to zero. With monetary policy having used up its ammunition of interest rate cuts, economists and policymakers increasingly looked elsewhere for a solution. In particular, they focused on fiscal policy and unconventional instruments of monetary policy. Traditional Keynesian economics suggests a startlingly simple solution: the government can increase its spending to make up for the shortfall in Copyright 2011, The Brookings Institution 210 Brookings Papers on Economic Activity, Spring 2011 private spending. Indeed, this was one of the motivations for the stimulus package proposed by President Barack Obama and passed by Congress in early 2009. The logic behind this policy should be familiar to anyone who has taken a macroeconomics principles course anytime over the past half century. Yet many Americans (including quite a few congressional Republicans) are skeptical that increased government spending is the right policy response. Their skepticism is motivated by some basic economic and political questions: If we as individual citizens are feeling poorer and cutting back on our spending, why should our elected representatives in effect reverse these private decisions by increasing spending and going into debt on our behalf? If the goal of government is to express the collective will of the citizenry, shouldn’t it follow the lead of those it represents by tightening its own belt? Traditional Keynesians have a standard answer to this line of thinking. According to the paradox of thrift, increased saving may be individually rational but collectively irrational. As individuals try to save more, they depress aggregate demand and thus national income. In the end, saving might not increase at all. Increased thrift might lead only to depressed economic activity, a malady that can be remedied by an increase in government purchases of goods and services. The goal of this paper is to address this set of issues in light of modern macroeconomic theory. Unlike traditional Keynesian analysis of fiscal policy, modern macro theory begins with the preferences and constraints facing households and firms and builds from there. This feature of modern theory is not a mere fetish for microeconomic foundations. Instead, it allows policy prescriptions to be founded on the basic principles of welfare economics. This feature seems particularly important for the case at hand, because the Keynesian recommendation is to have the government undo the actions that private citizens are taking on their own behalf. Figuring out whether such a policy can improve the well-being of those citizens is the key issue, and a task that seems impossible to address without some reliable measure of welfare. The model we develop to address this question fits solidly in the New Keynesian tradition. That is, the starting point for the analysis is an intertemporal general equilibrium model that assumes prices to be sticky in the short run. This temporary price rigidity prevents the economy from reaching an optimal allocation of resources, thus giving monetary and fiscal policy a possible role in helping the economy reach a better allocation through their influence on aggregate demand. The model yields several significant conclusions about the best responses of policymakers n. gregory mankiw and matthew weinzierl 211 under various economic conditions and constraints on the set of policy tools at their disposal. To be sure, by the nature of this kind of exercise, the validity of any conclusion depends on whether the model captures the essence of the problem being examined. Because all models are simplifications, one can always question whether a conclusion is robust to generalization. Our strategy is to begin with a simple model that illustrates our approach and yields some stark results. We then generalize this baseline model along several dimensions, both to check its robustness and to examine a broader range of policy issues. Inevitably, any policy conclusions from such a theoretical exploration must be tentative. In the final section we discuss some of the simplifications we make that might be relaxed in future work. Our baseline model is a two-period general equilibrium model with sticky prices in the first period. The available policy tools are monetary policy and government purchases of goods and services. Like private consumption goods, government purchases yield utility to households. Private and public consumption are not, however, perfect substitutes. Our goal is to examine the optimal use of the tools of monetary and fiscal policy when the economy finds itself producing below potential because of insufficient aggregate demand. We begin with the benchmark case in which the economy does not face the zero lower bound on nominal interest rates. In this case the only stabilization tool that is necessary is conventional monetary policy. Once monetary policy is set to maintain full employment, fiscal policy should be determined based on classical principles. In particular, government consumption should be set to equate its marginal benefit with the marginal benefit of private consumption. As a result, when private citizens are cutting back on their private consumption spending, the government should cut back on public consumption as well. We then examine the complications that arise because nominal interest rates cannot be set below zero. We show that even this constraint on monetary policy does not by itself give traditional fiscal policy a role as a stabilization tool. Instead, the optimal policy is for the central bank to commit to future monetary policy actions in order to increase current aggregate demand. Fiscal policy continues to be set on classical principles. A role for countercyclical fiscal policy might arise if the central bank both hits the zero lower bound on the current short-term interest rate and is unable to commit itself to expansionary future policy. In this case monetary policy cannot maintain full employment of productive resources on its own. Absent any fiscal policy, the economy would find itself in a 212 Brookings Papers on Economic Activity, Spring 2011 nonclassical short-run equilibrium. Optimal fiscal policy then looks decidedly Keynesian if the only instrument of fiscal policy is the level of government purchases: increase those purchases to increase the demand for idle productive resources, even if the marginal value of the public goods being purchased is low. This very Keynesian result, however, is overturned once the set of fiscal tools available to policymakers is expanded. Optimal fiscal policy in this situation is one that tries to replicate the allocation of resources that would be achieved if prices were flexible. An increase in government purchases cannot accomplish that goal: although it can yield the same level of national income, it cannot achieve the same composition of it. We discuss how tax instruments might be used to induce a better allocation of resources. The model suggests that tax policy should aim at increasing the level of investment spending. Something like an investment tax credit comes to mind. In essence, optimal fiscal policy in this situation tries to produce incentives similar to what would be achieved if the central bank were somehow able to reduce interest rates below zero. A final implication of the baseline model is that the traditional fiscal policy multiplier may well be a poor tool for evaluating the welfare implications of alternative fiscal policies. It is common in policy circles to judge alternative stabilization ideas using “bang-for-the-buck” calculations. That is, fiscal options are judged according to how many dollars of extra GDP are achieved for each dollar of extra deficit spending. But such calculations ignore the composition of GDP and therefore are potentially misleading as measures of welfare. After developing these results in our baseline model, we examine three variations. First, we add a third period. We show how the central bank can use long-term interest rates as an additional tool to achieve the flexible-price equilibrium. Second, we add government investment spending to the baseline model. We show that all government expenditure follows classical principles when monetary policy is sufficient to stabilize output. Moreover, even when monetary policy is limited, the model does not point toward government investment as a particularly useful tool for putting idle resources to work. Third, we modify the baseline model to include nonRicardian, rule-of-thumb households who consume a constant fraction of income. The presence of such households means that the timing of taxes may affect output, and we characterize the optimal policy mix in that setting. We find that the description of the equilibrium closely resembles the traditional Keynesian model, but the prescription for optimal policy can differ substantially from the textbook answer. n. gregory mankiw and matthew weinzierl 213 I. Introducing the Model In this section we introduce the elements of the baseline model. Before delving into the model’s details, it may be useful to describe how this model is related to a few other models with which readers may be familiar. Our goal is not to provide a completely new model of stabilization policy but rather to illustrate conventional mechanisms in a way that permits an easier and more transparent analysis of the welfare implications of alternative policies. First, the model is closely related to the model of short-run fluctuations found in most leading undergraduate textbooks. Students are taught that prices are sticky in the short run but flexible in the long run. As a result, the economy can temporarily deviate from its full-employment equilibrium, yet over time it gravitates toward full employment. Similarly, we will (in a later section) impose a sticky price level in the first period but allow future prices to be flexible. Second, this model is closely related to the large literature on dynamic stochastic general equilibrium (DSGE) models. Strictly speaking, the model is not stochastic: we will solve for the deterministic path of the economy after one (or more) of the exogenous variables changes. But the spirit of the model is much the same. As in DSGE models, all decisions are founded on underlying preferences and technology. Moreover, all decisionmakers are forward looking, so their actions will depend not only on current policy but also on the policy they expect to prevail in the future. There is, however, a key methodological difference between our approach and that in the DSGE literature. In recent years that literature has evolved in the direction of greater complexity, as researchers have attempted to match various moments of the data more closely. (See, for example, Christiano, Eichenbaum, and Evans 2005 and Smets and Wouters 2003.) By contrast, our goal is greater simplicity and transparency so that the welfare implications of alternative monetary and fiscal policies can be better illuminated. Third, the model we examine is related to the older literature on “general disequilibrium” models, such as those of Robert Barro and Herschel Grossman (1971) and Edmond Malinvaud (1977). As in these models, we will assume that the price level in the first period is exogenously stuck at a level that is inconsistent with full employment of productive resources. At the prevailing price level, there will be an excess supply of goods. But unlike this earlier literature, our model is explicitly dynamic. That is, we emphasize the role of forward-looking, intertemporal behavior in determining current spending decisions and the impact of policy. 214 Brookings Papers on Economic Activity, Spring 2011 I.A. Households The economy is populated by a large number of identical households. The representative household has the following objective function: ( ) max , 1 1 1 2 2 u C v G u C v G ( ) + ( ) + ( ) + ( ) [ ] { } β where Ct is consumption in period t, Gt is government purchases, and b is the discount factor. Households choose consumption but take government purchases as given. Households derive all their income from their ownership of firms. Each household’s consumption choices are limited by a present-value budget constraint: ( ) , 2 1 0 1 1 1 1 2 2 2 2 1 P T C P T C i Π Π − − ( ) + − − ( ) + ( ) = where Pt is the price level, Pt is profits of the firm, Tt is tax payments, and i1 is the nominal interest rate between the first and second periods. Implicit in this budget constraint is the assumption of a bond market in which households can borrow or lend at the market interest rate. I.B. Firms Firms do all the production in the economy and provide all household income. It is easiest to imagine that the number of firms is the same as the number of households and that each household owns one firm. For simplicity, we assume that capital K is the only factor of production. In each period the firm produces output with an AK production function, where A is an exogenous technological parameter. The firm begins with an endowment of capital K1 and is able to borrow and lend in financial markets to determine the future capital stock K2. Without loss of generality, we assume that capital fully depreciates each period, so investment in the first period equals the capital stock in the second period. The parameter A plays a key role in our analysis. In particular, we are interested in studying the optimal policy response to a decline in aggregate demand, and in our model the most natural cause of such a decline is a decrease in the future value of A. Such an event can be described as a decline in expected growth, a fall in confidence, or a pessimistic shock to “animal spirits.” In any event, in our model it will tend to reduce wealth and current aggregate demand, as well as reducing the natural rate of interest n. gregory mankiw and matthew weinzierl 215 (that is, the real interest rate consistent with full employment). A similar set of events would unfold if the shock were to households’ discount factor b, but it seems more natural to assume stable household preferences and changes in the expected technology available to firms. Before proceeding, it might be worth commenting on the absence of a labor input in the model. That omission is not crucial. As we will describe more fully later, it could be remedied by giving each household an endowment of labor in each period and making the simplifying assumption that capital and labor are perfect substitutes in production. That somewhat more general model yields identical results regarding monetary and fiscal policy. Therefore, to keep the results as clean and easily interpretable as possible, we will focus on the one-factor case. Firms choose the second period’s capital stock to maximize the present value of profits, discounting the second period’s nominal profit by the nominal interest rate: max . K P P i 2 1 1 2 2 1 1 Π Π + + ( )      