|
APPENDIX
Evidence on the Economic Effects of Fiscal Stimulus
The Congressional Budget Office (CBO) based its estimates of the economic effects of the American Recovery and Reinvestment Act of 2009 (ARRA) on information from a variety of sources: macroeconometric forecasting models, general-equilibrium models, and direct extrapolations of past data. Macroeconometric forecasting models incorporate relationships between aggregate economic variables that are based largely on historical evidence. General-equilibrium models, by contrast, are built on explicit assumptions about the decisionmaking of individuals and businesses. Another source of informationon the economic effects of fiscal stimulus is research that makes projections for the future by directly examining the correlations between economic variables in the past or by evaluating the effects of specific types of policy events in the past.
Macroeconometric Forecasting Models
In analyzing the economic effects of ARRA, CBO drew heavily on versions of the commercial forecasting models of two economic consulting firms, Macroeconomic Advisors and Global Insight, as well as on the FRB-US model used at the Federal Reserve Board. Those models assume that the economy has an underlying potential output determined by the size of the labor supply, the capital stock, and technology. They also assume that actual output can change relative to potential output because of shifts in aggregate demand for goods and services from households, businesses, and the government. With those basic assumptions, the details of interactions between economic variables in the models are based largely on historical relationships, informed by theories of how those variables are determined (for example, the theory that total consumption depends mostly on disposable income, wealth, and interest rates).1 Because they emphasize the influence of aggregate demand on output in the short run, the macroeconometric forecasting models tend to predict greater economic effects from demand-enhancing policies such as ARRA than some other types of models do.
Macroeconometric forecasting models of this sort are widely used, and they underlie most of the forecasts offered to paying clients of economic consulting firms. In addition, the models that CBO uses generally produce results that are roughly in line with the consensus of private-sector forecasters, as compiled in the Blue Chip Economic Indicators. However, some analysts criticize this sort of model for being based on historical relationships between aggregate economic variables, such as income and consumption, rather than being built up from clearly specified rules governing the behavior of households and firms. In particular, some critics argue that models based on historical relationships will not provide accurate predictions in the face of new policies or new circumstances. To address that concern and to reflect current economic conditions—in which uncertainty about the financial and economic outlook remains high, and interest rates are low and are expected to remain so for some time—CBO altered the models’ usual formulation to reduce the extent to which interest rates respond to increases in output.2
Some skeptics of the efficacy of stimulus policies have cited the results of an alternative class of models, which tend to imply more-modest economic effects for such policies. In those models, people are assumed to make decisions about how much to work, buy, and save on the basis of current and expected future values of the wage rate, interest rates, taxes, and government purchases, among other things. In the basic form of such models, stimulus policies tend to crowd out a significant amount of other economic activity, and multipliers tend to be less than 1—meaning that stimulative policies have less than a dollar-for-dollar impact on output.
Although some analysts favor the rigor of that approach to modeling behavior, other analysts view the assumptions underlying households’ and businesses’ decisionmaking in those models to be unrealistic and leading to unrealistic predictions. In particular, this type of model generally assumes that people are fully rational and forward-looking, basing their current decisions on a full lifetime plan. The forward-looking assumption implies that people expect to eventually pay for any increased government spending or reduced revenues in the form of future tax increases and that they incorporate those expected payments—even if far in the future—into their current spending plans. Thus, they are assumed to reduce their consumption when government spending rises, because their lifetime income has fallen by the amount of the eventual taxes. For the same reason, cash transfer payments and tax refunds have little or no effect on current consumption in such models. People in the models generally also have full access to credit markets, so they can borrow to maintain their consumption when faced with a temporary loss of income. This class of models does not typically incorporate involuntary unemployment: People can work as many hours as they choose at the wage rate determined by the market. Finally, in these models, monetary policy usually follows a fixed rule by which increased output or inflation implies higher real interest rates.
Recent research has shown that relaxing some of those modeling assumptions can result in much higher multipliers.3 CBO has incorporated the results of that research into its view of the effects of government policies. However, the research results appear to be too dependent on particular assumptions for CBO to rely on them heavily.
Extrapolations from Historical Data
Another type of research uses historical data to directly project how government policies will affect the economy on the basis of how economic variables such as output and consumption have behaved relative to government spending and revenues in the past. However, estimates of economic effects from this research vary widely and are sensitive to the time period and estimation strategy used. Many estimates of this sort suggest that in the case of government purchases, crowding-out effects dominate, and the impact on output tends to be less than one for one and tends to fade over time. Some estimates, however, suggest multipliers higher than the range estimated by CBO. Estimated multipliers for tax cuts are generally higher than those for spending and tend to grow over time.4
One pitfall of this approach is that the direction of causation between policies and the economy is not always clear. For example, poor economic conditions can prompt the government to enact policies such as ARRA in an effort to boost economic activity. If weak economic performance led to such a policy, it would not be accurate to ascribe that performance to the policy, rather than vice versa. Likewise, if states and localities reduced purchases and laid off employees when their budgets deteriorated in a recession, it would not be accurate to blame the cuts in government spending for causing the recession. When causation runs in both directions in this way, the historical correlation between variables may not be a good guide for predicting the effects of a newly proposed policy.
A strategy that has been used to try to overcome that obstacle is to identify policies, such as wartime spending, that are arguably unrelated to other economic conditions and try to isolate their impact on the economy. Wartime spending, however, may not be indicative of the effects of other increases in government spending. For example, during World War II, the rationing of many goods may have reduced the indirect effects of government spending on private consumption. More generally, historical evidence shows the effects of policies under average economic conditions. Under current conditions—in which interest rates are apt to be less affected than usual by expansionary government policies, and there are high levels of idle resources—effects may be greater than they were, on average, in the past.
The FRB-US model differs from the other two forecasting models that CBO used in that it explicitly incorporates the influence of expected future developments on current outcomes.
Stimulative policies such as ARRA can lead to higher interest rates in two ways. First, if they increase economic activity, they can prompt the Federal Reserve to raise interest rates to combat inflation. Currently, however, that effect is likely to be smaller than usual. The federal funds rate (the interest rate directly controlled by the Federal Reserve) is near zero and is unlikely to rise until economic conditions have substantially improved. Interest rates on short-term government securities tend to move closely with the federal funds rate, so they are also unlikely to rise. For that reason, CBO estimates that expansionary government policies are likely to have less effect on interest rates now than under more-normal conditions, which implies less crowding out. (With the federal funds rate as low as possible, the Federal Reserve has used other policies to try to increase the availability of credit in order to stimulate economic activity. If ARRA caused the Federal Reserve to reduce those efforts, the law’s effects would be offset to some extent even without affecting interest rates; whether the Federal Reserve would indeed respond in that way under current financial and economic conditions is unclear.) Second, stimulative policies can influence longer-term interest rates if they create expectations of higher interest rates or inflation in the future. Policies that imply steep increases in future deficits may lead to higher current interest rates to the extent that people expect that the deficits will crowd out private investment and result in a lower capital stock (which tends to imply both higher rates of return on capital and higher interest rates). However, the policies in ARRA are temporary and thus are unlikely by themselves to have a major impact on the size of the capital stock or interest rates in the future.
For examples of model estimates that incorporate a lower-than-usual response of interest rates to policy changes, see Lawrence Christiano, Martin Eichenbaum, and Sergio Rebelo, When Is the Government Spending Multiplier Large? Working Paper No. 15394 (Cambridge, Mass.: National Bureau of Economic Research, October 2009); Troy Davig and Eric M. Leeper, Monetary–Fiscal Policy Interactions and Fiscal Stimulus, Working Paper No. 15133 (Cambridge, Mass.: National Bureau of Economic Research, July 2009); and Robert E. Hall, By How Much Does GDP Rise If the Government Buys More Output? Working Paper No. 15496 (Cambridge, Mass.: National Bureau of Economic Research, November 2009). For examples of models that include liquidity-constrained agents, see Jordi Gali, J. David López-Salido, and Javier Vallés, “Understanding the Effects of Government Spending on Consumption,” Journal of the European Economic Association, vol. 5, no. 1 (March 2007), pp. 227–270; and Marco Ratto, Werner Roeger, and Jan in’t Veld, “An Estimated Open-Economy DSGE Model of the Euro Area with Fiscal and Monetary Policy,” Economic Modelling, vol. 26, no. 1 (January 2009), pp. 222–233. For model estimates in which government spending can contribute to future production, see Eric M. Leeper, Todd B. Walker, and Shu-Chun Susan Yang, Government Investment and Fiscal Stimulus in the Short and Long Runs, Working Paper No. 15153 (Cambridge, Mass.: National Bureau of Economic Research, July 2009).
See Olivier Blanchard and Roberto Perotti, “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output,” Quarterly Journal of Economics (November 2002), pp. 1329–1368; Andrew Mountford and Harald Uhlig, What Are the Effects of Fiscal Policy Shocks? Working Paper No. 14551 (Cambridge, Mass.: National Bureau of Economic Research, December 2008); Roberto Perotti, In Search of the Transmission Mechanism of Fiscal Policy, Working Paper No. 13143 (Cambridge, Mass.: National Bureau of Economic Research, June 2007); Valerie Ramey and Matthew Shapiro, “Costly Capital Reallocation and the Effects of Government Spending,” Carnegie–Rochester Conference Series on Public Policy, vol. 48, no. 1 (June 1998), pp. 145–194; and Robert J. Barro and Charles J. Redlick, Macroeconomic Effects from Government Purchases and Taxes, Working Paper No. 15369 (Cambridge, Mass.: National Bureau of Economic Research, September 2009). In interpreting the results of this research, it is important to note that the reported multipliers are generally “peak” multipliers—that is, the largest effect on output in any one quarter of a dollar change to policy that persists consistent with historical behavior—rather than the cumulative effect of a one-time dollar’s worth of policy change, as CBO defines its multipliers.
|
|