Introduction and summary
A classic question facing macroeconomists is: How does an increase in government purchases affect the economy? Our interest in this question is motivated by the desire to evaluate the properties of different rules and institutions for setting fiscal policy. For example, should government purchases vary systematically over the business cycle? What would the macroeconomic consequences of a balanced budget amendment be? What would the effect of a permanent decline in defense purchases be on aggregate employment and real wages? If we had observations on otherwise identical economies operating under the different fiscal policies that we are interested in evaluating, it would be easy to answer these types of questions. But we do not. So we have no choice but to attack them within the confines of economic models.
Which model should we use? We have at our disposal a plethora of competing business cycle models, each of which incorporates different views of the way the economy functions and makes different recommendations for macroeconomic policy. So one's views about the costs and benefits of different policy proposals depends critically on the model being used to assess the proposal. In this sense, research aimed at assessing the empirical plausibility of competing models is a crucial input to the policy process. One approach for choosing among competing models is to compare their predictions for the consequences of a shock for which we know how the actual economy responds.(1) To the extent that different models give rise to different predictions, some will be counterfactual and can be eliminated from the field of choice.
Shocks to government spending are likely to be useful in this regard. This is because many models give rise to different predictions for the effects of an increase in government purchases on real wages and average labor productivity (output per man hour). Neoclassical models of the sort discussed in Barro (1981), Aiyagari, Christiano, and Eichenbaum (1992) and Edelberg, Eichenbaum, and Fisher (1998) assume constant returns to scale and perfect competition. Models of this sort predict that real wages fall after an exogenous increase in government purchases, that is, after a change in government purchases that was not caused by other developments in the economy. For reasons discussed below, other models which deviate from the assumptions embedded in the neoclassical model generate different predictions. For example, models embodying increasing returns and imperfect competition of the sort considered by Devereaux, Head, and Lapham (1996) and Rotemberg and Woodford (1992) predict that real wages ought to rise. Which of the two predictions is correct?
Competing business cycle models also give rise to different predictions for how average labor productivity responds to an increase in government purchases. For example, some authors assume that average productivity of firms depends on the level of aggregate economic activity (for example, Baxter and King, 1992 and Farmer, 1993). Others assume that increasing returns to scale occur at the firm level (see Farmer, 1993). These models predict that average labor productivity should rise after an exogenous increase in government purchases. This prediction also emerges in models that allow for labor hoarding and variable capital utilization rates (Burnside, Eichenbaum, and Rebelo, 1993 and Burnside and Eichenbaum, 1996). Standard neoclassical models with constant returns to scale production functions (Aiyagari, Christiano, and Eichenbaum, 1992) predict that average labor productivity should fall. As with real wages, the key question is: Which prediction is correct?
The major difficulty in answering this question is identifying exogenous changes to government purchases. Simply observing what happens to real wages and average labor productivity after government purchases change does not reveal the effects of the changes in government purchases per se. …