Empirical estimates of the short-run aggregate supply and demand curves for the post-war U.S. economy.

AuthorGamber, Edward N.
  1. Introduction

    This paper presents estimates of the short-run aggregate supply and demand curves for the post-war U.S. economy using a structural vector autoregression (SVAR). Following the work of Blanchard and Quah [4] (henceforth BQ) I assume that aggregate demand shocks have no long-run impact on the log of real output. In contrast to BQ I use aggregate output and prices while they use aggregate output and the unemployment rate. By replacing unemployment with prices I can estimate the slopes of the aggregate supply and demand curves for the U.S. economy.

    My purpose in extending the work of BQ in this direction is threefold. First, to investigate whether the decomposition method proposed by BQ yields "textbook" aggregate demand and supply curves; second, to investigate whether the supply and demand shocks derived from this method correspond to historical demand and supply shocks; and third, to test the stability of the slope of the aggregate supply curve.

    The once-well-accepted stylized characterization of prices as procyclical has recently been questioned (Kydland and Prescott [13] and Wolf [23]). Procyclical prices were previously thought to be evidence that the cycle was aggregate demand driven. The studies by Kydland and Prescott and Wolf show that prices are either acyclical or countercyclical thus indicating that aggregate supply shocks may play a role in business cycle fluctuations. In this paper I test whether, once output movements and prices are decomposed into demand and supply driven components, the aggregate demand induced movements in prices are procyclical and the aggregate supply induced movements in prices are countercyclical. I find that the aggregate supply curve does slope upward and the aggregate demand curve does slope downward. Thus, the evidence presented here is consistent with the notion that the lack of a consistent cyclical pattern in the price level is due to the fact that both demand and supply shocks generate business cycle fluctuations.

    Secondly, I extend BQ by investigating whether the movements in output and prices due to aggregate demand and supply shocks correspond to specific episodes of demand and supply shocks during the post-war era as defined by independent sources. For example, does this estimation technique attribute movements in output due to oil price shocks to aggregate supply and the movement in output during the Volker-recession to aggregate demand? For the most part, oil price shocks and tight monetary policy regimes do correspond to aggregate supply and demand shocks respectively.

    The third and final purpose of this extension is to test the stability of the aggregate supply curve. According to Lucas [14] the AS curve will steepen as the variance of absolute price shocks rises relative to the variance of relative price shocks. Over the post-war period the slope of the aggregate supply curve is not stable. I find that the aggregate supply curve was essentially flat in the mid-1960s, steepened throughout the 1970s and then flattened slightly in the 1980s.

    This paper is organized as follows. Section II describes the evolution of the literature on decomposing movements in output into permanent and temporary components. Section III describes the Blanchard-Quah technique. Section IV presents the estimates of the supply and demand curves and the results of the stability tests. Section V concludes the paper.

  2. Evolution of the Literature

    The traditional, pre-1980, method of decomposing output movements into cycle and trend typically assumed that the trend followed a linear or smoothly evolving path. This trend line was assumed to be a function of growth factors such as labor, capital and technology while the remaining cycle was assumed to be a function of aggregate demand.

    The work of Nelson and Plosser [15] showed that most macroeconomic time series contain a stochastic rather than deterministic trend. Their work cast doubt on the simple deterministic trend decomposition of output and suggested that at least part of the quarterly fluctuations in aggregate output are due to aggregate supply factors. Although this finding suggests that part of the quarterly change in aggregate output is determined by permanent aggregate supply factors it by no means suggests that all of the quarterly change is a function of these factors. In fact, the work by Nelson and Plosser gave no indication of the amount of output changes due to permanent and temporary shocks - they simply identified the fact that permanent shocks occur each time period. The question of the proportion of the variance in output growth due to each of the two factors was left to subsequent literature.

    Figure 1 shows a scatter plot of quarterly growth rates of real GDP and the GDP deflator, 1949:2 through 1992:4. One would expect that if quarterly movements were dominated by aggregate demand the points would form a pattern from the southwest to the northeast thus tracing out an aggregate supply curve.(1) If the movements were dominated by aggregate supply one would expect the points to form a pattern from the northwest to the southeast thus forming an aggregate demand curve. Neither pattern emerges from these points which leads one to conclude that either the model is wrong or both demand and supply shocks are important in determining the position of these points.

    The literature on measuring the contribution of aggregate demand and supply can be roughly divided into two types: univariate and multivariate approaches.(2) The univariate approaches typically estimate an ARIMA model for GDP and then assume either that permanent and temporary disturbances are perfectly correlated [2] or orthogonal [22]. The multivariate approaches typically impose long-run restrictions on the impulse responses from a vector autoregression. Examples of this literature include Shapiro and Watson [18] who restrict the aggregate labor supply curve to be vertical in the long-run and BQ who restrict aggregate demand shocks to have no long-run impact on the log of real output. Estimates of the percent of the variation in quarterly GDP growth due to permanent shocks vary widely from a low of less than 1 percent (BQ) to a high of 93 percent (Stock and Watson, [21]).(3)

  3. The Blanchard-Quah Technique

    The Blanchard-Quah decomposition method begins with the estimation of the following moving average representation of output growth ([Delta]y) and the prime age male unemployment rate (u):

    [Mathematical Expression Omitted]

    where the [Epsilon]'s are mean zero innovations with covariance matrix [Omega] (individual elements of this matrix are denoted by [[Omega].sub.ij]).

    This representation is obtained by first estimating the unconstrained vector autoregression and then inverting that representation by subjecting each equation to a one unit shock and tracing the effects of that shock through both equations. The resulting unconstrained impulse responses are represented by the polynomials in the lag operator [c.sub.ij](L). C(L) denotes the entire matrix of polynomials. By construction, C(0), the matrix of contemporaneous responses, is the identity matrix.

    The innovations ([Epsilon]'s) are, in general, contemporaneously correlated. It follows that the impulse responses represented by the [c.sub.ij]'s do not show the responses to independent...

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT