Academic journal article Economic Review (Kansas City, MO)

Is the Great Moderation over? an Empirical Analysis

Academic journal article Economic Review (Kansas City, MO)

Is the Great Moderation over? an Empirical Analysis

Article excerpt

I. EVIDENCE OF INCREASED VOLATILITY

Many studies have now documented the Great Moderation (e.g., Blanchard and Simon, Kim and Nelson, McConnell and Perez-Quiros, and Stock and Watson 2002, 2003). During the Great Moderation, growth rates of real GDP were sharply less variable than in the 1960s and 1970s (Chart 1). (1) From 1960 through the early 1980s, the United States experienced quarterly GDP growth rates as high as 15 percent and as low as -8 percent. Starting in 1984, the variability of GDP growth was much lower, as swings in growth became much more muted (Kim and Nelson, McConnell and Perez-Quiros, and Stock and Watson 2002, 2003).

More recently, though, the recession has produced sharp declines in GDP growth reminiscent of the 1960s and 1970s (Chart 1). Growth plummeted to roughly -6 percent in each of 2008:Q4 and 2009:Q1. These changes in GDP growth suggest a rise in volatility back toward the level that prevailed prior to the Great Moderation.

[GRAPHIC 1 OMITTED]

While Chart I suggests an increase in volatility, it does not quantify the magnitude of the change, or the precise timing. For a more formal assessment, this section follows Stock and Watson (2002, 2003) in estimating statistical models that allow volatility to vary over time. (2) The analysis covers GDP growth and a wide range of other macroeconomic indicators for the United States. These estimates reveal a partial or complete reversal of the Great Moderation in many sectors of the U.S. economy. Appendix 1 provides similar evidence for GDP growth in the other Group of Seven (G7) economies: Recent events have caused estimates of volatility to move significantly higher in all of the nations, reversing much of the Great Moderation.

The statistical model

For each variable of interest (generically denoted y), the statistical model used to assess changes in volatility relates the current value of the variable to past values and a shock that captures unexplained, sudden movements in the variable:

[y.sub.t] = [[beta].sub.0,t] + [[beta].sub.1,t] [y.sub.t - 1] + [[beta].sub.2,t] [y.sub.t - 2] + [[beta].sub.3,t] [y.sub.t - 3] + [[beta].sub.4,t] [y.sub.t - 4] + [e.sub.t].

The model captures the dependence of the volatility (standard deviation) of each variable y, on the coefficients on past values of y, and the standard deviation of the error term.

More specifically, the model recognizes that shifts in either coefficients or the standard deviation of the error term could change the standard deviation of the variable. A rise in the variability of shocks would lead directly to increased volatility of the variable. For example, an increase in the variability of GDP growth could be due to bigger shocks to GDE An increase in the (non-intercept) coefficient values will also generally lead to greater volatility. An upward shift in the coefficients would increase the influence of past values on the current value of the variable, making the variable more sluggish. As a result, following a shock, the variable would more slowly return to baseline. The longer-lived departure from baseline would result in a higher standard deviation of the variable. For instance, an increase in the variability of GDP growth could be due to more drawn-out responses of GDP to shocks, reflected in larger model coefficients.

To capture periodic changes in volatility, the model treats the coefficients on past values of the variable and the variance of the error term as evolving smoothly over time (from quarter to quarter). In turn, the model-implied volatility of the variable can vary from quarter to quarter. In this section, the model-implied volatility is an instantaneous standard deviation. In any quarter, the instantaneous standard deviation of a variable is what the standard deviation would be for all time in the future if the variance of shocks and the coefficients on lagged values of the variable stayed at their current levels. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.