عدم قطعیت اقتصاد کلان، تورم و رشد : اثرات وابسته به رژیم در G7

We analyze the causal effects of real and nominal macroeconomic uncertainty on inflation and output growth by considering whether these effects are cycle phase specific. Employing a bivariate Smooth Transition EGARCH-M model for the G7 countries during 1957–2009, we find strong nonlinearities. First, uncertainty regarding the output growth rate is related with a higher average growth rate mostly in a low-growth regime, supporting the theory of “creative destruction”. Second, higher inflation uncertainty diminishes growth rates, mainly at a high-inflation regime. Finally, real uncertainty has mixed effects on average inflation, while the effect of nominal uncertainty is typically positive, especially so during inflationary periods. Our findings suggest that these relationships are sufficiently complex to require treatment with nonlinear models.

Considerable research has been devoted in recent years in examining the effects of macroeconomic (nominal and real) uncertainty on the rates of inflation and output growth. The literature offers many competing theories as to the sign of these effects, while empirical support for these theories is just as varied, offering an inconclusive outcome. For this reason, a few studies have attempted to investigate whether these effects depend on the rates of inflation and output growth, thus stressing the importance of the state of the economy.1 Amongst them, Baillie et al. (1996) and Chang and He (2010) examine the effect of inflation uncertainty on inflation and output growth, respectively, as a function of the rate of inflation, while Henry and Olekalns (2002) focus on the effect of recessions on the relationship between output variability and growth. The current study combines the above considerations and examines the presence of regime-dependent effects by jointly considering the impact of real and nominal uncertainty on inflation and output growth outcomes within a simultaneous estimation model. An advantage of the model is that the regime-switching thresholds of inflation and output growth are determined endogenously, thus avoiding issues of potential misspecification bias. The theoretical literature that examines the effects of macroeconomic uncertainty on inflation and growth is rich in offering competing explanations.2 In particular, there are theories that support an effect of inflation uncertainty on output growth that takes up a negative sign (Friedman, 1977, Pindyck, 1991 and Huizinga, 1993) while some others support a positive sign (Abel, 1983 and Dotsey and Sarte, 2000).3 Similarly, the influence of nominal uncertainty on the rates of inflation has been deemed to be either positive, by Cukierman and Meltzer (1986) and Cukierman (1992), or negative by Holland (1995). In a similar fashion, output growth uncertainty has had supporters of a positive (Deveraux, 1989) but also of a negative (Taylor effect jointly with the Cukierman–Meltzer hypothesis) influence on inflation. Finally, the impact of real uncertainty on output growth has been theorized to go in all possible directions covering positive (Black, 1987), zero (Friedman, 1968), and negative (Bernanke, 1983 and Pindyck, 1991) effects. All these theoretical contributions assume that the influence of macroeconomic uncertainty on the rates of inflation and growth is linear, or constant. An implication of this is that when the theories are brought into the test, between each pair of suggestive explanations, one theory should prevail as being correct. One, however, cannot dismiss the possibility that the impact of uncertainty on inflation and growth may vary so that all theories are relevant to an extent. This possibility is corroborated by the mixed and often contradictory findings of the empirical literature. It can also be supported by a few recent studies that examine the conditional, or regime-dependent, impact of real and nominal uncertainties. There exists a sizable literature investigating the above relationships with the use of time series Generalized Autoregressive Conditional Heteroskedasticity (GARCH) techniques.4 The findings of the studies that utilize bivariate GARCH-M models, in particular, are found to vary considerably. For example, in the case of the USA which is the most studied of the countries, three of the most well-known studies in the literature agree only on the negative impact of inflation uncertainty on output growth (Grier and Perry, 2000, Grier et al., 2004 and Bredin and Fountas, 2005). The significance of the remaining three estimates varies across studies. This situation, however, is not limited to the United States. For Japan, Bredin and Fountas (2005) and Wilson (2006) agree only on the sign and significance of the impact of real uncertainty on inflation, while the same is also true for the United Kingdom between the studies of Bredin and Fountas (2005) and Bredin et al. (2009). Between these last two studies, one can also observe differences as to the impact of macroeconomic uncertainty for France and Germany. One, therefore, can conclude that the empirical evidence is mixed, even for studies that use similar estimation techniques.5 More recently, some studies have considered the presence of nonlinearities, or regime shifts in the examination of these effects. Henry and Olekalns (2002) examine the nonlinear impact of real volatility on economic growth and provide evidence that recessions result in increased output uncertainty, which dampens subsequent growth. As the economy expands, the impact of real uncertainty on growth vanishes. Other studies consider the nonlinear impact of nominal uncertainty. Baillie et al. (1996) show that higher uncertainty about the rate of inflation leads to higher inflation rates only in high-inflation countries. Chen et al. (2008), on the other hand, in an examination of the four little dragons, show that inflation uncertainty induces a “Laffer curve” effect on inflation in the case of Taiwan. Finally, Chang and He (2010) employ a bivariate Markov regime switching model for the US economy and demonstrate that nominal uncertainty inhibits growth in both low-and high-inflation regimes. The size of the effect, however, is greater (by threefold) in the high-inflation regime. Importantly, when there is no distinction between regimes, the use of a single-regime GARCH model shows the effect of inflation uncertainty to be negligible, pointing to a misspecification bias.6 These GARCH-related studies imply non-monotonic effects of real uncertainty on output growth and of nominal uncertainty on both output growth and inflation. The findings of this set of papers is the starting point of the current paper, where we consider the impact of nominal and real uncertainty as being dependent on the rates of inflation and output growth respectively. Our choice to anchor parameter instability on inflation and growth regime switches is in line with the literature. As described above, the use of the rate of inflation as the measure by which inflation uncertainty influences inflation and growth outcomes is adopted by Baillie et al. (1996) and Chang and He (2010). The use of the growth rate of output as the transition variable in estimating the impact of output growth uncertainty on inflation and output growth, in turn, draws from a long literature that uses this variable to distinguish between periods of economic expansions and contractions (see, among others, Teräsvirta and Anderson, 1992, Teräsvirta, 1994 and van Dijk et al., 2002). These considerations may offer a compromise between the conflicting views of the existing theoretical hypotheses. Keeping with the above reasoning, our methodology employs an asymmetric bivariate Smooth Transition VAR EGARCH-M model that is used to generate the conditional variances of inflation and output growth as proxies for inflation and output growth uncertainty. These measures are then used to test for the effect of real and nominal uncertainty on inflation and output growth. It is important to note that our model also accounts for asymmetric and spillover effects in the conditional variances to avoid a potential source of misspecification. This permits the measure of macroeconomic uncertainty to respond to the direction of change in inflation and output growth (see, for example, Wilson, 2006). The main innovation in our technique is that we generalize the conditional mean and the conditional variance to allow for the possibility that they are affected by the state of the economy as captured by the rate of inflation and the business cycle. An advantage of this specification is that the regime is recognized at each point in time by using the smoothed transition process without exogenously assuming structural change points and threshold values. In this way, our technique allows us to obtain more reliable and accurate estimates of macroeconomic uncertainty. Thus, our model can be viewed as connecting the two strands of research on macroeconomic uncertainty and regime-switching. Our findings offer strong support to the presence of threshold, or regime-dependent, effects for most of the hypotheses tested. First, there is evidence that output growth uncertainty enhances mean output growth. The effect, however, is mainly restricted in the low-growth regime during economic contractions. Second, higher inflation uncertainty diminishes mean output growth during periods of high inflation. Third, output growth uncertainty has a mixed effect on mean inflation across the G7. Most countries, however, experience a negative effect in one of the growth regimes. Fourth, with the exception of France, the effect of inflation uncertainty on mean inflation is positive for one of the inflation regimes. At the same time, there is a small trend for an amplifying impact as economies move to the high-inflation regime. These results offer support to the use of non-linear models in the examination of the macroeconomic impact of real and nominal uncertainties. They also offer a new perspective to the interpretations of the implications of the various competing theories. The paper is organized as follows. Section 2 presents the dataset while Section 3 outlines our econometric model. In Section 4 we report the results of the single-regime model and of the regime-dependent model. These results are then discussed in the context of some recent studies. Finally, Section 5 summarizes our conclusions.

In this paper, we develop an asymmetric bivariate Smooth Transition VAR EGARCH-M system of inflation and industrial production growth in the G7 countries. Our goal is to investigate the presence of nonlinearities in the effects of real and nominal uncertainty on average inflation and output growth. By making the effects conditional on the rate of inflation and the economy’s position along the business cycle, we find significant regime-switching effects. We can summarize our findings as follows. First, there is evidence that output growth uncertainty enhances mean output growth with the effect being mainly restricted in the low-growth regime during economic downturns. Second, higher inflation uncertainty diminishes output growth especially during high inflation periods. Third, output growth uncertainty has mixed effects on mean inflation with most economies experiencing a negative effect in one of the two growth regimes. Finally, there is a positive effect of inflation uncertainty on mean inflation for at least one of the regimes that mainly increases in size in inflationary periods. Our results are important as they demonstrate the existence of significant regime-dependent effects of real and nominal uncertainty on the rates of inflation and output growth. The introduction of switching regimes could explain why competing hypotheses about macroeconomic uncertainty provide varying degrees of support across countries. This would suggest that the empirical relationship between macroeconomic uncertainty and macroeconomic performance is quite complex and, thus, needs to account for nonlinearities. But, of course, the variation in the results across different (mainly industrialized) countries could also be due to country-specific characteristics. This would represent a valuable future research avenue.