RDP 2009-05: Macroeconomic Volatility and Terms of Trade Shocks 4. The Impact of Terms of Trade Volatility on Macroeconomic Volatility
October 2009
To understand the effect of terms of trade volatility on macroeconomic volatility, we first estimate Equation (2) without the interactive terms. The results in Table 3 suggest that terms of trade volatility has a statistically significant positive effect on the volatility of output growth and inflation. The point estimates imply that if the volatility of annual terms of trade growth was greater by one standard deviation, the volatility of shocks to annual GDP growth would be 1.1 percentage points greater and the volatility of annual inflation shocks would be 1.2 percentage points greater.^{[13]}
Regression | ||||||
---|---|---|---|---|---|---|
Dependent variable | ||||||
Output volatility | Inflation volatility | |||||
[3.1] | [3.2] | [3.3] | [3.4] | [3.5] | ||
Terms of trade variables | ||||||
σ Terms of trade _{t} | 0.13^{**} | 0.21^{**} | 0.19^{***} | |||
Control variables | ||||||
Openness _{t−1} | −0.05 | 0.00 | −0.29 | −0.21 | −0.25 | |
Credit _{t−1} | 0.13 | 0.12 | −0.07 | −0.07 | ||
Floating exchange rate _{t−1} | 0.12^{*} | 0.13^{*} | −0.15 | −0.14 | −0.16 | |
Strict monetary policy _{t−1} | −0.25^{*} | −0.27^{**} | −0.08 | −0.11 | −0.30 | |
Inflation _{t−1} | 0.24^{**} | 0.24^{*} | ||||
Currency crisis _{t} | 1.51^{***} | 1.47^{***} | 1.56^{***} | |||
Number of countries/observations | 71/411 | 71/411 | 71/402 | 71/402 | 71/402 | |
R^{2} within | 0.21 | 0.23 | 0.50 | 0.51 | 0.47 | |
Notes: ***, **, and * indicate that coefficients are significant at the 1, 5 and 10 per cent levels, respectively, using robust standard errors. All regressions include country- and time-fixed effects. |
The estimated effects of the control variables on output volatility generally accord with our prior expectations. In particular, adopting a strict monetary policy regime reduces the volatility of shocks to annual output growth in the next five years by around 0.24 percentage points. This finding is consistent with Kent et al (2005), and demonstrates the stabilising role that credible monetary policy plays in general.
As in Easterly, Islam and Stiglitz (2000), the point estimates in our regressions suggest that (other things equal) floating exchange rates are associated with higher output volatility, although this effect is not statistically significant. More trade openness is associated with less output volatility while more developed financial institutions are positively associated with output volatility, although neither effect is statistically significant. The insignificance of the credit term is not unexpected given that the theoretical relationship between financial market development and output volatility is ambiguous, and is contingent on the nature of the shocks (Beck et al 2006).
In the inflation volatility equations, we find that economies with higher rates of inflation also tend to experience more inflation volatility, a standard result in the literature. The coefficient on the currency crisis dummy is also positive and significant, illustrating the disruptive effect of large exchange rate depreciations (often associated with the abandonment of a fixed exchange rate regime) on domestic prices. The other control variables have negative coefficients, but these are not statistically significant. In the case of strict monetary policy, this result is surprising in light of existing evidence that inflation targeting reduces the volatility of inflation (Calderon and Schmidt-Hebbel 2009, for example). While this result could partly reflect the relative crudeness of our measure of strict monetary policy, the problem of multi-collinearity could also be a factor, given the relatively strong correlation between strict monetary policy and the credit to GDP variable (the correlation coefficient is 0.60; see Table C1). Moreover, to the extent that strict monetary policy stabilises inflation volatility by reducing the average rate of inflation, the lagged CPI inflation term – which is highly significant in Regression [3.4] – could well be a proxy for the impact of strict monetary policy. Indeed, when we exclude the credit and lagged inflation terms (Regression [3.5]), the size of the strict monetary policy coefficient almost triples and becomes marginally significant (with a p-value of 0.11), with only a minor reduction in explanatory power overall (the R^{2} within falls from 0.51 to 0.47).^{[14]}
Having determined that variation in terms of trade shocks matters for output and inflation volatility, we now ask how policies affect stability in the face of such terms of trade volatility. To do this, we interact our measure of terms of trade volatility with the structural indicators described in Section 3.
Table 4 shows how our various structural variables affect the relationship between terms of trade volatility and output volatility. The main result is that adopting a floating exchange rate regime helps to stabilise an economy subject to a more volatile terms of trade (Regression [4.2]). The estimates imply that given a one standard deviation increase in the volatility of terms of trade shocks, other things equal, annual output volatility will be around 0.15 percentage points lower in economies with floating exchange rates than in economies with fixed exchange rate regimes.^{[15]} Indeed, given that the non-interacted floating exchange rate coefficient is positive and significant in this regression, it appears that offsetting terms of trade shocks are the main way in which a floating exchange rate helps to stabilise output volatility.
Regression | |||||
---|---|---|---|---|---|
[3.2] | [4.1] | [4.2] | [4.3] | [4.4] | |
Terms of trade variables | |||||
σ Terms of trade _{t} | 0.13^{**} | 0.39^{*} | 0.17^{***} | 0.15^{***} | 0.37 |
σ Terms of trade _{t} ^{*} Credit _{t−1} | −0.07 | −0.05 | |||
σ Terms of trade _{t} ^{*} Floating exchange rate _{t−1} | −0.15^{**} | −0.14^{*} | |||
σ Terms of trade _{t} ^{*} Strict monetary policy _{t−1} | −0.17 | −0.06 | |||
Control variables | |||||
Openness _{t−1} | 0.00 | 0.02 | −0.03 | 0.01 | −0.01 |
Credit _{t−1} | 0.12 | 0.26^{*} | 0.14 | 0.12 | 0.25 |
Floating exchange rate _{t−1} | 0.13^{*} | 0.14^{**} | 0.41^{**} | 0.13^{*} | 0.39^{***} |
Strict monetary policy _{t−1} | −0.27^{**} | −0.33^{***} | −0.31^{**} | −0.17 | −0.31^{*} |
Wald tests (p-values) | |||||
H0: terms of trade coefficients (jointly) = 0 | 0.02 | 0.01 | 0.03 | 0.02 | |
H0: institutional coefficients (jointly) = 0 | 0.24 | 0.02 | 0.02 | 0.00 | |
H0: institution interaction coefficients (jointly) = 0 | 0.10 | ||||
Number of countries/observations | 71/411 | 71/411 | 71/411 | 71/411 | 71/411 |
R^{2} within | 0.23 | 0.23 | 0.24 | 0.23 | 0.24 |
Notes: ***, **, and * indicate that coefficients are significant at the 1, 5 and 10 per cent levels, respectively, using robust standard errors. All regressions include country- and time-fixed effects. |
The coefficients on both the strict monetary policy and financial market development interaction terms are also negative, although insignificant. The strict monetary policy interaction term, however, is jointly significant and negative when considered with the non-interaction strict monetary policy term. The credit result is broadly consistent with Beck et al (2006), who find only weak evidence for the idea that greater financial development dampens the impact of terms of trade volatility on output volatility. In all equations, the coefficient on the terms of trade volatility variable is larger than in the model with no interaction terms. The results for the regression with all of the interactive terms included together are broadly similar to the regressions with each interactive term by itself. However, a Wald test for the significance of all of the institutional terms (both interacted and non-interacted in Regression [4.4]) suggests that these variables jointly have a significant moderating influence on the effect of terms of trade volatility on output volatility. Overall, we interpret these results as providing evidence that terms of trade shocks can increase the volatility of output, but that institutional settings can help to diminish the impact of these shocks.
Table 5 considers the impact of labour market flexibility on output volatility. Overall, the results are fairly weak. In the specification that employs the Economic Freedom of the World Index, the interaction terms suggest that high labour market flexibility tempers the impact of terms of trade shocks on output volatility, though this effect is not statistically significant. When we use union density to proxy labour market flexibility, the interaction term is also statistically insignificant, though the effect goes the other way (that is, higher union density is associated with lower output volatility in the presence of terms of trade shocks). The only statistically significant coefficient of interest is the (non-interacted) union density term, which implies that more regulated labour markets – as proxied by higher union density – tend to raise output volatility. It is important to note, however, that this effect jointly captures the extent to which labour market flexibility affects the responsiveness of the economy to all other shocks (unrelated to the terms of trade), as well as the size of these shocks. This result is broadly consistent with Kent et al (2005), although they use a different measure of labour market flexibility – the number of days lost to labour disputes.
Regression | ||
---|---|---|
[5.1] | [5.2] | |
Terms of trade variables | ||
σ Terms of trade _{t} | 0.15^{*} | 0.42 |
σ Terms of trade _{t} ^{*} Medium labour market flexibility _{t−1} | −0.01 | |
σ Terms of trade _{t} ^{*} High labour market flexibility _{t−1} | −0.09 | |
σ Terms of trade _{t} ^{*} Union density _{t−1} | −0.11 | |
Control variables | ||
Openness _{t−1} | −0.01 | −0.34 |
Credit _{t−1} | 0.00 | 0.31^{**} |
Floating exchange rate _{t−1} | 0.15^{**} | 0.16^{*} |
Strict monetary policy _{t−1} | −0.30^{**} | −0.34^{**} |
Medium labour market flexibility _{t−1} | 0.02 | |
High labour market flexibility _{t−1} | 0.19 | |
Union density _{t−1} | 0.40^{*} | |
Number of countries/observations | 71/394 | 56/251 |
R^{2} within | 0.23 | 0.31 |
Notes: ***, **, and * indicate that coefficients are significant at the 1, 5 and 10 per cent levels, respectively, using robust standard errors. All regressions include country- and time-fixed effects. |
Table 6 presents the results when we interact our structural variables with terms of trade volatility in regressions whose dependent variable is inflation volatility. Once again, we find that adopting a flexible exchange rate helps to moderate the effect of terms of trade volatility. The estimates imply that given a one standard deviation increase in the volatility of terms of trade shocks, other things equal, annual inflation volatility will be around 0.22 percentage points lower in economies with floating exchange rates than fixed exchange rate regimes (see Regression [6.2]). Coefficients on interactions between terms of trade volatility and private credit, and terms of trade volatility and strict monetary policy both produced positive – though insignificant – coefficients.^{[16]} To further investigate the role of monetary policy regimes and to abstract from the multi-collinearity concerns raised above, Regression [6.5] excludes the credit and lagged inflation terms. While the coefficient on the interaction between terms of trade volatility and strict monetary policy remains insignificant, the strict monetary policy term by itself is negative and significant at the 5 per cent level. This suggests that monetary policy regimes that have become relatively more strict on inflation have played a role in reducing the volatility of inflation, as well as output.
Regression | ||||||
---|---|---|---|---|---|---|
[3.4] | [6.1] | [6.2] | [6.3] | [6.4] | [6.5] | |
Terms of trade variables | ||||||
σ Terms of trade _{t} | 0.21^{***} | 0.20 | 0.26^{***} | 0.21^{***} | 0.23 | 0.23^{***} |
σ Terms of trade _{t} ^{*} Credit _{t−1} | 0.01 | 0.01 | ||||
σ Terms of trade _{t} ^{*} Floating exchange rate _{t−1} | −0.21^{**} | −0.22^{**} | −0.21^{*} | |||
σ Terms of trade _{t} ^{*} Strict monetary policy _{t−1} | 0.08 | 0.11 | 0.18 | |||
Control variables | ||||||
Openness _{t−1} | −0.21 | −0.21 | −0.26 | −0.22 | −0.27 | −0.29 |
Credit _{t−1} | −0.07 | −0.08 | −0.04 | −0.07 | −0.06 | |
Floating exchange rate _{t−1} | −0.14 | −0.14 | 0.24 | −0.14 | 0.25 | 0.21 |
Strict monetary policy _{t−1} | −0.11 | −0.11 | −0.16 | −0.17 | −0.24 | −0.48^{**} |
Inflation _{t−1} | 0.24^{**} | 0.25^{***} | 0.25^{***} | 0.24^{**} | 0.25^{**} | |
Currency crisis t | 1.47^{***} | 1.47^{***} | 1.51^{***} | 1.47^{***} | 1.52^{***} | 1.61^{***} |
Wald tests (p-values) | ||||||
H0: terms of trade coefficients (jointly) = 0 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | |
H0: institutional coefficients (jointly) = 0 | 0.71 | 0.07 | 0.70 | 0.31 | 0.01 | |
H0: institution interaction coefficients (jointly) = 0 | 0.25 | 0.14 | ||||
Number of countries/observations | 71/402 | 71/402 | 71/402 | 71/402 | 71/402 | 71/402 |
R^{2} within | 0.51 | 0.51 | 0.51 | 0.51 | 0.51 | 0.48 |
Notes: ***, **, and * indicate that coefficients are significant at the 1, 5 and 10 per cent levels, respectively, using robust standard errors. All regressions include country- and time-fixed effects. |
Overall then, we interpret these results as suggesting that adopting a floating exchange rate regime reduces the influence of terms of trade volatility on macroeconomic volatility. While the results for the other institutional variables are less robust, and depend somewhat on the specification, the point estimates suggest a more obvious role in moderating output volatility, rather than inflation. To obtain a better understanding of how these output effects operate, we adopt a disaggregated approach, and estimate how terms of trade shocks and economic institutions affect the volatility of the various expenditure components of GDP.
Footnotes
A one standard deviation increase in the volatility of annual terms of trade shocks is equivalent to 0.90 log points (based on the result shown in Table 2), while the coefficient on the terms of trade term is 0.13 in the output volatility regression and 0.21 in the inflation volatility regression (Table 3). Given that these regressions are estimated in logarithmic form, a one standard deviation increase in terms of trade volatility increases output volatility and inflation volatility by e^{0.9x0.13}=1.1 percentage points and e^{0.9x0.21}=1.2 percentage points, respectively. [13]
Note that the strict monetary policy coefficient becomes significant at the 5 per cent level when we include interaction terms (see Regression [6.5]). [14]
In Regression [4.2], the coefficient on σTerms of trade _{t} is 0.17 while the coefficient on σTerms of trade _{t}* Floating exchange rate _{t−1} is −0.15. Accordingly, a one standard deviation increase in terms of trade volatility raises output volatility by e^{0.9x0.17}=1.17 percentage points in a fixed exchange rate regime, and by e^{(0.9*0.17)+(0.9*−0.15)}=1.02 percentage points in a floating exchange rate regime. Therefore, given a one standard deviation increase in the volatility of the terms of trade, annual output volatility is 0.15 percentage points lower in a floating exchange rate regime, compared with a fixed regime. In conducting this thought experiment, we abstract from the non-interacted floating exchange rate coefficient in Regression [4.2] to the extent that it captures how floating exchange rate regimes condition the responsiveness of the economy to all other (non terms of trade-related) shocks. [15]
We also estimated models containing the various measures of labour market flexibility from Table 5, though none of these variables turned out to be significant explanators of inflation volatility. [16]