RDP 2012-01: Co-movement in Inflation 5. Implications for Modelling Inflation in Australia
March 2012 – ISSN 1320-7229 (Print), ISSN 1448-5109 (Online)
- Download the Paper 814KB
The significant correlation between inflation rates across countries is something that could be exploited for understanding the dynamics of domestic inflation in Australia. Furthermore, the results presented so far in this paper suggest that movements in global inflation may reflect, in part, information other than what can be gleaned from data on foreign real activity and movements in commodity prices. The significance of global inflation for explaining domestic inflation, however, will depend on a number of factors. For example, while movements in commodity prices are a potential key driver of co-movement in inflation rates across countries, in Australia (a commodity exporter) the floating exchange rate could work to offset somewhat the impact of higher commodity prices on inflation, since a positive shock to commodity prices is likely to be associated with an appreciation of the Australian dollar. Pass-through to domestic prices may be delayed, however, for a number of reasons (such as, foreign exchange hedging, pricing to market), in which case exchange rate fluctuations need not perfectly offset a change in foreign prices. It remains an empirical question, therefore, as to whether or not we may still see significant co-movement in inflation rates in the near term.
To test the significance of foreign inflation in explaining domestic inflation in Australia, models for both headline and underlying (trimmed mean) inflation following Norman and Richards (2010) were estimated. The Phillips curve and mark-up model specifications were augmented with the average of quarterly G7 headline inflation, similar to the measure of global inflation used in the panel VAR (which averages G7 inflation over two periods), as an additional explanatory variable.[16] The same oil and non-fuel commodity price variables used in the panel VAR were also included. As Australia can be considered a small open economy, G7 inflation enters the regressions contemporaneously (and with one lag), as do both commodity price variables.
The results are reported in Tables 2 and 3, over the estimation sample from 1990:Q1 to 2011:Q1. A positive and statistically significant coefficient on the G7 inflation variable would be evidence in favour of foreign inflation being an important explanator of inflation in Australia.
| Dependent variable | Headline inflation | Trimmed mean inflation | |||
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | ||
| Lagged inflation | −0.377** | −0.409* | 0.233 | 0.233 | |
| Inflation expectations | 0.598*** | 0.395** | 0.311*** | 0.188** | |
| Unemployment rate | 0.107** | 0.123*** | 0.102*** | 0.108*** | |
| Change in unemployment rate | −0.003 | −0.003* | −0.002*** | −0.002*** | |
| Import price inflation | 0.119 | 0.104 | 0.035 | 0.032 | |
| Oil price inflation | 0.014*** | 0.007** | 0.004* | 0.002 | |
| Non-fuel commodity price inflation | 0.018** | 0.013 | −0.007* | −0.010* | |
| G7 inflation (t) | 0.562*** | 0.192** | |||
| G7 inflation (t−1) | −0.115 | 0.052 | |||
| Adjusted R2 | 0.354 | 0.401 | 0.712 | 0.723 | |
| Notes: Models (2) and (4) include G7 inflation; sum of coefficients and p-values from Wald tests for joint significance are reported; ***, ** and * indicate significance at the 1, 5 and 10 per cent level, respectively; two lags were included for domestic inflation; import price inflation enters as a polynomial distributed lag; oil and non-fuel commodity price inflation and G7 inflation enter contemporaneously andwith one lag. Both the headline and trimmed mean inflation measures exclude interest, tax and healthpolicy changes and exclude the deposits and loans component of the CPI; see Norman and Richards(2010) for more details. | |||||
| Dependent variable | CPI inflation | Trimmed mean inflation | |||
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | ||
| Lagged inflation | −0.335* | −0.347* | 0.378** | 0.410*** | |
| Inflation expectations | 0.464** | 0.280 | 0.153* | 0.049 | |
| Output gap | 0.112 | 0.130 | 0.099* | 0.116* | |
| Growth in real unit labour costs | 0.182 | 0.196 | 0.123** | 0.104** | |
| Import price inflation | 0.175 | 0.166 | 0.079 | 0.077 | |
| Oil price inflation | 0.015*** | 0.009** | 0.005* | 0.003 | |
| Non-fuel commodity price inflation | 0.016** | 0.011 | −0.005* | −0.008 | |
| G7 inflation (t) | 0.534** | 0.174* | |||
| G7 inflation (t−1) | −0.155 | 0.038 | |||
| Adjusted R2 | 0.333 | 0.374 | 0.697 | 0.706 | |
| Notes: Models (2) and (4) include G7 inflation; sum of coefficients and p-values from Wald tests for joint significance are reported; ***, ** and* indicate significance at the 1, 5 and 10 per cent level, respectively; two lags were included for domestic inflation; growth in real unit labour costs and importprice inflation enter as a polynomial distributed lag; oil and non-fuel commodity price inflation and G7 inflation enter contemporaneously and with one lag. Both the headline and trimmed mean inflationmeasures exclude interest, tax and health policy changes and exclude the deposits and loans component ofthe CPI; see Norman and Richards (2010) for more details. | |||||
The regression results suggest that foreign inflation does have explanatory power for domestic inflation in Australia (in line with the results for the G7 presented earlier), particularly headline inflation. The coefficient on contemporaneous G7 inflation is highly significant in the headline inflation regressions and including G7 inflation improves the in-sample fit of both models considerably (as measured by the adjusted R-squared). Over the 1990:Q1–2011:Q1 sample, the contemporaneous impact of a 1 percentage point increase in G7 headline inflation is to increase domestic headline inflation by around 0.5 percentage points. While also statistically significant (at around the 5 per cent level), foreign inflation seems to offer little additional explanatory power in the trimmed mean inflation regressions. The impact of a 1 percentage point increase in G7 inflation on underlying inflation is also smaller, with trimmed mean inflation increasing by around 0.2 percentage points in the same quarter. The results also find some evidence of commodity prices affecting headline inflation, in particular, in the same quarter.[17] However, the coefficients on oil and non-fuel commodity price inflation are smaller (and become less significant or insignificant) when the G7 inflation variable is included. The role of inflation expectations for explaining domestic inflation also falls somewhat when G7 inflation is included as an additional explanator.
In line with the earlier results, foreign inflation seems to offer more information for explaining headline inflation than underlying inflation. Given headline inflation is more sensitive to fuel and food prices than is trimmed mean inflation, this suggests a key role for commodity prices in driving inflation co-movement. It remains somewhat of a puzzle, however, as to why G7 inflation is still found to be significant in the above regressions. Notably, controlling for commodity prices and import prices (and global activity – although this variable was found to be insignificant and is not included in the above results) should account for explanations emanating from business cycle correlations and pass-through of shocks to traded goods prices. An alternative explanation may relate to the difficulties in measuring variables such as import prices, and in particular the dynamics of their impact on inflation, and so foreign inflation may help to proxy for these impacts.
In the Australian data, it also seems that the bulk of the explanatory power of G7 inflation is coming through the contemporaneous term (when only the first lag of G7 inflation is included its coefficient was also found to be insignificant in each of the regressions). This result seems consistent with the idea that the endogenous response of the exchange rate could be working to offset (with some lag) the effect of an impulse to global inflation. Well-anchored inflation expectations are also likely to have helped limit second-round type effects.
Finally, the result that movements in international inflation have explanatory power for domestic inflation is also an interesting finding from a forecasting perspective. Stock and Watson (2007) discuss the fact that in recent times inflation has become harder to forecast, in the sense that it is difficult to find useful predictors of inflation. Cicarelli and Mojon (2010) consider a forecasting exercise (which includes data for Australia) and find that including information on global inflation can improve forecasting accuracy relative to simple univariate benchmark models. As a simple ‘nowcasting’ exercise, we also included in the above regressions a measure of G7 inflation constructed using less than the full quarter's worth of inflation data. Measures using only the first two months, and first month, of inflation data for the relevant quarter were also generally found to be significant in the above regressions (although the estimated coefficients were somewhat lower), suggesting the monthly frequency of inflation data in the G7 could be used to inform predictions of (quarterly) Australian inflation. However, history would suggest it would be wise to be cautious about the stability of various relationships for predicting inflation. The emergence of Asia, for example, presents a key structural change taking place in the global economy, with potentially important implications for global inflation dynamics.[18]
Footnotes
The results are qualitatively similar if the common inflation indicator from the panel VAR is used instead. The simple average is also very close to the first principle component of G7 inflation rates (as principle component analysis assigns close to equal weight to each of the series). Using an OECD-wide measure was also found to give similar results. [16]
These results were found to be robust to the inclusion of the common real activity indicator (contemporaneously and with lags) from the panel VAR as a proxy for international economic activity. [17]
Neely and Rapach (2008) found that country characteristics such as the degree of openness, institutional quality, financial development and real GDP per capita were all associated with an increasing role for global inflation in explaining domestic inflation. [18]