RDP 2012-01: Co-movement in Inflation 2. Global Inflation

2.1 Empirical Evidence

A number of papers have found that there is statistically significant co-movement in inflation rates across countries. Ciccarelli and Mojon (2010) use a panel of 22 OECD countries' inflation rates and find that different measures of global inflation (such as a simple average, the official OECD measure, or the first principal component of the data) can explain up to 70 per cent of the movement in domestic CPI inflation rates. There are reasons to think, however, that this result represents somewhat of an upper bound on the contribution of international influences to inflation. Ciccarelli and Mojon use aggregate inflation data in year-ended terms and a long sample from 1961:Q1 to 2008:Q2 that includes large common shocks, such as the oil price shocks in the 1970s, and possible (synchronised) regime changes. Consistent with this, Ciccarelli and Mojon estimate the contribution of international influences to be closer to 30 per cent after de-trending their data to highlight business cycle frequencies.

Monacelli and Sala (2009) use monthly disaggregated (to the product level) CPI data for the United States, France, Germany and the United Kingdom over the sample 1991 to 2004 and find that between 15–30 per cent of the variation in domestic inflation can be attributed to international factors. Since Monacelli and Sala's data are both monthly and disaggregated they will tend to be noisier and so should have lower measured co-movement independent of the underlying relationships in the data. This result might then be thought of as a lower bound to the variance in domestic inflation that can be explained by ‘global inflation’.

Neely and Rapach (2008), using aggregate data and a similar methodology to Kose et al (2003) (who study common movements in real activity), decompose national inflation rates into common, regional and idiosyncratic components. Their paper incorporates a large cross-section of 64 countries, including data from Latin America, Asia, Africa and the Middle East, along with North America and Europe, allowing them to distinguish between common and regional effects. Over the sample 1951 to 2009 they find that, on average across countries, 35 per cent of the variation in domestic inflation rates can be explained by the global factor, with a further 15 per cent explained by regional factors. Neely and Rapach also find that the North American and European regional factors have become more important since the 1980s.

Other work includes Wang and Wen (2007) who find that cross-country quarterly inflation rates (for 18 OECD countries) are highly correlated (correlation of around 0.6 on average) and more so than quarterly output growth (correlation of around 0.2 on average). Mumtaz et al (2011) work with a very long sample, going back as far as the 1800s for some countries, and find that the share of inflation variation due to a common factor has increased post-1985. Work by Hyvonen (2004) documents the convergence of inflation rates using a large sample of IMF member countries and the role played by inflation targeting in driving this result.

2.2 Determinants of Global Inflation

The majority of the papers highlighted above focus on the statistical result that inflation rates tend to move together across countries. There has been less work looking at the possible determinants of this observed co-movement. From an economic theory point of view, it is not necessarily obvious how developments in inflation in other countries might influence domestic inflation. In the long-run, a central bank with independent monetary policy should determine the level of domestic inflation (see, for example, Woodford (2009)). Furthermore, most structural economic models are unable to capture the phenomenon of inflation co-movement (see, for example, Cicarelli and Mojon (2008) and their discussion of the work by Clarida, Galí and Gertler (2002)). Wang and Wen (2007) show that neither a two-country New Keynesian sticky-price model nor a sticky-information model can explain the phenomenon of co-movement in inflation rates unless monetary shocks across countries are themselves correlated. They also document that inflation rates across countries tend to be more highly correlated than output, which is the opposite of what real business cycle theory would predict.

In the short to medium run, however, there are a number of reasons why inflation rates across countries could move together. Increased trade integration (or globalisation more generally) has been highlighted as a key mechanism influencing inflation rates in a number of countries (see Helbling, Jaumotte and Sommer (2006) for work done by the IMF and also Bean (2007)). For example, as east Asian economies have become more integrated into the global trade network, the declining relative price of manufactured goods has been a global phenomenon. Bernanke (2007) links increased trade integration and inflation co-movement via two channels: a direct terms of trade channel that increases or decreases import prices; and a more indirect pro-competitive effect, working to reduce the pricing power of domestic firms and lower mark-ups. Ball (2006) presents an alternate view and argues that globalisation has had no material impact on the dynamics of inflation.

One natural explanation for the correlation in inflation rates is that real activity is also correlated across countries. That is, co-movement in business cycles could lead to co-movement in inflation as domestic inflation responds to correlated changes in domestic demand. The evidence in favour of this explanation, however, is somewhat mixed. Borio and Filardo (2007) have argued that global factors are becoming an important determinant of inflation dynamics. They include a measure of global slack into standard Phillips curve type equations of domestic inflation and find it adds considerable explanatory power. Ihrig et al (2007), however, show that this finding is not robust to alternate specifications of the Phillips curve, nor the measure of global slack. Eickmeier and Moll (2009) estimate factor-augmented Phillips curves for 24 OECD economies and allow global forces to impact inflation indirectly through common movements in domestic demand and supply. They find that the common component to changes in unit labour costs is a significant determinant of inflation but less so the common component to the output gap. Eickmeier and Moll also find that the first principal component explains less of the variation in output gaps across countries than is the case for inflation, suggesting there are other mechanisms driving the co-movement in inflation than simply business cycle correlations.

Common shocks represent another potential driver of the co-movement in inflation. As discussed by Ciccarelli and Mojon (2008), common shocks will be more likely to generate co-movement if they account for a large share of the variance in inflation and inflation rates respond in a similar fashion across countries. Common shocks to commodity prices, therefore, would seem to be one potential explanation for the global inflation phenomenon. Commodity prices are largely determined in global markets, and so large price fluctuations can be experienced in a number of economies at the same time. Also, food and energy prices make up a substantial share of consumer consumption baskets around the world, meaning that movements in these prices could explain a large share of the variation in overall inflation. Movements in global commodity prices could also influence inflation expectations in a number of economies in a similar way, which could in turn affect realised inflation.

If there are substantial structural differences between economies, however, common shocks to commodity prices may not necessarily lead to significant co-movement in inflation rates. For example, a large positive oil price shock is likely to see inflationary pressures increase in a number of countries; directly through higher fuel costs but also indirectly as increased production costs are passed through to consumers. The full extent of the inflationary impulse, however, will depend on a number of other things, including the exchange rate regime (fixed or floating), the degree of competition among firms, domestic policy responses and income effects (which will be of opposite sign for an oil importer versus an oil exporter). Another common commodity shock is to food prices, which is likely to have a greater impact on inflation in developing rather than developed economies, since food makes up a higher share of consumption in the developing world.

Finally, similarities in monetary policy reaction functions (Neely and Rapach 2008; Henriksen, Kydland and Sustek 2009) could also be important in explaining the observed co-movement in inflation rates. For example, with authorities becoming more focused on achieving low inflation outcomes around the early 1990s, the move to a low and more stable inflation environment around that time was a common trend in a number of countries. Also, if central banks respond similarly to common shocks or movements in the global business cycle then this could also induce co-movement as domestic inflation responds to the change in policy. Neely and Rapach (2008) find that the common inflation factor in their model is more important for explaining domestic inflation in countries with a greater degree of central bank independence, which could be indicative of similar reaction functions across countries being important.^[3]

The explanations for the phenomenon of inflation co-movement outlined above are by no means an exhaustive list, nor are they mutually exclusive. Also, different drivers of global inflation might be more important at different times. We now outline a statistical model that can be used to investigate further the issue of global inflation and some of its potential determinants.

Footnote

Neely and Rapach (2008) also note, however, that greater central bank independence could plausibly reduce the estimated importance of global inflation. [3]