James P. Walsh
The rapid rise in food prices since 2003 has faced policymakers with a difficult predicament. In general, the high volatility of food inflation complicates monetary policy decision making by obscuring underlying signals about inflation, but their transient nature limits the long-term impact. At the same time, with prices set globally, food price shocks are often viewed as supply shocks, and thus unlikely to be affected by traditional central bank tools. The broad-brush conclusion is that the role of food prices should be minimized in policymaking: central banks should focus on core measures of inflation that exclude food, both in their assessments of inflation and in monetary policy decisions. Broadly, this is the approach most often supported by the IMF, as discussed in the September 2011 World Economic Outlook (WEO). However, while this will be justified in many cases, in others, as the WEO notes, extenuating circumstances may call for a focus on headline inflation.
Recent IMF research has looked at some of these assumptions for a range of applications. Walsh (2011) notes that core inflation is intended to eliminate statistical noise to focus on underlying trends, either by minimizing the weight of components displaying extreme changes, or components with relatively transitory shocks. Either measure rests on the assumption that headline and core inflation have the same long-run mean (otherwise core understates true inflation) and that non-core inflation has no long-run effect on core inflation. But simply eliminating food prices from headline inflation can violate these assumptions in three important ways:
- Sustained high food inflation. If food prices rise faster than nonfood prices over a long period, then core inflation will underestimate headline inflation.
- Persistent food inflation. If food shocks do not dissipate, they will affect inflation expectations and thus headline inflation.
- Second round effects. If food shocks affect nonfood prices, accommodated food shocks can have an important impact on nonfood inflation.
These conditions can be found in many emerging or developing economies. Looking at a very wide sample of countries, Walsh finds that the difference between long-run average food and nonfood inflation tends to be minimal in advanced economies, but can be sustained and large elsewhere. A non-food core measure can thus show lower inflation than headline, even in the long run.
Second, three different measures of persistence are derived from fitted autoregressive models. In rich countries, with relatively credible central banks, persistence under all three measures is low or even negative as shocks are quickly countered. But in poorer countries, food and nonfood inflation are often persistent; thus, excluding either from a core inflation measure is difficult to justify.
Finally, second-round effects strengthen the case for an earlier monetary policy response to limit pass-through to nonfood inflation. But fitted VARs for food and nonfood inflation show that while second round effects are small and quickly reversed in rich countries, they may not be reversed in poorer countries, and can have a significant impact on nonfood prices.
“In many rich countries, the assumptions required to exclude food inflation from core measures are likely to hold.”
Thus in many rich countries, the assumptions required to exclude food inflation from core measures are likely to hold. But in poorer countries, persistence, high means (likely due to rising incomes and demand), and second-round effects signify that core measures should be developed from first principles of reducing volatility or transience where it might be; mere exclusion of food can lead policymakers to underestimate the impact of price shocks on headline inflation, possibly leading to a weaker policy response.
Looking at optimal monetary policy more broadly, Anand and Prasad (2010) question whether targeting core inflation under imperfect markets yields higher welfare than alternate policies. They note that in many emerging markets and low income countries, not only is the share of food in the CPI very high, but the price elasticity of demand is extremely low, and the income elasticity very high. As in Walsh (2011), they note that both the level and volatility of core and headline inflation also tend to be higher in poorer countries than in richer ones.
To model these differences, they incorporate novel features into a basic dynamic sticky price model: a nontrivial share of credit-constrained consumers who produce food, and a base subsistence level of food consumption. While unconstrained consumers can smooth consumption between periods, credit-constrained consumers must finance consumption out of current wages.
The central bank uses a Taylor rule weighing inflation, the output gap, and a preference for interest-rate smoothing, and the model is evaluated under four regimes: strict core or headline targeting (the central bank values only interest rate smoothing and inflation stabilization) and flexible core or headline targeting (the central bank also stabilizes output).
Under complete markets, targeting strict core inflation maximizes welfare. As inflation rises, the central bank raises interest rates. Consumers save more, reducing aggregate demand and bringing inflation back down. Targeting headline inflation thus results in a higher volatility of output and consumption, analogous to other findings in the inflation targeting literature.
On the other hand, when some households are credit-constrained and cannot smooth consumption, flexible headline inflation targeting maximizes welfare. Higher interest rates in this model lead unconstrained consumers to reduce their aggregate demand as above, but credit constrained consumers cannot respond. Additionally, since their incomes come from food, their consumption may increase when food prices rise. Under strict core targeting, the central bank does not react to food price shocks, and this higher demand aggravates inflation. But under strict headline targeting, the central bank reacts to those higher food prices by raising rates, and the falling consumption by unconstrained consumers outweighs rising demand from constrained consumers. This fact, that inflation and output can move in opposite directions, means that stabilizing output (flexible headline targeting) raises welfare further. Thus when some consumers are credit constrained, as in many developing countries, relative food prices affect not only aggregate supply but also aggregate demand, and central banks can raise welfare by acknowledging this.
Catão and Chang (2010) look at setting monetary policy in small open economies (SOEs). Like Anand and Prasad (2010), they note that food often constitutes a large and relatively inelastic share of the consumption basket, and further note that much of this is imported, so food price shifts can have large terms of trade implications.
They employ a dynamic stochastic general equilibrium (DSGE) model, with some important features. Monopolistic competition and nominal rigidities allow domestic policies to affect the real exchange rate and terms of trade. Traditionally, targeting PPI raises welfare relative to CPI targeting, since food is not modeled differently from other goods, and thus has a high intertemporal elasticity of substitution.1 However, Catão and Chang assume food is imported, priced exogenously, and enters the utility function in a discrete manner.
This changes the welfare ranking. When food price shocks are large and the weight of food in the utility function is high, ignoring food prices, either by targeting a (nominal) exchange rate peg or the PPI, raises the volatility of the real exchange rate and lowers the terms of trade. These, in turn, produce more volatile but on average lower consumption, reducing overall welfare. On the other hand, targeting the CPI takes the effects of external food prices into account: if international food prices rise, the central bank tightens more than in a PPI targeter, leading to a more stable (and more appreciated) real effective exchange rate and more stable consumption path over the long run. Moreover, the welfare result of this strict CPI targeting can be improved under certain parameterizations by also placing some weight in the output gap in the monetary policy function.
Each of these three papers highlights a different way in which failing to recognize the distinctiveness of food in assessing economic conditions can lead to suboptimal outcomes. In countries where food is a small share of the consumption basket, this distinctiveness will likely be unimportant. But in many emerging and developing economies this will not be the case. In such countries, Walsh (2011) shows that looking at core inflation measures that exclude food price inflation can lead to a substantial underestimation of inflationary pressures and mislead central banks on the size of underlying inflationary pressures. Anand and Prasad (2010) show that an environment where many food producers are credit constrained is one with significantly weaker monetary policy transmission, and ignoring food price developments can lead to higher and more volatile inflation. Finally, Catão and Chang (2010) show that when food is imported and not easily substitutable, ignoring food prices in setting monetary policy can reduce welfare by leading to more volatile and reduced consumption.
Anand, Rahul and Eswar S.Prasad,2010. “Optimal Price Indices for Targeting Inflation Under Incomplete Markets,” NBER Working Paper 16290 (Cambridge, Massachusetts: National Bureau of Economic Research).
Catão, Luis A.V. and RobertoChang,2010. “World Food Prices and Monetary Policy,” IMF Working Paper 10/161 (Washington: International Monetary Fund).
International Monetary Fund, World Economic Outlook, September2011, (Washington).
Walsh, James P.,2011. “Reconsidering the Role of Food Prices in Inflation,” IMF Working Paper 11/71 (Washington: International Monetary Fund).
In this context, CPI can be thought of as analogous to headline inflation, while PPI inflation more closely approximates core inflation.