Can The IMF Commodity Price Index Forecast US Inflation?

Economists agree on the relevant role of monetary policy in the process of maintaining sustained economic growth. That same growth is constrained and weakened by high inflation due to changes in the investment decision of the related economic agents. Price stability and support for economic growth are therefore inevitably erected as the priorities of monetary authorities. In terms of inflation, central banks use a common measure known as “inflation-targeting policy” in order to maintain the inflation rate a 2%. (Bernanke and Woodford, 2005). In this framework, an intermediate objective of central banks is the use of forecasting future pressures on the price level in a short-term horizon in order to adjust the instruments of monetary policy. This way, future inflationary trends effects can be mitigated or taken better into account. For this purpose, economists are interested in commodity prices since they are known to be leading indicators for inflation, also because they have real effects on the economy. Indeed, rising prices of energy commodities such as crude oil petroleum, gasoline and natural gas are considered to be disturbing due to their large share in household’s budgets. (Edelstein, 2007) However, although some economists confer on the commodities predictive capacity of inflation, the empirical reality remains divided on the subject. Indeed, according to the literature a debate started in the 1980s (Blomberg and Harris, 1995) when researches revealed that commodity prices were leading predictors of inflation during the 1970s and mid 1980s. But since the mid 1980s, commodity prices predictive ability decreased and after that period, they were no longer considered as a leading indicator for inflation. (Furlong and Ingenito, 1996) Nonetheless, even though this relation seemed to be somewhat out-dated, the debate recently revived again when commodity prices such as energy, oil and food increased and reached all time highs mid-2008, which raised up inflation rates above central banks targets around the world (Bermingham, 2008). Reversely, due to the recession, commodity prices also decreased rapidly during the 2008 subprime crisis. The link between commodity prices and inflation seems highly established, many authors in the literature have tried through research to estimate this link in order to predict inflation using various kinds of forecasting models. According to the empirical literature, forecasting inflation has been for a long time held in a linear framework. The specifications commonly adopted are the univariate autoregressive (AR) and the multivariate autoregressive (VAR) models. Therefore, the research hypothesis raised in this thesis is that a linear model such as a vector autoregressive (VAR) can assess the forecasting ability of commodity prices to forecast inflation. In the past, several empirical results have demonstrated this hypothesis. Indeed, Blomberg and Harris, (1995), Acharya (2010) and other authors have assessed the predictive ability of commodity prices to forecast inflation through a linear methodology and they found interesting and encouraging results for their commodity price based model performance. The inflation forecasting model that we propose in order to estimate this relationship is the vector autoregressive (VAR) model. We made the choice for this model thanks to its known ability to make good inflation forecasts and because it integrates macroeconomic variables and variables based on financial assets prices. However, despite their theoretical appeal, vector autoregression models are, like all models, the subjects of criticism. Several authors believe that this type of model makes bad forecasts of inflation and it suffers from a lack of stability. The aim of this thesis is to investigate whether it is possible to improve forecasts results of the autoregression model (VAR) when integrating commodity prices indicators. The reason behind this approach provides from the empirical literature that demonstrate that the link between these variables can be captured by linear models. Similarly, the theoretical literature provides a set of corroborating reasons that explain the close relationship between commodities and inflation. The remainder of this thesis is structured in four chapters. The first chapter, we will conduct a brief review of the literature, in which we discuss the economic and empirical theories that justify the linear relation between commodity prices and inflation. This chapter also focuses on the various models for forecasting inflation. Then, the second chapter will present the methodology for forecasting inflation, where we will detail the creation of our economic model, which will incorporate relevant macroeconomic information such as commodity price index and consumer price inflation. Then chapter three will present the transformation process of the data. Finally, the fourth chapter will present the forecast results of our VAR model. Thereby, our study’s main objective was to provide comprehensive analysis of the predictive content of commodity price index through a VAR model in order to improve US inflation forecasts from those of an Autoregressive model during a horizon of 8 months (April to December 2013). The analysis revealed that the outputs obtained are inline with previous authors who used the same linear model in the literature (Achayra, 2010; Blomberg and Harris, 1995). Indeed, our commodity-based model was slightly outperformed by the autoregressive model forecasts. However, it did outperform the benchmark on two forecast periods. Thus we can qualify these results as mixed. Nevertheless, it is important to point out that our work did involve limitations that may have affected our results, and thus also the sharpness of our interpretations. One of these limitations would be the selection of commodity index. In fact, considering the whole commodity index for predicting US inflation can lead to inaccurate information since US consumer price index is balanced differently. Moreover, the selection of the forecast period could also be a limitation because of its short forecasts horizon as consumer price index always reacts with a lag to variation in commodity prices. Thus, choosing a longer forecasts horizon could capture better the impact of commodity prices on inflation. Having detected these limitations, it would be interesting to use them in a future investigation. We could therefore bounce back these limitations as an extension of this thesis, especially by using other alternatives models presented in chapter one. Going beyond these limitations would enable us to expand the reflection on the relation between commodity prices and inflation.