There isn't a standard way of forecasting inflation in the industry. In the risk-modeling world, the models have to be submitted to a regulatory body to receive the approval for using an advanced modeling approach for credit risk. This has somehow ensured that, across the world, analysts and modelers are using a similar suite of statistical techniques to manage risk. However, in the case of inflation, there isn't much exchange of ideas or a platform available from where we can choose the best modeling technique. Usually, the central bank publishes some inflation figures and these can be used as a benchmark for inflation-forecasting accuracy checks. Most of the models built by organizations are proprietary in nature. Moreover, these organizations are chiefly concerned about benchmarking a model to the central bank inflation index rather than agreeing to and conforming to an industry-level approach to forecasting methodology. In some organizations, modelers prefer to use the past inflation data only to forecast future levels. In other countries, the modelers have preferred to lag some particular variables, as they believe that there is a lag between the driver and inflation growth.

In this chapter, we will introduce new ways to prepare time series data for inflation forecasting. Although the SAS University Edition (the software used for this book) doesn't provide the full suite of SAS ETS procedures in this version of the software, we will try to maximize the use of the available procedures. Let us try and explore the concept of inflation before we try to solve a business problem.

The chapter consists of the following topics:

• Inflation definition and overview
• Reasons for inflation
• Inflation outcomes and the Philips curve
• Leveraging the time ID procedure for data quality checks
• The modeling methodology: the past trend (univariate analysis) and methodology based on predictors (multivariate regression)