Until now, we have discussed various aspects of inflation and some SAS procedures to evaluate the time series component. However, to deal with the business problem at hand, the modeling team needed to come up with a solution. The two approaches that the modeling team wanted to consider are as follows:

• Using a regression model to understand the interaction of various factors on CPI.
• Forecasting CPI for some periods based on the current CPI time series data. This did not involve understanding the interaction with various variables that can influence CPI.

There were pros and cons to both approaches. Whereas the multivariate regression model would help understand what aspect of consumer spending is statistically significant in predicting CPI, it at times isn't the best method to be used for forecasting. Multivariate regression could in this instance lead to a better understanding of the influencer variables and their statistical significance. However, due to the presence of autocorrelation, it may not be the best prediction method. However, the modelers could still use the resultant regression equation to forecast a few periods ahead. Using the second method of leveraging the CPI variable would mean the use of the Autoregressive Integrated Moving Average (ARIMA) model or some other similar procedure. Autocorrelation could be dealt with in a robust manner. Both the multivariate regression model and a model that leveraged the CPI variable only were built. Let's examine the outcome of the different modeling approaches.