Forecasting Business Financial Performance
This webinar is about the forecasting of business financial performance, that is: forecasting business revenues and forecasting business costs. How should we perform this task? There are four possible approaches to use. First, we can use simple financial projections, which is the approach taken in business planning, and which comes from accounting. Second, we could use an approach that stems from traditional business forecasting, such as ARMA methods (where AR stands for autoregressive and MA stands for moving average) and multiple equation extensions of ARMA’s, or VARs (vector autoregressions). Third, we could use predictive analytics. Fourth, we could use a behavioral approach. The latter three approaches make use of multiple linear regression. The first doesn’t. The approach we take in this webinar is a behavioral approach.
If we are to use multiple linear regression to obtain a forecast, there are two types of variables that we will encounter: the variable(s) to be forecast, termed the predicted variable(s). The variables that are used to make the forecast, termed the predictor variables. The forecast that we make is such that the predicted variable(s) are conditional upon the predictor variable(s).
To forecast anything it is necessary to use causal relationships such that the predictor variables cause the predicted variables. Then when predictor variables change we can forecast the implied change in the predicted variables that come from the causal relationship. However, since we are using multiple regression analysis, we can only be sure that the predictor variables are correlated (or associated) with the predicted variables. The problem is how can we get causation when all we have is correlation? The answer to this conundrum is to model the behavior of business firms (when forecasting costs) and the behavior of consumers (when forecasting revenues). If we model behavior we need to derive the forecasting equations from the behavioral model such that predictor variables cause predicted variables. If these derived forecasting equations can be statistically estimated, then causation becomes correlation and correlation becomes causation. This approach is an alternative to predictive analytics where we only have correlation and not causation. Please see our webinar on predictive analytics.
In this seminar, we first do a thorough treatment of the behavioral approach to forecasting. Next we compare this approach to the use of simple financial projections, which we call forecasting by assumption. The use of simple financial projections carries an air of certainty because they do not incorporate any measures of uncertainty. But, just because measures of uncertainty are not included does not mean that uncertainty doesn’t exist. For example, the assumptions that are used in forecasting by assumption are wrong almost certainly (with 100 percent probability). In every form of forecasting, uncertainty is captured in the confidence interval. The width of the interval captures the uncertainty inherent in the forecast. Using inferential statistics as branch of classical statistics estimating parametric models via multiple linear regression automatically incorporates uncertainty when estimating variances. Forecasting by assumption does not do this, but can be added if one uses Bayesian statistics.
Download a flyer
We are a 501 (c)(3) public charity. All revenues that we receive that are in excess of the costs of producing and marketing our forecasting services are used to subsidize our research function.
The Boston Forecasting Research Institute, Forecasting, Economic Forecasting, Macroeconomic Forecasting, Business Forecasting, Forecasting for Business, Business Planning, Strategic Planning, Business Analytics, Marketing Analytics, Predicitve Analytics, Forecasting Research, Budgeting for Business, Financial Planning for Business, Optimal Budgeting for Business