# Econometric Forecasting Models

Econometric model building holds considerable promise as a method of forecasting demand. The best starting point towards an understanding of the basis of econometric forecasting is regression analysis. But the difficulty with regression analysis is that it is used to forecast a single dependent variable based on the value and the relations between one or more independent variables and each of these independent variables is assumed to be exogenous or outside the influence of the dependent variable. This may be true in many situations. But unfortunately, in most broad economic situations an assumption that each of the variable, is independent is unrealistic.

For example, let us assume that demand is a function of Gross National Product (GNP), price and advertising. In regression terms we would assume that all three independent variables are exogenous to the system and hence are not influenced by the level of demand itself or by one another. This is fairly correct assumption so far as GNP is concerned. If, however, we consider price and advertising, the same assumption may not hold good, for instance, of the per unit cost is of some quadratic form, a different level of cost. Again, advertising expenditures will often influence the price of the product, since the production and selling cost influence the per-unit price. The price, in turn, is influenced by the magnitude of demand, which can also influenced by the magnitude of demand, which can also influence the level of advertising or promotional expenditure. All of these point to the independence of all four of the valuables in our equation. When this independence is strong, regression analysis cannot be used. If we want to be accurate, we must express this demand relationship by developing a system of four simultaneous equation that to be accurate, we must express this demand relationship by developing a system of four independent equations.

Thus is econometric form we can have;

1. Demand = f (GNP, price and advertising)
2. Cost = f (production and inventory levels)
3. Selling expenses = f (advertising and other selling expenses)
4. Price = f (cost and selling expenses)

That is, instead of one relationship, we now have four. As in regression analysis, we must (a) determine the functional form of each of the equations (b) estimate in a simultaneous manner the value of their parameters and (c) test for the statistical significance of the results and validity of the assumption. It should be realized that the advantages of econometric forecasting is that is provides the values of several of the independent variables from within the model itself, thus freeing he forecaster from having to estimate them exogenously.

The estimation of the equation parameters involves problems for more complex than those encountered in regression analysis. This makes the application of econometric forecasting difficult and expensive.

An econometric model includes a number of simultaneous equation that can be of different types and functional forms. The translation of econometric theory to the right type of form of equation and their development into a set of functional relationship is termed as specification. The accurate and most appropriate specification of an econometric model is a key step in use of this technique of forecasting.

A major part of specification is the identification of the exogenous and endogenous variables, one must arbitrarily decide on the degree of influence of the different factors and choose those that are least determined within the system as exogenous factors. This is kind to the distinction made in regression analysis between the independent variables and the dependent variables. in an econometric model we will want to separate those factors that are most strongly influenced by one another into the endogenous group and those than can be assumed to be determined outside the system of simultaneous equations into the exogenous group.

Once the choice of endogenous and exogenous variables has been made, on equation must be specified for each of the endogenous variables. When the number of the equations specified is equal to the number of exogenous variables, the model is said to be just specified. When the number of equations, the model is under specified and one or more of the variables has to be set arbitrarily to some initial value. these variables then become exogenous variables is greater than the number of equations, the model that is most often used for estimating the parameters of a set of simultaneously equations.

Econometric forecasting models are used most widely to forecast macro series of inter related economic data such as income, consumption and capital spending and much less for business forecasts.

The great advantage of econometric forecasting models are indirect. It can be used to predict the direction and extent of change of the overall economic activity or any its components. This information can then become the input required to estimate the independent variables of a single equation forecasting model. Since this information can be obtained from outside sources, organizations do not have to develop their own models but can rely on outsiders to provide them with forecasts when they are required. Thus individual companies can forego all the high costs associated with developing maintaining, and running a large scale econometric model and obtain he information it offers through third parties.