Econometric Models

Overview

Econometric models typically refer to a model that consists of a set of equations that relate various economic quantities. Models can run from a single equation, to hundreds of variables and equations. Larger size models do not necessarily perform better. In fact, as a general rule, increasing complexity can harm a model, and not necessarily just from overfitting.

Sample Model

The seciont presents a sample simple econometric model. The model consists of the following

• {% Y %} - national income
• {% C %} - consumer spending
• {% G %} - Government spending
• {% I %} - investment spending
• {% r %} - the interest rate

It consists of the following equations:

This expresses that consumer spending is a fraction of the national income. That is, there is a marginal propensity to consume, which is a number less than one, which describes the fraction of ones income that is typically spent. The parameters in this equation are estimated using a regression. Investments are a fraction of the nations income, and dependent on the interest rate. This equation is also estimated with a regression. The national accounts identity

Model Components

• Identities - equations that are identically true, many times are simply definitions. Example: {% Y = C + I + G %}
• Exogenous Variables - exogenous variables are variables in the model whose value is an observed value that is just plugged into the equations. In the above model, government spending {% G %} is an exogenous variable.
• Structural Equations - equations that express relationships between variables that may only be true on average, and typically need to be estimated with a regression.

Implementation and Forecast

Given the above model, an analyst can use the equations to forecast the effects of various policies. The implementation shows how to script the equations using the simultaneous equations library.

Time Frames

Economic Time Frames are a crucial concern when modeling the economy. That is, the model may assume certain variables, such as prices, are flexible or fixed based on the time frame under consideration. In fact, it may be advisable to build a short term, medium term, and a long term model and then interpolate between the models to construct a full time series model.