In chapter 8 the simple linear regression model was extended to cover the introduction of two or more explanatory variables, and in chapter 9 the model was given its essential stochastic form.

In the present chapter two further extensions are described. First, §10.1 is devoted to the use of dummy variables. This is a procedure developed to enable us to include in a regression a variable that cannot be measured in the same way as a continuous numerical value (for example, income, or age at marriage) but is instead represented by two or more categories (for example, single, married, or widowed). This special form of a nominal (or categorical) scale is known as a dummy variable.

Secondly, in §10.2 we develop the implications of the idea that it may be appropriate for one or more of the explanatory variables to refer to an earlier period than the one to which the dependent variable relates. Such lagged values recognize the fact that there may be a delay before the changes in the explanatory variable make their full impact.

Dummy variables

Dummy variables with two or more categories

In historical research we often want to take account of factors in a regression that are not measurable in the usual way, but can be expressed as representing one of two (or more) categories. This is a special form of a nominal scale and is called a dummy variable.