Common Adjustments to Multiple Regressions

Abstract

Many variables which constitute causal factors for other variables are of qualitative nature. Their qualitative nature means that they cannot be directly measured by numbers. However, if they influence the behavior of other variables, they must be dealt somehow in process of regression model building. We might imagine hundreds of such qualitative variables; however, examples of the most commonly used in economic analyses are:

• Variables used in marketing research (mostly in cross-sectional models), relating to peoples’ profiles and impacting peoples’ behavior as consumers:

  • Sex (male vs. female).

  • Marital status (married vs. non-married).

  • Having children vs. not having children.

  • Occupation (“white-collar” vs. “blue-collar” vs. other).

  • Education (higher vs. secondary level vs. primary level).

  • Place of living (big city vs. small town vs. rural).

• Variables used in modeling economic and financial results of companies (both in time-series as well as in cross-sectional models):

  • Company’s industry according to its SIC (Standard Industrial Classification) code (belonging to a given industry vs. belonging to other industries).

  • Company’s stock market listing status (listed on stock exchange vs. belonging to private shareholders).

  • Company’s shareholding status (state-owned vs. owned by private shareholders).

  • Occurrence or not of one-off events such as employees’ strike in a period for which economic results of a company are reported.

• Variables used in modeling macroeconomic processes (mostly in time-series models):

  • Occurrence or not of a recession in a period.

  • Occurrence or not of a one-off event impacting inflation rate, such as flood or drought.

  • Seasonal factors (first quarter of a year vs. second quarter vs. third quarter vs. fourth quarter).