Measurement

Measurement Definition

Measurement in a sociological context involves navigating fundamental epistemological challenges concerning how empirical reality is translated into formal, numerical systems. Sociologists must address four primary issues: representation (determining which properties of the empirical world can be effectively modelled), uniqueness (ensuring that resulting measurement numbers are distinct), appropriate statistics (selecting indices that legitimately summarise the measures), and meaningfulness (understanding what the resultant numerical values signify). Fundamentally, measurement concerns the exact relationship between the ‘Empirical Relational System’—the observed reality—and the ‘Formal (or Numerical) Relational System’ chosen to represent it. This process involves determining how subjective relational statuses or positions can be mapped onto mathematical operators, such as ‘greater than’ or ‘less than’, and understanding that not all social attributes possess inherent numerical properties, leading to a distinction between qualitative/non-metric variables and quantitative/metric properties like wealth or intelligence.

The Hierarchy of Measurement

The structure of measurement is defined by a hierarchy of levels of increasing complexity, established by thinkers such as S. S. Stevens. These levels dictate the types of mathematical transformations that are permissible while preserving the properties of the underlying concepts:

• Nominal Level: Data is categorised and labelled (e.g., male = 0, female = 1). Any one-to-one re-assignment of numbers maintains information about the categories, but no order is implied.
• Ordinal Level: Categories possess a meaningful order (e.g., a Guttman scale), meaning that order-preserving transformations are legitimate.
• Interval Level: Equal differences between objects correspond to equal intervals on the scale (as seen in temperature). Linear transformations must preserve these differences.
• Ratio Level: This highest level allows for the preservation of ratios between distances, such as moving from miles to kilometres.

Scaling and Quantification

The process of transforming raw empirical data into higher levels of measurement is known as scaling or quantification. While unidimensional scaling applies when the representation can be mapped along a single straight line (as in Likert scales), multidimensional scaling is required when multiple dimensions are necessary for accurate representation. Procedures for moving between these levels are crucial, and textbooks on survey research typically detail the appropriate statistical techniques corresponding to each level of measurement.