Time Series Design

Time Series Design Definition

A time series design is a research methodology in which measurements of identical variables are collected across multiple points in time, typically with the aim of studying temporal social trends. These designs are distinguished from cross-sectional designs, where data is collected only once, as they allow researchers to observe how variables change over a period.

Application and Inference

This design is highly effective when analysing official data, such as plotting crime rates for the same geographical area across different time intervals (e.g., monthly or annually). This method allows for the identification of trends in levels of these variables. Furthermore, it permits the simultaneous plotting of multiple variables to investigate potential relationships between them, such as correlating unemployment rates with crime rates. However, while a time series design can establish an association between changes in variables, it does not inherently provide sufficient evidence to infer causality—that is, it cannot definitively prove that one change causes another.

Time Series Sampling

The term is also applied to survey methodologies where equivalent samples of the population are drawn at various points in time to collect data. This involves gathering information from sample members repeatedly using consistent principles and criteria. An example is the British Crime Survey (BCS), which has been repeated at regular intervals since 1982, allowing for longitudinal tracking of crime levels. Such surveys are valuable because they provide reliable indicators of societal trends, particularly when official recorded data is deficient, such as in estimating the ‘dark figure’ of unrecorded crime.

Evaluation

Time series designs are primarily descriptive rather than explanatory. While they excel at mapping and describing changes and trends within a society, they do not offer direct evidence for explaining the underlying causes of those changes. Although these designs can be used to theorise about trend evolution, they lack the necessary empirical basis to establish causality. Consequently, while useful for macro-level analysis, they are generally insufficient for making inferences about the individual development trajectories of specific groups or individuals; for such purposes, longitudinal cohort surveys are required.