6.1 Statistical Control

Randomized experiments often considered the gold standard in scientific research.However, it is often infeasible, difficult or impossible to manipulate the putative effect of interest:

• cannot randomly resettle individuals into different strata of society
• cannot assign couples to stay married or get a divorce
• cannot randomize children to different levels of adversity

Researchers have employed a number of different strategies in response to the limitations of observational data:

• surrogate interventions
  • item the real-life cause of interest cannot be manipulated, often a proxy can be randomized in the lab.
• avoiding causal language
  • item write about associations and relationships despite substantive question about causal effects
• statistical control
  • include control variables in analysis but choice of variables often difficult and statistical control itself insufficient