"Too often social scientists skip the step of defining the estimand. Instead, they leap straight to describing the data they analyze and the statistical procedures they apply."

Lundberg, Johnson, and Stewart (2021) call this "hypotheses about regression coefficients," which "defines the research goal inside a particular statistical model."

"If your research goal is a coefficient of a particular [statistical] model,

• It becomes impossible for the reader to know whether those procedures were appropriate.
• The methodological approach becomes tautological: ... It becomes impossible ... to reason about other approaches to achieve the goal. ... It cuts off productive consideration of a broader class of models that could accomplish the same goal.
• A goal defined entirely within a model bears a connection to theory that is questionable at best."