Abstract
Management research routinely makes causal claims about organisations, behaviour,
routines, strategy, and innovation, yet the field lacks a common upstream framework
for specifying what a causal problem actually is. The result is fragmentation, as behavioural
studies emphasise psychological mechanisms, strategy research adopts structural
or econometric logics, qualitative traditions foreground meaning and institutions,
and machine learning approaches prioritise prediction. These perspectives often model
different implicit causal problems without shared primitives, constraining the coherence
of theory-building, identification, and intervention design.
This paper introduces Causal Problem Modelling (CPM) as a domain-general methodological
framework that clarifies and formalises the causal assumptions underlying management
theories and empirical designs. CPM defines a causal problem through five
primitives, the causal architecture, feasible intervention set, data-generating situation,
identification surface, and counterfactual constraint set. These primitives provide an integrative
language for representing organisational phenomena, specifying mechanisms,
and analysing the conditions under which causal claims are meaningful or testable.
We show how CPM strengthens management scholarship by (1) making causal architectures
explicit in theories of routines, leadership, identity, innovation, and strategy;
(2) refining intervention semantics in organisational and policy contexts; (3) improving
1
construct clarity through architectural grounding; (4) enabling multi-method integration
across qualitative, quantitative, computational, and machine learning approaches;
and (5) revealing when identification is robust, fragile, or impossible within a given
organisational context.
CPM offers management researchers a coherent, intervention-ready, method-agnostic
framework for designing, diagnosing, and improving causal explanations in complex organisational
and socio-technical systems. It positions causal problem formulation, not
data, method, or model, as the primary unit of analysis, advancing the foundations of
causal reasoning in management research.