Uncovering Hidden Configurations: Specifying Outcomes in National Innovation Systems with Archetypal Analysis

Siman CAI, The Chinese University of Hong Kong
Junru Chen, The Chinese University of Hong Kong

Comparative research on national innovation systems is organized around familiar ideal types—liberal, coordinated, or developmental models—that structure how configurational analysis proceeds. These types define the categories into which countries are sorted and the outcomes whose causes are investigated. Yet configurational methods presuppose the existence and adequacy of such types; they ask which conditions lead to a given model, but not whether the models themselves accurately capture the range of configurations present in the data. Without a procedure for deriving types empirically, causal analysis risks resting on typological foundations that are incomplete or theoretically inherited rather than observed. We introduce Archetypal Analysis (AA) as a methodological extension to configurational research. AA identifies boundary-defined types inductively from data and represents cases as convex combinations of them, enabling outcomes to be specified as structured mixtures rather than unitary categories. Using Global Innovation Index data for 147 countries (2013–2020), we identify 44 innovation archetypes—distinct boundary-defined innovation modes that serve as empirical analogues to Weberian ideal types—16 of which persist throughout the period. These include familiar cases such as the United States, Japan, and China, as well as less-recognized configurations such as Romania and Latvia. At the same time, the global innovation space contracts by 8.5 percent, indicating convergence toward a more standardized model even as multiple archetypes endure Methodologically, this framework formalizes a step long assumed but rarely operationalized in configurational research: the empirical specification of outcome types prior to causal analysis. Empirically, the findings demonstrate that global innovation systems are characterized by both persistent structural diversity and measurable convergence. This dual pattern provides a clearer basis for investigating why some innovation modes endure, why others emerge or disappear, and how countries move within an evolving configuration space.

See extended abstract

 Presented in Session 204. Property Rights, Land Markets, and Cultural Industries