Model Structure

The transactionality index combines nine normalised input variables $\tilde{x}_i = x_i / 10$ with expert-elicited weights $w_i$ (summing to 1.0), two multiplicative interaction terms, and a convex autonomy adjustment:

where $\gamma_1 = 0.08$ and $\gamma_2 = 0.06$ are interaction coefficients and $\ell$ is normalised autonomy. The index maps to four posture regimes via heuristic thresholds:

• $T < 25$: Rules-first / institutionalist
• $25 \leq T < 45$: Selective engagement
• $45 \leq T < 65$: Hedging / insurance
• $T \geq 65$: Hard transactionalism

Interaction Terms

A purely additive model misses crucial conditional dynamics. Walt's balance-of-threat theory implies that high threat with a credible alliance is qualitatively different from high threat without one — the interaction term $\tilde{x}_{\text{thr}} \cdot (1 - \tilde{x}_{\text{all}})$ captures this compounding effect. Similarly, Thorhallsson's shelter theory predicts that great-power rivalry in a region without institutional cover produces amplified pressure beyond what the additive terms alone would suggest.

Crisis Regime

Under the crisis regime, weight mass is reallocated so that the institutional shelter deficit weight increases (+0.02). This follows Thorhallsson (2011), who argues that shelter is most critical precisely during crises — a shelter deficit is costlier when the threat is acute. Conversely, reputational capital weight decreases (−0.03) because short-term survival overrides reputation investment under acute threat.

Theoretical Sources

Factor selection draws on established IR concepts: Thorhallsson's institutional shelter theory (political, economic, and societal shelter dimensions); Kuik's hedging framework (structural uncertainty as scope condition, domestic legitimation as primary variation driver); Walt's balance of threat (proximity, capability, intentions); and Hirschman's dependency concentration analysis (operationalisable via Herfindahl-Hirschman Index).

Note: the reputational capital factor reflects the Weisiger/Yarhi-Milo position that reputations matter in IR. This is actively contested — Press and Mercer argue reputations matter less than commonly assumed. The model implicitly takes a side in this ongoing debate.

Limitations

This is an expert-judgment heuristic, not an empirically calibrated model. The specific weights are elicited rather than regression-derived — comparable in methodology to the Fragile States Index. The posture thresholds (25/45/65) are decision-theoretic boundaries, not empirically observed cutpoints.

Missing factors include geographic proximity (Walt), regime type (democratic vs authoritarian states hedge differently per Kuik), economic structure (resource-rich vs service economies), and normative identity. Sanctions exposure presents an endogeneity concern — it is partly a consequence of posture rather than purely an input to it. Domain-specific posture variation (a state may be transactional on energy security while rules-first on border norms) would require sector-weighted extensions.