Hanson/Sigman: Leviathan’s Latent Dimensions: Notes
Core idea: State capacity is multidimensional and influential across economic development, civil conflict, democracy, and security. The authors address conceptual and measurement challenges by identifying three core dimensions, arguing they are mutually reinforcing, and estimating a general-purpose latent measure of state capacity using Bayesian latent-variable analysis.
Key contribution: A new, broadly comparable measure of state capacity with strong internal interrelationships among dimensions and demonstrated validity for wide empirical use.
Defining and Conceptualizing State Capacity
Core idea: State capacity embodies the state’s power to implement goals or policies and is distinct from normative judgments about legitimacy or the preferred scope of state action.
Infrastructural power (Mann, 1984): The capacity of the state to penetrate society and implement decisions across the realm. Soifer (2008) highlights three lenses on infrastructural power: state resources (material), effects on society, and territoriality.
Distinction emphasized in this work: Focus on capacities that exist within the state’s organizational structures and its territorial reach, not merely the effects of state action on social relations.
State capacity vs regime power: The authors separate state capacity from the broader power of the regime or the process of political decision-making (which can be shaped by democratic or autocratic features). This separation helps avoid conflating capacity with policy priorities or societal involvement in decisions.
Two key definitional debates:
Nature of state power: how to define and measure power without normative bias.
Functions (scope) of a capable state: Should capacity be treated as a general trait or as disaggregated across many functions?
Final definitional stance: State capacity is the state’s ability to perform the core functions necessary for modern states: protection from external threats, internal order, administration/provision of basic infrastructure, and revenue extraction.
Rationale: This middle-ground definition avoids normative judgments about what states should do and keeps measurement focused on core functions, enabling cross-national comparability.
Core functions summarized: protection, internal order, infrastructure provision, revenue extraction.
Dimensions of State Capacity
The literature identifies many potential dimensions (coercive, fiscal, administrative/implementation, transformative/industrializing, relational/territorial, legal, political, etc.).
The dimensionality of state capacity can be understood in three basic patterns:
1) Dimensions treated as independent (disaggregated approaches).
2) One dimension used as a proxy for the whole (e.g., taxation as a stand-in for capacity).
3) Dimensions that are mutually dependent and interrelated (complementary underpinnings that jointly enable a broad set of functions).Empirical challenges: Attempts to disaggregate have produced ambiguous results (e.g., distinct indicators loading on the same factor; difficulty separating infrastructural from coercive capacity).
The authors’ stance: While acknowledging multiple possible dimensions, they identify three core, minimally necessary, plausibly distinct dimensions that together enable modern state functions and are likely mutually constitutive. These three are:
Extractive capacity: ability to raise and mobilize resources (notably revenue/tax collection).
Coercive capacity: ability to enforce rules, protect borders, maintain internal order, and sustain a coercive apparatus.
Administrative capacity: organizational and bureaucratic capabilities to design, implement, deliver, and regulate public policy.
These three map onto Skocpol’s framing of general underpinnings: plentiful resources, administrative-military control of territory, and skilled, loyal officials.
The central argument: The three dimensions are interdependent in practice; strong extractive capacity supports coercive power and administrative machinery, while coercive and administrative capacities facilitate revenue extraction. This interdependence motivates measuring state capacity as a latent construct arising from the conjunction of the three capacities.
Policy/analytic implication: If dimensions are empirically inseparable, focusing on a broad latent capacity may be more informative than analyzing a single dimension in isolation. If they are separable, researchers should be cautious about using broad proxies for narrow concepts.
Indicators and Measurement Strategy
Purpose: Build a latent-variable measure of Capacity from indicators tied to the three core dimensions, with broad geographic and temporal coverage and minimal aggregation that could blur distinct concepts.
Indicator pool (21 indicators) spans 1960–2015 and up to 163 countries per year (overall: 56 years, 94,135 data points; 99% of country-years have at least six indicators; median indicators per country-year = 12).
Indicators by dimension (examples and sources):
Extractive capacity indicators:
Total tax revenue as % of GDP (ICTD/IMF/OECD)
Taxes on income as % of taxes; Taxes on international trade as % of taxes
Proportion of tax revenue from income taxes; Proportion from trade taxes
Tax revenue as a portion of GDP (overall extractive capacity)
Expert-coded: Efficiency of revenue mobilization (World Bank CPIA)
Fiscal capacity (V-Dem v9)
Coercive capacity indicators:
Military personnel per 1,000 population (COW, WDI)
Military expenditures per capita (log) (SIPRI, COW)
Police officers per 1,000 population (UN)
Monopoly on use of force (BTI)
Law and order (PRS International Country Risk Guide)
State territorial control/territorial authority (V-Dem v9)
State antiquity index (Bockstette, Chanda, Putterman)
Administrative capacity indicators:
Administrative efficiency (Adelman and Morris 1967)
Bureaucratic quality (PRS)
Quality of budgetary/financial management (CPIA)
Quality of public administration (CPIA)
Impartial public administration (V-Dem)
Census frequency (UN census data)
Information capacity (Brambor et al. 2020)
Statistical capacity (World Bank CPIA-derived, World Bank data)
Weberianness (Rauch and Evans 2000)
Indicator selection criteria:
Conceptual fit to the three core dimensions
Broad geographic and temporal coverage (1960–2015, many countries)
Avoid reliance on purely aggregate indexes; prefer components or indicators linked to a single dimension when possible
Information-gathering capabilities indicators:
Census frequency (US Census Bureau counts)
Information capacity index (Brambor et al. 2020): presence of statistical agency, civil register, population register, census/yearbook capabilities
World Bank statistical capacity index
Latent Variable Model and Estimation Details
Model concept: State capacity is latent and arises from the intersection of extractive, coercive, and administrative dimensions. Observable indicators are linear reflections of latent dimensions plus error.
Model variants and identification challenges:
They attempted to fit models with J dimensions where J ∈ {1, 2, 3}
In repeated runs, only the one-dimensional model (J = 1) converged consistently; two- and three-dimensional models failed to converge (chains did not stationary). Consequently, the authors interpret results as a single latent capacity dimension, Capacity, at the conjunction of the three core dimensions.
This empirical finding aligns with the idea that extractive, coercive, and administrative capacities are interrelated in practice, even if conceptually distinct.
Resulting measure:
Capacity is a latent, year- and country-specific value inferred from indicators across dimensions
The latent Capacity measure is designed to capture general-purpose state capacity for cross-national, long-run comparative research
Results: Capacity as a Single Latent Dimension
Key empirical finding: The data converge on a single latent dimension (Capacity) rather than distinct separate factors for the three dimensions. This supports the view that extractive, coercive, and administrative capacities are mutually constitutive.
Number of observations: 8,254 observations in total for Capacity estimates (across countries/years and indicators).
Coverage: Capacity estimates offer broader coverage than many existing general indicators of state capacity for postwar/postcolonial analysis.
Scale and distribution: Capacity is scaled from approximately 22.31 down to 2.96; with a mean around 0.26 and a standard deviation near 0.95.
Interpretation: Capacity is a general-purpose, cross-national, time-series measure that aggregates information across indicators spanning the three core dimensions and serves as a broad gauge of state capability.
Correlations with Base Indicators (Convergent Validity)
Capacity correlates strongly with a broad set of independent state-capacity-related measures, indicating convergent validity:
Statistical capacity (r ≈ 0.83)
Bureaucratic quality (PRS) (r ≈ 0.81)
Rigorous and impartial public administration (V-Dem) (r ≈ 0.80)
Law and order (PRS) (r ≈ 0.77)
Quality of public administration (CPIA) (r ≈ 0.74)
Monopoly on the use of force (BTI) (r ≈ 0.74)
Fiscal capacity (V-Dem) (r ≈ 0.73)
Quality of budgetary and financial management (CPIA) (r ≈ 0.71)
Administrative efficiency (Adelman and Morris) (r ≈ 0.70)
Military expenditures per capita (log) and other security-related indicators also show meaningful but somewhat weaker associations
Weaker associations observed for purely personnel-based coercive indicators (e.g., military personnel per 1,000; police per 1,000), suggesting that sheer sizes of security personnel are less informative about capacity than the administrative organization and information-handling dimension of coercive power.
A notable exception: Taxes on international trade as % of tax revenue shows a negative and weaker relationship with Capacity, consistent with informality that such taxes are easier to collect and may reflect different administrative challenges.
Face, Temporal, and Predictive Validity (Validity Checks)
Face validity: In 2015, capacity rankings place Singapore among the top quartile, while high-conflict or fragile states (e.g., Somalia, Yemen, Central African Republic) sit at the lower end, aligning with intuitive expectations about state functionality. The distribution across countries like Somalia and Singapore supports the intuition that Capacity tracks practical state function rather than simply democratic governance indicators.
Temporal variation: Capacity scores across years show strong positive autocorrelation; countries with high Capacity in 1975 tend to have high Capacity in 2015. The overall pattern indicates broad improvements in many countries over the postwar era, with notable declines in places experiencing conflict or state collapse (e.g., Somalia, Libya, Syria) and marked gains in others (e.g., Uganda, Bolivia, Rwanda, Lesotho, Nicaragua).
Convergent validity tests (non-MCMC indicators): Capacity correlates with widely used governance and development indices not used in the estimation process (e.g., WGI Government Effectiveness, Rule of Law, Regulatory Quality; Rothstein–Teorell’s impartial public administration; CPIA public-sector management; Hendix’s rational-legal index; BTI Stateness). These correlations provide convergent support for Capacity as a measure of state capacity rather than other constructs.
Nomological validity (predictive power for outcomes): The authors conduct six regression models to assess whether Capacity predicts development outcomes after controlling for GDP per capita and other factors. Key findings include:
Shadow economy: A 1-point higher Capacity in 1960 is associated with a reduction of approximately 1.96 percentage points of GDP in the shadow economy by 2010, controlling for GDP per capita.
Public services and state functionality: Higher Capacity predicts lower Myers index values (legibility/information) and improved public-service indicators in 2010.
Postal efficiency proxy: A 1-point Capacity increase is associated with a lower share of letters returned (i.e., higher administrative/operational efficiency) and fewer days for undelivered letters to be returned.
E-government readiness: A 1-point Capacity increase associates with about a 0.11-point increase in the UN e-government readiness index (roughly half a standard deviation).
All six models show Capacity as a significant predictor at the 1% level, even after controlling for initial GDP per capita, average polity score, and average tax revenue share.
Robustness checks: The paper includes an analysis where Capacity in 1960 is used as a predictor for 2010 outcomes, with controls for GDP and democracy, showing robust associations between early Capacity and later development indicators.
Implications and Interpretation
Interrelated dimensions: The empirical finding that a single latent Capacity dimension best fits the data supports the view that extractive, coercive, and administrative capacities are mutually reinforcing rather than strictly separable in practice.
Measurement advantages: A latent Capacity measure that incorporates indicators across dimensions provides a more comprehensive tool than single-dimension proxies, enabling broader temporal and cross-national analyses (especially useful in postwar/postcolonial contexts).
The chicken-and-egg question: The results invite further research on the causal order among dimensions (which dimension comes first, if any) and how these capacities co-evolve in different institutional and historical contexts.
Practical utility: The Capacity measure has been used in diverse studies (state-building, resilience in electoral authoritarian regimes, the democracy–capacity relationship, stock-market development) and serves as a meaningful control variable in cross-country analyses.
Limitations and future directions: The authors acknowledge that a one-dimensional latent solution is a pragmatic outcome given convergence, but emphasize potential for expanding measures, improving indicator coverage, and ensuring focus on core state functions rather than aggregation that conflates policy preferences or decision processes.
Summary of Core Takeaways
State capacity is best understood as a latent construct emerging from the interaction of extractive, coercive, and administrative capabilities.
A Bayesian latent-variable approach, drawing on indicators from multiple sources and across three domains, yields a general-purpose Capacity measure with broad country-year coverage and demonstrated validity.
The capacity measure correlates strongly with established indicators of governance and capacity, shows face and temporal validity, and predicts a range of development outcomes beyond GDP controls.
Empirically, the three core dimensions appear mutually reinforcing, suggesting that researchers should consider capacity as a unified, cross-cutting construct when feasible, while still recognizing context-specific variation across dimensions.
Notes on Terminology Used in this Summary
Capacity: The latent variable estimated in the study, reflecting overall state capacity across extractive, coercive, and administrative dimensions.
Extractive capacity: Revenue collection and financial sufficiency to sustain state functions.
Coercive capacity: Ability to enforce law, maintain order, and project power (military, police, monopoly on force).
Administrative capacity: Bureaucratic organization, efficiency, data and information handling, budgetary/administrative governance.
Infrastructural power (Mann): A related concept focusing on the state's reach and organizational capacity to implement decisions across society.