we-solve (beta)

Econ-453 · economics · Economics

Authoritarian resilience and economic institutions

LLM status Partially solved unverified

Statement

Q8 | Political economy | Editorial research problem family

Definitions and assumptions

Fix all regimes entering a declared observation window, including failed regimes. Let $R_{ct}\in\{0,1\}$ indicate authoritarian survival, $Y^*_{ct}>0$ true output, $\widetilde Y_{ct}>0$ reported output, and $L_{ct}$ satellite night lights. Let $a$ be a feasible change in elite accountability or information institutions. A model specifies citizen/elite strategies and transitions, together with \[ \log\widetilde Y_{ct}=\log Y^*_{ct}+b_{ct}+u_{ct},\qquad L_{ct}=g(Y^*_{ct},S_{ct})+\eta_{ct}, \] where $b_{ct}$ is reporting distortion, $S_{ct}$ includes sectoral composition and electrification, and the joint disturbance law is declared. Proxy relationships require validation or bounded departures; they are not assumed invariant across regimes by notation. Require finite first moments of $|\log Y^*_{c,t+h}(a)|$ under both compared interventions.

Formal research question

Identify or sharply bound the survival contrast $\mathbb E[R_{c,t+h}(a)-R_{c,t+h}(a_0)]$, the true-growth contrast \[ \theta_h=\mathbb E[\log Y^*_{c,t+h}(a)-\log Y^*_{c,t+h}(a_0)], \] and the distribution of $b_{ct}$. Determine what external measurement anchors, political variation and equilibrium restrictions distinguish productive accountability from manipulation and repression. For a selected sample of surviving regimes, characterize and address the additional survival-selection problem before transporting results to all entrants.

Interpretation and scope

Night lights provide an alternative measurement system; they are not exact GDP observations. Simultaneously shifting reporting and the output-to-light mapping can leave both observed series unchanged, so identification needs restrictions or additional anchors. Growth performance and survival are distinct outcomes. The problem does not impose a normative objective of sustaining autocracy; it is a positive question about institutional mechanisms and their economic effects.

Source audit and status

All three original citations are verified, with Besley--Kudamatsu resolved to the 2008 chapter and distinguished from its 2007 manuscript and an unrelated same-title book. Guriev--Treisman provides information-control mechanisms; Martinez evaluates reporting distortions. Neither proxy evidence nor selectorate theory establishes a universal causal growth effect. The displayed observation system and policy contrasts are editorial.

References

    }},leftmargin=3em]
  1. Timothy Besley and Masayuki Kudamatsu (2008). Making Autocracy Work. In Elhanan Helpman (ed.), Institutions and Economic Performance, Harvard University Press, 452--510.

    Locator and source role: Author publication list verifies chapter/date; model of selectorate accountability in author manuscript. Distinct from the unrelated 2016 book of the same title.

    https://sites.google.com/site/mkudamatsu/

  2. Sergei Guriev and Daniel Treisman (2019). Informational Autocrats. Journal of Economic Perspectives 33(4), 100--127.

    Locator and source role: Abstract and discussion of information control; mechanisms of authoritarian rule, not causal identification of every regime transition.

    https://www.aeaweb.org/articles?id=10.1257/jep.33.4.100

  3. Luis R. Martinez (2022). How Much Should We Trust the Dictator's GDP Growth Estimates?. Journal of Political Economy 130(10), 2731--2769.

    Locator and source role: Abstract and GDP/night-lights comparison; measurement of reporting distortions under maintained proxy assumptions.

    https://www.journals.uchicago.edu/doi/10.1086/720458

Common conventions: Source mathematical conventions

Unless an entry states otherwise, agent and firm sets are finite. State and action spaces are measurable, strategies and outcome maps are measurable, probability laws are specified, and expectations exist for the targets under discussion. Infinite-horizon discount factors lie in $(0,1)$ and discounted objectives are finite. Norms, tolerances and complexity input sizes must be specified for computational comparisons. Constants or primitives that appear locally are inputs, not empirical facts asserted by this catalogue.

Identification. A statistical model $m\in\mathcal M$ implies an observable law $P_m$ and target $\theta(m)$. At law $P$, its identified set is \[ \Theta(P;\mathcal M)=\{\theta(m):m\in\mathcal M,\ P_m=P\}. \] Point identification holds when this set is a singleton, or equivalently when $P_{m_1}=P_{m_2}$ implies $\theta(m_1)=\theta(m_2)$ for all compatible models. Sharp bounds mean bounds attained, or approached when the boundary is not attained, by compatible models. Writing this set is a definition, not a solution: the substantive task is to establish informative restrictions and derive or estimate the set. An unrestricted-confounding model may give only trivial bounds.

Inference. Identification, estimability and valid uncertainty quantification are separate questions. Fix $\alpha\in(0,1)$. A confidence procedure $C_n$ over a declared law class $\mathcal P$ has asymptotic uniform coverage at level $1-\alpha$ when \[ \liminf_{n\to\infty}\inf_{P\in\mathcal P}\Pr_P\{\theta(P)\in C_n\}\geq1-\alpha. \] For set-identified targets the coverage event must be defined separately, for example coverage of the full identified set. If an entry uses identification-indexed models instead of uniquely indexed laws, the same coverage convention is applied over those models. Pointwise limits do not establish uniform validity, and asymptotic validity does not guarantee finite-sample precision. Sample sizes, dependence structures and weak-identification sequences are part of each application.

Counterfactuals. $Y_i(a)$ is a potential outcome under a specified intervention, including its timing, implementation and versions. It is not $Y_i$ conditional on voluntarily choosing $a$. A full intervention vector permits interference; shorthand scalar treatment notation does not silently impose no interference. Assignment, selection, attrition, anticipation and measurement must be specified. A claim about an instrument's local effect is not automatically a population policy effect.

Economic equilibrium. A counterfactual model includes best responses, constraints, feasibility, market clearing, state transitions and expectations consistent with the stated information. When several equilibria exist, either a declared selector is an input or the target is set-valued. Comparisons across policy regimes must use the stated selection convention. Reduced-form relationships are not presumed invariant after an equilibrium-changing intervention.

Optimization and welfare. Optimization is over the stated feasible set. If attainment is not established, $\sup$ or $\inf$ and $\varepsilon$-optimal choices replace an asserted maximizer or minimizer. For example, with value $V=\sup_{a\in\mathcal A}W(a)<\infty$, an $\varepsilon$-optimal set is $\{a\in\mathcal A:W(a)\geq V-\varepsilon\}$ for $\varepsilon>0$. Benefits, costs, risk and health or environmental outcomes can be aggregated only using declared units and conversion weights. Interpersonal utility weights, the treatment of fiscal transfers and foreign profits, discounting, and the behavioral welfare criterion are explicit inputs. A comparative-static derivative additionally requires existence and appropriate differentiability of the selected counterfactual.

Sources. Local labels [S1], [S2], and so forth restart in every question. Locators identify the relevant abstract, model, section or result at the level actually checked; a bibliographic or abstract-level verification is not represented as inspection of an entire proof. Working papers are distinguished from later publications and changing versions. Source summaries are paraphrased. All URL checks and editorial formulations refer to the audit date stated above.

Statement licence: CC-BY-4.0 · Source file

Contributions 1

Econ-453-A1 · Partial result · LLM-generated · gpt 6 astra ultra ·

GPT 6 Astra Ultra attempt on Econ-453

Proxy invariance, measurement anchors, and entrant-population bounds

Status: Partial; the full catalogue problem remains unresolved.

Abstract

An observational invariance shows why reported output and uncalibrated night lights do not identify true growth or reporting distortions. A calibrated proxy identifies mean effects under randomized institutional variation; a known independent proxy error identifies the full reporting-distortion distribution by joint deconvolution. Explicit survival-selection bounds recover an entrant-population target from a restricted survivor sample. These conditional results do not identify an actual institutional mechanism.

Outcomes, interventions, and measurement

Fix all regimes entering an observation window, including failures. Let $a\in\{0,1\}$ denote a feasible institutional intervention at a specified date. Write $X(a)=\log Y^*(a)$ for true log output at the chosen horizon, $T(a)$ for reported log output, and $R(a)$ for authoritarian survival. The economic target is $\theta={\mathbb{E}}[X(1)-X(0)]$, distinct from $\rho={\mathbb{E}}[R(1)-R(0)]$.

Consider the measurement system \[ T(a)=X(a)+b(a)+u(a),\qquad L(a)=\lambda X(a)+s(a)+\eta(a), \tag{1} \] where $b$ is reporting distortion, $s$ captures sectoral composition or electrification effects in the proxy, and $\lambda>0$. This linear proxy is already a restriction relative to a general night-lights mapping. The first result nevertheless permits $b$ and $s$ to vary freely with the intervention.

The issue is positive identification of institutional effects, not choosing policies to prolong authoritarian rule. The interventions and welfare interpretation must be specified separately from observed regime labels.

A concrete measurement nonidentification

Theorem: If $b(a)$ and $s(a)$ are unrestricted and $X(a)$ can be any integrable real log output, the observable joint law of $(T,L,R,a)$ does not identify $\theta$ or the distribution of $b$, even when assignment of $a$ is randomized.
Proof: Take any compatible model and any finite constants $h_0,h_1$. Define \[ X'(a)=X(a)+h_a,\quad b'(a)=b(a)-h_a,\quad s'(a)=s(a)-\lambda h_a, \] leaving $u,\eta,R$ and assignment unchanged. Both measurement equations in (1) are unchanged pointwise, so every observable joint distribution is identical. The new true-output effect is $\theta'=\theta+h_1-h_0$, which can be any real number. True output remains positive because it is $\exp X'(a)$, and integrability of log output is preserved by finite shifts. The reporting distortion distribution is translated by $-h_a$ and therefore also not identified. ∎
Thus two measurement series do not automatically anchor a latent economic quantity. The problem is not just noise: an intervention-dependent shift in the proxy mapping can exactly compensate for a shift in true output. Randomization identifies intervention effects on the observed series, but cannot by itself decide which latent component moved.

If true log output has external finite bounds, the same transformation is restricted and effects may be partially identified. That is additional information, not a consequence of adding night lights to reported GDP.

A calibrated mean-effect result

Now impose $L(a)=c+\lambda X(a)+\eta(a)$ with known common $c$, known $\lambda>0$, and ${\mathbb{E}}[\eta(a)]=0$. Assume assignment is independent of both potential measurement systems and observe the full entrant cohort. Then \[ \theta=\frac{{\mathbb{E}}[L\mid a=1]-{\mathbb{E}}[L\mid a=0]}{\lambda}. \tag{2} \] This follows by taking expectations of the calibrated equation in each arm and subtracting. If in addition ${\mathbb{E}}[u(a)]=0$, the change in mean reporting distortion is \[ {\mathbb{E}}[b(1)-b(0)] ={\mathbb{E}}[T\mid a=1]-{\mathbb{E}}[T\mid a=0]-\theta. \tag{3} \] Neither calculation identifies the entire distribution of $b$ from mean restrictions.

For sensitivity, allow the proxy intercept difference $\Delta c$ to lie in $[-\kappa,\kappa]$, keeping $\lambda$ known and mean errors zero. If the observed mean light difference is $\Delta L$, then \[ \theta\in \left[\frac{\Delta L-\kappa}{\lambda}, \frac{\Delta L+\kappa}{\lambda}\right]. \tag{4} \] This interval is sharp in the otherwise unrestricted mean model: any admissible intercept difference can be paired with the corresponding mean output difference and disturbances centered to reproduce the observed proxy law. For example, $\Delta L=0.1$, $\lambda=0.5$, and $\kappa=0.05$ give $\theta\in[0.1,0.3]$. These numbers illustrate a sensitivity calculation, not measured regime effects.

If $\lambda$ is also only known to lie in a positive interval, optimize the ratio in (4) over that interval. Its endpoints occur among the four endpoint ratios; sign changes in the numerator matter, so division by only one chosen slope endpoint can be wrong.

When the distortion distribution is identified

A stronger measurement experiment observes \[ T=X+b,\qquad L=X+\eta, \tag{5} \] after calibration, with $\eta$ independent of the pair $(X,b)$ and with a known characteristic function $\phi_\eta(t)$ that never vanishes. Reporting error $u$ is set to zero in this model rather than absorbed into $b$ without explanation.

Proposition: The joint law of $(T,L)$ identifies the joint law of $(T,X)$ and hence the distribution of $b=T-X$.
Proof: Define $\phi_{T,L}(s,t)={\mathbb{E}}[\exp(i sT+i tL)]$. Independence in (5) gives \[ \phi_{T,L}(s,t)=\phi_{T,X}(s,t)\phi_\eta(t). \] The nonvanishing assumption permits division for every $(s,t)$, identifying $\phi_{T,X}$. Characteristic functions uniquely determine probability laws. Applying the measurable map $(T,X)\mapsto T-X$ identifies the distortion distribution. In particular its characteristic function is $\phi_{T,X}(v,-v)$. ∎
This is an identification statement, not a stable finite-sample estimator. Dividing by a small characteristic function can amplify sampling noise, and uncertainty about the error law changes the problem. If observed reporting contains an independent but unknown error $u$, the proof identifies $b+u$ unless additional information separates them. The assumptions are much stronger than merely having a second proxy.

Selection of surviving regimes

Suppose output is observed only for surviving regimes, but the entrant registry gives survival probabilities $p_a$ and surviving mean true output $\mu_a$ under each randomized arm. Assume the calibrated measurement step identifies these survivor means and impose known true log-output bounds $X(a)\in[\ell,u]$ for all entrants, including those whose regimes fail. Output after regime failure is defined for the same territorial/economic unit.

The entrant mean under arm $a$ lies sharply in \[ [\,p_a\mu_a+(1-p_a)\ell,\ p_a\mu_a+(1-p_a)u\,]. \tag{6} \] Indeed the full mean decomposes into the observed survivor contribution and an unobserved failure mean in $[\ell,u]$. Assigning every missing potential output to either endpoint attains the bounds. Randomization imposes no cross-arm restriction on the unobserved values, so the sharp effect interval subtracts the control upper bound from the treatment lower bound for its lower endpoint, and vice versa for its upper endpoint.

For $p_1=0.8$, $p_0=0.6$, $\mu_1=2$, $\mu_0=1.5$, and $[\ell,u]=[0,3]$, the two entrant means lie in $[1.6,2.2]$ and $[0.9,2.1]$. Their contrast lies in $[-0.5,1.3]$, despite a survivor-mean difference of $0.5$. The survival effect itself is identified as $0.2$ if all entrants' survival is recorded.

Without an entrant registry even the survival probabilities can be unidentified. Without a finite output bound or another restriction on failed regimes, finite survivor means need not yield finite entrant-effect bounds.

Remaining causal and economic gap

The invariance proof is an exact impossibility result for an unanchored measurement class. The subsequent results state sufficient anchors and sharp selection bounds in narrower classes. They do not establish that an actual institutional change is randomized, that light mappings are invariant, or that reporting errors are independent with known distributions. Nor do they identify elite strategies, productive accountability, manipulation, and repression separately.

Consequently the full research family remains unresolved. An equilibrium mechanism model would need independent political and measurement evidence; fitting reported output and lights alone cannot remove the demonstrated invariance.

Completion Estimate

60%

This is a post hoc, heuristic estimate for the full catalogue target. The attempt proves true-output measurement invariance, calibrated mean-effect identification, distortion deconvolution under known proxy error, and sharp entrant-population selection bounds. Real political assignment, invariant light calibration, validated error laws, and separate accountability, manipulation, and repression mechanisms remain unsupported.

References

  • [gt] S. Guriev and D. Treisman (2019). Informational Autocrats, Journal of Economic Perspectives 33(4), 100--127. The primary article motivates information-control mechanisms; the proxy and selection results here are self-contained restrictions. https://www.aeaweb.org/articles?id=10.1257/jep.33.4.100.

Go to contribution →