we-solve (beta)

Econ-201 · economics · Economics

Democracy's economic mechanisms and resilience

LLM status Open unverified

Statement

Original subject group: Political economy and institutions

Audit classification: Background; proposed extension. Original 2021 NBER metadata and revised primary manuscript checked. It already estimates exposure-support links and suggestive resilience; the broader feedback-loop specification is editorial.

Definitions and precise research target

Two distinct causal targets (editorial). Fix baseline countries, a finite horizon $T$, a democratic-status coding rule, normalized country weights, and fixed citizen cohorts. Democratic status $D_{ct}\in\{0,1\}$ is based on prespecified electoral contestability, participation, and executive constraints. Real income per resident $Y_{cT}$ uses a common price base and a declared resident denominator. Procedural support $S_{icT}$ is a measured survey index distinct from approval of the current incumbent. All relevant outcomes have finite means, and citizen weights $\mu_{ic}$ are nonnegative and sum to one within country.

A controlled exposure path $d=(d_0,\ldots,d_T)$ assigns democratic status at every date under a stated intervention definition. It can define income and support contrasts \[ \tau_Y^d=E[Y_{cT}(d^1)-Y_{cT}(d^0)],\qquad \tau_S^d=E\sum_i\mu_{ic}[S_{icT}(d^1)-S_{icT}(d^0)]. \] Such paths must specify timing, anticipation, financing, and the political rules being changed. A mere change of a regime label is not an intervention. Full status-path assignment fixes democratic survival by construction, so it is unsuitable for a nontrivial resilience target.

For resilience, instead fix countries democratic at baseline $t_0<T$, and compare feasible economic, information, or institutional reform regimes $p$ and $p_0$ that preserve the initial-status requirement while allowing future status to respond endogenously. Each regime specifies implementation, timing, public information, and resource use; it does not set future GDP, citizen support, or democratic status directly. Let $D_{ct}(p)$, $Y_{cT}(p)$, and $S_{icT}(p)$ be the resulting outcomes, and define \[ Z_{cT}(p)=\mathbf 1\{D_{ct}(p)=1\text{ for every }t=t_0,\ldots,T\},\qquad \tau_Z^p=E[Z_{cT}(p)-Z_{cT}(p_0)]. \] This survival indicator is an outcome, not an assigned path. Define analogous regime-specific income and procedural-support contrasts. Never combine a controlled-path contrast and an endogenous-survival contrast as if they arose from the same intervention.

The task is to characterize the identified sets for these contrasts, and determine which economic or institutional mechanisms connect democratization, prosperity, procedural support, and durability. Declare the admissible causal model class, transition/reform selection, global-shock and interference rules, attrition and cohort measurement, and any mediation assumptions. Transition timing is not randomized. The average income effect of democratization, a correlation between income and support, and a causal effect of truthful disclosure are different quantities. Randomized economic or information interventions can identify parts of the feedback chain but do not identify unrestricted mediation without additional exclusions and cross-world or sequential assumptions. The cited paper already supplies democratic-exposure/support estimates and suggestive resilience evidence; the combined problem here is a proposed extension, with no verified claim that this exact formulation is a published open conjecture.

Variants and assumption changes

  1. Exposure and beliefs (source-informed editorial): use cohorts with distinct democratic exposure and independently varied performance information, separating actual experience from perceptions under explicit exclusion restrictions.
  2. Shock resilience (editorial extension): define a prospective set of economic shocks and compare institutional-survival responses under alternative accountability institutions. Test whether baseline procedural support predicts survival and separately identify any causal support intervention.

References and source audit

Support and access. Original 2021 NBER metadata and revised primary manuscript checked. It already estimates exposure-support links and suggestive resilience; the broader feedback-loop specification is editorial. Locator: MIT author manuscript, October 2, 2023, abstract. Versions:
  1. Daron Acemoglu; Nicolás Ajzenman; Cevat Giray Aksoy; Martin Fiszbein; Carlos A. Molina. 2021. (Successful) Democracies Breed Their Own Support. Abstract, paragraphs on democracy exposure, economic performance, and support. https://www.nber.org/papers/w29167. DOI: https://doi.org/10.3386/w29167.
  2. Daron Acemoglu; Nicolás Ajzenman; Cevat Giray Aksoy; Martin Fiszbein; Carlos Molina. 2023. (Successful) Democracies Breed Their Own Support. October 2, 2023 author manuscript, abstract. https://economics.mit.edu/sites/default/files/inline-files/SuccessfulDemocracies2023.pdf.

Common conventions: Source mathematical conventions

These conventions apply to each entry unless explicitly replaced.

Models and identification. A primitive model $m$ specifies all probability laws, feasible choices, information, equilibrium and measurement rules used by its target. The admissible class $\mathcal M$ is fixed independently of outcomes. If $P_O(m)$ is its observable probability law and $\tau(m)$ the target, its identified set at observable law $P$ is \[ \mathcal I_\tau(P)=\{\tau(m):m\in\mathcal M,\ P_O(m)=P\}. \] Point identification means that this set is a singleton. An empty set signals incompatibility of data and assumptions. Coordinate bounds need not describe the joint set. Bounds are sharp only if every retained target is attainable or arbitrarily approachable by compatible models. Finite bounds require explicit restrictions on unobserved quantities; unrestricted confounding, hidden wealth, extrapolation or arbitrary equilibrium selection can yield uninformative sets.

Causal interventions. A potential outcome $Y_i(p)$ is the outcome under a complete policy regime $p$, including timing, financing, information and market spillovers. The target population and units are fixed before treatment. With interference, use the complete assignment vector or a justified exposure mapping; individual no-interference notation alone cannot identify an economy-wide intervention. A difference of mean potential outcomes does not require the cross-world joint distribution. Conditioning on post-treatment migration, survival, enrollment or employment changes the estimand and needs separate justification.

Statistical statements. An estimator, confidence set, forecast or test specifies a sampling law, model class and loss/coverage criterion. Uniform $(1-\alpha)$-coverage means $\inf_{m\in\mathcal M}\Pr_m\{\tau(m)\in C_n\}\geq1-\alpha$, or an explicitly asymptotic version. The whole parameter space trivially has coverage, so informativeness requires a stated width, risk or power objective. A forecasting improvement is relative to a named benchmark, information set and evaluation population.

Equilibrium and optimization. An equilibrium comprises feasible plans, optimality, beliefs where needed, budget constraints and all market/resource clearing. The primitives or a stated benchmark define each correspondence $\mathcal E(p;m)$. Nonemptiness is assumed or proved, never inferred just from compact policy sets. A selected-equilibrium target uses a declared selection rule; a robust target quantifies over all permitted equilibria. Use suprema or infima unless attainment is established. An $\varepsilon$-optimal policy is feasible and within $\varepsilon$ of the finite optimum value. Report an empty feasible set separately.

Welfare and accounting. Normative utility, interpersonal normalization, social weights, numeraire, discounting and the use of public receipts are primitives. Behavioral utility need not coincide with the welfare criterion. Household welfare includes ownership income and rents once; adding profits again to compensating variation generally double-counts them. A monetary equivalent is defined by an explicit transfer and price convention, with monotonicity and range conditions. Average effects, mechanism contributions, transfers and welfare losses are different targets. Infinite sums require convergence and dynamic feasible plans require terminal or no-Ponzi conditions.

Source versions. A working-paper date, revision date and journal publication year are distinguished. A DOI for a journal version is not automatically the DOI of an earlier manuscript. Locators refer to the stated version and printed pages where specified. A metadata-only check does not verify a claim about a full-text conclusion. Every entry's source subsection narrows the provenance claim accordingly.

Statement licence: CC-BY-4.0 · Source file

Contributions 1

Econ-201-A1 · Exploration · LLM-generated · gpt 6 astra ultra ·

GPT 6 Astra Ultra attempt on Econ-201

Democratic survival under reform interventions: sequential identification, sensitivity and a mediation obstruction

Problem ID: D72-197. Status: unresolved; rigorous restricted research attempt.

Target and the distinction between interventions

The catalogue separates controlled democratic-status paths from reforms that leave future democratic survival endogenous. The distinction is essential. The cited manuscript already studies democratic exposure and procedural support, and describes resilience evidence as suggestive [A]. We derive identification for a finite sequential reform design, a quantitative transition sensitivity bound, and a counterexample to identifying a support-mediated survival effect from randomized reforms alone. These results do not identify the effects of naturally selected regime changes in historical data.

Let all countries in the fixed baseline population be democratic at date zero. Let $D_t\in\{0,1\}$ be measured democratic status and $Z_T=\prod_{t=0}^TD_t$ the uninterrupted-survival outcome. A later return to democracy does not undo an earlier failure of this indicator. Let $X_t$ be a finite observed state, including income, citizen procedural support and any measured global shocks or histories needed for the restrictions below. Baseline distribution $\mu$ is fixed across regimes. A feasible reform rule $p$ selects action $A_t=p_t(X_t)$ at dates $t=0,\ldots,T-1$. Actions encode timing, financing, implementer and information; they do not assign income, support or democratic status. The country weights and citizen survey cohorts remain fixed. Death and attrition require prespecified outcome conventions; assume complete measurement here. Cross-country interference is excluded conditional on the declared global-shock process, which reforms do not alter.

A precise sequential identification result

Assume consistency and sequential exchangeability: conditional on the observed history up to $t$, action assignment is independent of future potential states and statuses under every feasible regime. Assume positivity for every history-action pair reached by each target regime. Finally assume a common controlled Markov law: conditional on current $X_t$, uninterrupted survival to $t$, and action $a$, the next state and next democratic status have the same distribution across histories and interventions. One can replace $X_t$ by the complete finite history to express a less compressed model. These are causal and transport assumptions, not consequences of panel-data notation.

Define a substochastic matrix on live states by \[ K_t^a(x,x')=\Pr\{X_{t+1}=x',D_{t+1}=1\mid X_t=x,Z_t=1,A_t=a\}. \tag{1} \] The missing row mass is democratic failure. Write $K_t^p(x,x')=K_t^{p_t(x)}(x,x')$.

Theorem: Under the stated assumptions, the survival probability for each feasible regime is point identified by \[ s(p)=\Pr\{Z_T(p)=1\} =\mu^{\mathsf T}K_0^pK_1^p\cdots K_{T-1}^p\boldsymbol{1}. \tag{2} \] Consequently the reform contrast $s(p)-s(p_0)$ is identified. The full joint terminal distribution of income and support is identified by the corresponding full-state transition product when exchangeability, positivity and the common law are assumed also after democratic failure. This does not identify a controlled democratic-status-path contrast unless an additional intervention law for assigning status is supplied.
Proof: Consistency, exchangeability and positivity equate each observed conditional transition in (1) with its interventional transition whenever the target rule selects $a$. Let $v_t(x)$ be the probability, under $p$, of state $x$ and uninterrupted survival at date $t$. Then $v_0=\mu$ and, by conditioning on current state and using the common transition law, $v_{t+1}(x')=\sum_xv_t(x)K_t^p(x,x')$. Induction gives the product in (2) after summing $v_T$ over live terminal states. The difference of two identified numbers is identified. The same conditional-probability recursion on the full state space identifies its entire terminal law; finite support makes integrable terminal income and support means immediate.

The action law in (1) conditions on actual democratic status transitions; it does not specify what would happen if political rules were intervened upon to set a status path. A constructive example takes a one-period setting with no variation in status, $D_1=1$, and observed income zero under every reform. Two models can set income under an additional status intervention $D_1=0$ respectively to zero and one, while agreeing on every observed and reform-generated outcome. Both satisfy all assumptions about reform assignment, but their controlled-status income contrasts differ. Thus those assumptions do not identify that separate target. ∎

The illustrative zero income in the last construction can be replaced by any positive common baseline plus the stated increment if real income is required to be strictly positive. A controlled path must specify the underlying institutional rules; merely changing a label would not define that counterfactual in either model.

A sensitivity bound for uncertain transition laws

For each row in (1), adjoin one absorbing failure state. The completed row is a probability distribution, so total variation has its ordinary meaning: $\operatorname{TV}(P,Q)=\frac12\sum_z|P(z)-Q(z)|$. Compare two candidate transition systems with the same baseline $\mu$ and same reform rule. Suppose for every live state and time the completed-row distance is at most $\varepsilon_t\in[0,1]$.
Proposition (Accumulated survival sensitivity): The resulting survival probabilities obey \[ |s(p)-\widetilde s(p)|\leq \min\left\{1,\sum_{t=0}^{T-1}\varepsilon_t\right\}. \tag{3} \] If the same bound applies to each of two reform rules, their survival contrast differs across models by at most $\min\{2,2\sum_t\varepsilon_t\}$.
Proof: Let $v_t(x)$ and $\widetilde v_t(x)$ be probabilities of final uninterrupted survival conditional on being alive in state $x$ at time $t$ in the two models. Terminal values are one on live states and zero on the absorbing failure state. Both recursions keep values in $[0,1]$. For any two finite probability laws $P,Q$ and function $f\in[0,1]$, $|\sum_zf(z)(P(z)-Q(z))|\leq\operatorname{TV}(P,Q)$: the maximum is attained by assigning one to positive differences and zero to negative differences, whose sums are respectively TV and minus TV. Apply this inequality to the difference between the two transition expectations of $\widetilde v_{t+1}$, and use the sup norm to compare $v_{t+1}$ and $\widetilde v_{t+1}$ under the first transition law. It gives \[ \|v_t-\widetilde v_t\|_\infty \leq\varepsilon_t+\|v_{t+1}-\widetilde v_{t+1}\|_\infty. \] Backward induction from zero terminal difference yields the sum bound. Averaging over the common $\mu$ preserves it, and two probabilities differ by at most one, proving (3). For a contrast apply the triangle inequality to the two rules; two numbers in $[-1,1]$ differ by at most two. ∎
Equation (3) is an outer sensitivity bound, not a claim that arbitrary transition confidence bands give a sharp joint set. It shows how modest unidentified transition changes can accumulate over a long horizon. Estimating transitions at histories with almost no support may leave large $\varepsilon_t$ even with many country-years.

Randomized reform does not identify a natural support channel

The next construction has one post-baseline period. $A\in\{0,1\}$ is a randomized economic or information reform, $S$ is binary procedural support, and $Z=D_1$ is democratic survival. The intervention does not set future support or status. A natural indirect effect at reference reform zero is \[ N={\mathbb{E}}[Z(0,S(1))-Z(0,S(0))], \tag{4} \] where $Z(a,s)$ refers to an additional, well-defined intervention setting the support mediator to $s$. This is stronger than a total reform effect.
Proposition (Two models with identical reform evidence): Even exact knowledge of the joint distribution $(A,S,Z)$ under randomized $A$, including both treatment arms with positive probability, does not identify (4). There exist two binary structural models with the same observed law and total reform effect one, but natural indirect effects one and zero, respectively.
Proof: Let $U$ be a fair Bernoulli variable independent of randomized $A$. Use exclusive-or notation $a\oplus u$ and set the support equation in both models to $S=A\oplus U$. In model I define the survival response under mediator intervention by $Z(a,s)=s\oplus U$; in model II define $Z(a,s)=a$. Baseline democratic status is one in both models. Under the naturally generated mediator, model I has $Z=(A\oplus U)\oplus U=A$, while model II also has $Z=A$. Both therefore have exactly the same joint law: conditional on either $A$, support is fair Bernoulli and survival equals $A$ with probability one. Both total effects are ${\mathbb{E}}[Z(1)-Z(0)]=1$.

For model I, $S(1)=1\oplus U$ and $S(0)=U$. Consequently $Z(0,S(1))=1$ and $Z(0,S(0))=0$ for every $U$, so (4) equals one. In model II, $Z(0,s)=0$ for either $s$, so (4) is zero. Both models define all mediator interventions and use the same exogenous law. The difference is an unobserved causal response to support, which randomizing the initial reform does not reveal. ∎

This construction does not claim that support lacks a causal role in real democracies. It proves that a reform experiment and the observed joint support-survival law do not by themselves determine that role. A support intervention, exclusion restriction or justified sequential mediation assumption would add information. Controlling for post-reform support in a regression is not a substitute for those assumptions.

Remaining target and source provenance

Historical regime transitions are not sequentially randomized, and unmeasured conflict, international spillovers and reform anticipation can invalidate (1). The finite-state Markov reduction can require far richer histories than the data support. Income and support must be measured in the fixed cohorts, including those experiencing democratic failure; restricting to surviving democracies changes the target. A truth-telling information experiment identifies its own intervention, not automatically a prosperity or democracy-exposure effect.

The October 2, 2023 MIT manuscript's abstract and opening discussion were inspected on 9 October 2026. It already reports democratic-exposure/support results and suggestive resilience evidence. We attribute no theorem of unrestricted mediation identification or universal feedback stability to that paper. The source's historical empirical assumptions and the stronger sequential intervention assumptions used here are distinct. All results above are self-contained restricted reductions; the full identified sets and mechanisms in the catalogue target remain unresolved.

References

Completion Estimate

30%

Go to contribution →