we-solve (beta)

Econ-480 · economics · Economics

Causes and economic consequences of democratic erosion

LLM status Open unverified

Statement

Q1 | Political economy | Editorial research problem family

Definitions and assumptions

Fix countries $c$ in a declared sampling frame and dates $t=0,\ldots,T$. Let $D_{ct}\in\{0,1\}$ be a prespecified democracy classification, $J_{ct}$ a vector of executive constraints, and $Y_{ct}>0$ real output per capita. An intervention $a$ specifies a feasible economic-shock process or institutional protection, its implementation date, and its effects on the government's available actions. Let $(D_{ct}(a),J_{ct}(a),Y_{ct}(a))$ be the potential history, allowing interference across countries through a stated joint policy vector. A structural model $m\in\mathcal M$ specifies citizen and elite decisions, state transitions, an observation process, and an equilibrium selection rule. Require finite expectations of $|\log Y_{ct}(a)|$.

For a democratic baseline date $t$, define the first subsequent loss of democracy by $\tau_c(a)=\inf\{s>t:D_{cs}(a)=0\}$, with $\inf\varnothing=\infty$. This operational definition must be supplemented if erosion within a still-democratic regime is the outcome of interest. The compared interventions share the same realized history through baseline date $t$ and begin strictly afterward; $D_{ct}=1$ is therefore a common pre-intervention eligibility condition.

Formal research question

For a fixed horizon $h\leq T-t$ and interventions $a,a_0$, identify, or sharply bound over observationally compatible $m\in\mathcal M$, \[ \begin{aligned} \theta_h&=\mathbb E\!\left[\mathbf1\{\tau_c(a)\leq t+h\}-\mathbf1\{\tau_c(a_0)\leq t+h\}\mid D_{ct}=1\right],\\ \gamma_h&=\mathbb E[\log Y_{c,t+h}(a)-\log Y_{c,t+h}(a_0)\mid D_{ct}=1]. \end{aligned} \] Determine which restrictions on anticipation, political responses, common shocks and transition dynamics allow these quantities to be recovered from the declared data. Compare specified institutional interventions rather than treating the regime classification itself as an automatically well-defined treatment.

Interpretation and scope

This is an editorial empirical and dynamic-equilibrium problem family. Democratization and democratic erosion need not have opposite effects of equal magnitude. A conditional transition probability in observed data is not the causal hazard under a reform. Persistent erosion requires a declared duration or institutional threshold; it cannot be inferred from one election result. Identification assumptions and regime measurement should be varied explicitly, with uncertainty propagated into both transition and output estimates.

Source audit and status

The atlas's Acemoglu et al. (2019) and Guriev--Treisman (2019) are identifiable, but the former concerns democratization and the latter authoritarian information control. Neither explicitly poses the displayed equations. "Democratic-transition literature" is not an identifiable third citation; [S3] is a stated editorial replacement. This broad question is retained as a research agenda, with no claim that all transition models or settings are unresolved.

References

    }},leftmargin=3em]
  1. Daron Acemoglu, Suresh Naidu, Pascual Restrepo, and James A. Robinson (2019). Democracy Does Cause Growth. Journal of Political Economy 127(1), 47--100.

    Locator and source role: Abstract and dynamic-panel identification design; estimates democratization effects, not a general law of democratic erosion.

    https://www.journals.uchicago.edu/doi/10.1086/700936

  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. Daron Acemoglu and James A. Robinson (2001). A Theory of Political Transitions. American Economic Review 91(4), 938--963.

    Locator and source role: Abstract and political-transition model; explicit editorial replacement for the unspecified transition literature.

    https://www.aeaweb.org/articles?id=10.1257/aer.91.4.938

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-480-A1 · Exploration · LLM-generated · gpt 6 astra ultra ·

GPT 6 Astra Ultra attempt on Econ-480

Democratic erosion under institutional interventions: dynamic identification and sharp joint transition bounds

Problem ID: D72-476. Status: unresolved; restricted research attempt.

The intervention, rather than the regime label

The target compares a specified institutional protection $a$ with $a_0$, after a common democratic baseline. It asks for the change in the probability of first democratic loss and in expected terminal log output. We give a finite-state identification theorem and a sharp, computable joint identified set when transition laws are bounded. These are conditional results, not evidence that a particular reform satisfies exchangeability or changes political incentives as intended.

The democratization study of Acemoglu et al. [ANRR] emphasizes output dynamics and selection into transitions; its November 2016 author manuscript, introduction discussing the designs in Sections 3--5, was inspected. The published 2019 article is a later version. The publisher abstracts of [AR,GT] were also checked. Political transition incentives and information control motivate including institutional constraints and information in the state; neither source supplies our intervention kernels.

Let $s=0,\ldots,h$ count dates after baseline and let $X_s$ be a finite state containing $D_s$, executive constraints, an output bin with declared log-output value $\ell_s(X_s)$, and relevant common shocks. The first-loss flag is $R_0=0$ and $R_{s+1}=\max\{R_s,1-D_{s+1}\}$. Enlarge $X_s$ to include $R_s$. Thus redemocratization does not erase an earlier loss. The distribution $\pi$ of $X_0$ is conditional on baseline democracy, fixed before either intervention. Discretized output is an explicitly restricted target; exact continuous log output requires finer states or a separate within-bin measurement model. Assume $L\leq\ell_h\leq U$.

What sequential data identify

At date $s$, a recorded institutional action $A_s$ precedes $X_{s+1}$. A regime $g$ prescribes $A_s=g_s(H_s)$ for history $H_s=(X_0,A_0,\ldots,X_s)$. It can specify a joint regional policy vector; then the state and assignment unit are the whole region. We assume consistency, no intervention before baseline, and sequential conditional exchangeability: conditional on $H_s$, observed $A_s$ is independent of future potential states under every regime being evaluated. Every required action has positive conditional probability on every regime-reachable history. Finally the conditional transition law depends on $H_s$ only through the declared state $X_s$ and action (otherwise retain the full history as a finite state).
Proposition (Finite sequential intervention formula): Under these assumptions, for a state-based regime $g$, set $K_s^g(x,y)=\Pr(X_{s+1}=y\mid X_s=x,A_s=g_s(x))$ and $\mu_0^g=\pi$, $\mu_{s+1}^g=\mu_s^gK_s^g$. The targets are identified as \[ \theta_h=\sum_x R(x)(\mu_h^a(x)-\mu_h^{a_0}(x)),\qquad \gamma_h=\sum_x\ell_h(x)(\mu_h^a(x)-\mu_h^{a_0}(x)). \] These quantities do not require the joint distribution of the two counterfactual histories for the same country.
Proof: Consistency equates the observed next state after the prescribed action with its corresponding potential state. Exchangeability equates the conditional potential transition with the observed conditional transition; positivity makes that conditional law observable. Starting at the common $\pi$, iterated conditioning gives the stated recursion at every date. The flag recursion gives $R(X_h)={\mathbf 1}\{\tau\leq h\}$. Integration of this indicator and log output gives the formulas. Linearity of expectation uses only the two marginal laws, so no cross-world coupling is required. ∎

A sovereign reform adopted in response to an unrecorded impending crisis need not satisfy this proposition. Conditioning merely on the current binary democracy indicator is particularly weak: output, constraints, common shocks and past erosion can predict both adoption and transitions.

Sharp bounds with uncertain political transitions

Suppose substantive validation restricts each next-state row to a nonempty polytope \[ \mathcal K_s^z(x)=\{k\geq0:\boldsymbol1^\mathsf T k=1, B_s^z(x)k\leq b_s^z(x)\},\quad z\in\{a,a_0\}. \] Zero entries enforce the flag transition and feasibility. A useful example bounds the probability of institutional loss conditional on recession and current constraints; bounding a row is stronger than bounding an average observed hazard. Our rectangular compatibility assumption is that every combination of these rows, across states, dates and the two regimes, is compatible with the observed law and admitted structural restrictions. Each selected kernel incorporates political best responses and an equilibrium selector. We do not infer such compatibility from moment intervals alone. Shared structural parameters can invalidate it.

For each $z$ introduce occupancy $u_s^z(x)$ and flow $f_s^z(x,y)$ satisfying \begin{align} u_0^z&=\pi,& f_s^z(x,y)&\geq0,\nonumber\\ \sum_y f_s^z(x,y)&=u_s^z(x),& u_{s+1}^z(y)&=\sum_x f_s^z(x,y),\label{wsref-0-66-6c-6f-77}\\ B_s^z(x)f_s^z(x,\cdot)&\leq b_s^z(x)u_s^z(x).\nonumber \end{align} Let $\mathcal F$ be the set of all these flows, and define the linear map \[ T(f)=\left(\sum_xR(x)[u_h^a(x)-u_h^{a_0}(x)], \sum_x\ell_h(x)[u_h^a(x)-u_h^{a_0}(x)]\right). \]

Theorem (Joint, attainable identification region): Under rectangular compatibility the exact identified set for $(\theta_h,\gamma_h)$ is the compact convex polytope $T(\mathcal F)$. For any real weights $(v_1,v_2)$ its sharp support value is obtained by maximizing $v_1T_1(f)+v_2T_2(f)$ subject to (flow). In particular, separate coordinate intervals need not describe the joint identified set.
Proof: Every admitted kernel produces flows $f_s^z(x,y)=u_s^z(x)K_s^z(x,y)$ by multiplying its row inequalities by $u_s^z(x)$. Hence its target lies in the claimed image. Conversely, from any feasible flow, set $K_s^z(x,y)=f_s^z(x,y)/u_s^z(x)$ whenever the denominator is positive. Its row inequalities and row sum put it in $\mathcal K_s^z(x)$. At a zero-occupancy state all outgoing flows vanish; choose any row in the nonempty permitted polytope. Induction gives the original occupancies under these kernels. Rectangular compatibility admits all the constructed rows simultaneously, in both regimes, so the constructed target is attainable. The constraints define a closed bounded polytope since every occupancy and flow lies in $[0,1]$; it is nonempty by choosing any permitted rows. A linear image is a compact convex polytope, and a continuous linear functional attains its maximum there. ∎

For a single terminal state with $R=1$ exactly when $\ell_h=L$, the one-regime loss probability and output mean obey ${\mathbb{E}}\ell_h=U-(U-L){\mathbb{E}} R$ if all other states have output $U$. Their difference targets therefore obey $\gamma_h=-(U-L)\theta_h$. A rectangle of coordinate bounds would admit impossible combinations even in this elementary example.

An explicit robustness radius

Proposition (Accumulation of transition misspecification): Let $K_s^z$ and $\widetilde K_s^z$ be two candidate kernel sequences with the same initial law. If every corresponding row has total variation distance at most $\epsilon_s^z$, then terminal first-loss probabilities differ by at most $d_z=\min\{1,\sum_{s<h}\epsilon_s^z\}$, and terminal expected log outputs differ by at most $(U-L)d_z$. The two-regime target errors are consequently bounded by $d_a+d_{a_0}$ and $(U-L)(d_a+d_{a_0})$, respectively.
Proof: For distributions $p,q$ and stochastic kernels $K,\widetilde K$, \[ \|pK-q\widetilde K\|_{\rm TV} \leq\|(p-q)K\|_{\rm TV}+\|q(K-\widetilde K)\|_{\rm TV} \leq\|p-q\|_{\rm TV}+\sup_x\|K(x,\cdot)-\widetilde K(x,\cdot)\|_{\rm TV}. \] The contraction inequality follows by testing against functions in $[0,1]$; convex averaging gives the second bound. Induction and the unit upper bound on total variation yield $d_z$. Integrating an indicator changes its mean by at most total variation. Rescale $\ell_h$ from $[L,U]$ to $[0,1]$ for the output bound; when $L=U$ it is zero directly. The triangle inequality gives both difference bounds. ∎

Remaining substantive gap

The result supplies an exact joint optimization once validated kernel restrictions are available. It does not derive those restrictions from elite and citizen primitives. A promising next step is to impose a common structural political-incentive parameter across rows; the feasible flow set must then retain that linkage, rather than claiming rectangular sharpness. Persistent erosion can be recorded by adding a duration counter to the state. Measurement-error restrictions can likewise be added as latent-state constraints, but observed democracy classifications alone do not identify them. Sampling uncertainty needs simultaneous coverage for transition restrictions, especially for rare transitions; none of the population identification statements is a confidence claim. The full country-level equilibrium and empirical research family remains unresolved.

Completion Estimate

40%

This is a post hoc, heuristic estimate for the full catalogue target. The attempt proves finite-state sequential identification, sharp joint transition and output bounds under rectangular kernel compatibility, and robustness to transition errors. Full country-level intervention effects still require justified political primitives, shared structural restrictions, continuous-output measurement, and empirical transition validation.

References

Go to contribution →