{"id":"b3e678e0-9559-4ecf-bdfa-6303db2a56d7","arxiv_id":"2607.15119","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Vote shares in EU and French elections are shown to follow a Rayleigh-Jeans thermal distribution, with the Gini coefficient set by a fitted energy parameter.","lead":"This paper applies a statistical-mechanics model of energy sharing to election results and claims that party vote shares in EU and French elections follow a Rayleigh-Jeans thermal distribution. The work suggests that vote inequality across parties behaves like wealth inequality, with a few large parties and many small ones.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'describes very well' claim rests on unquantified visual fits: ε is calibrated to the Gini coefficient, Np is small, and RJS/RJE selection is post hoc.","rationale":"The reader's weakest assumption concerns the absence of a microfoundation for the oscillator/thermalization analogy. This is a legitimate concern about the explanatory framing, but it is not the most load-bearing point for the paper's central descriptive claim: even if the analogy is granted, the empirical evidence that the RJ model 'describes very well' is not quantitatively established. The most decisive weakness is statistical: the single model parameter is fitted to the Gini coefficient, which is itself the primary global summary of the Lorenz curve, and the remaining agreement is judged visually. Small Np further inflates the apparent plausibility of smooth curves. The proposed synthetic-election test would settle whether the observed agreement is within sampling noise of the fitted model. Since the reader's verdict already conditions on the need for stronger statistical validation, my concern does not change the verdict, so UNCHANGED is appropriate. I partially agree with the reader because they also mention lack of statistical tests, but I would not make the missing microfoundation the central condition for acceptance; the descriptive claim can be evaluated independently of it.","tokens_in":13779,"tokens_out":7801,"duration_ms":91525,"concrete_test":"For each election in Table 1, generate 10,000 synthetic elections by sampling Np vote shares from the fitted RJS (or RJE) density with the reported ε (and a) from Table 1/SupMat, normalizing to sum 1. For each synthetic election compute the same geometric Lorenz-curve distance D used in the SupMat RJE fitting, and compute the empirical Gini. Compare the real election's D to the synthetic distribution of D: if the observed D exceeds the 95th percentile for more than a small fraction of the elections, the 'describes very well' claim fails at conventional significance; if it lies inside the band for most elections, the concern is resolved and the visual agreement is consistent with finite-N sampling from the RJ model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the TTV/RJ model 'describes very well' the distribution of vote shares. The evidence for this is visual agreement of Lorenz/Pareto curves, but the model's parameter ε is fixed by matching exactly the real-data Gini coefficient (RJC section; Table 1). Thus the first moment of the Lorenz curve is matched by construction, and the residual shape agreement is never quantified. For the RJE cases, a second parameter a is chosen post hoc to minimize the geometric Lorenz-curve distance (SupMat), again without any null distribution. With Np between 6 and 41, smooth one-parameter curves can plausibly look acceptable by eye even when the underlying distribution is not the RJ model. Consequently, the paper's core empirical statement is not supported by any statistical test, and the current 'good description' could be an artifact of small-sample visual tolerance plus fitting the dominant summary statistic.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a 'thermodynamic theory of voting' (TTV) in which party vote shares are modeled as occupation probabilities rho_m = T/(E_m - mu), the Rayleigh-Jeans (RJ) distribution, for a classical oscillator system with conserved total energy and norm. The resulting Lorenz and Pareto curves are compared with vote shares from seven EU elections in Germany and France (1994-2024), the 2024 EU elections in the ten largest EU member states, and eleven first-round French presidential elections (1965-2022). The single model parameter epsilon (rescaled energy) is set by matching the empirical Gini coefficient; in four EU 2024 cases an additional spectral parameter a is fitted by minimizing the Lorenz-curve distance. The paper reports visual agreement and concludes that the RJ thermalization framework 'describes very well' the statistical distribution of votes.","tokens_in":14027,"tokens_out":7313,"duration_ms":81926,"significance":"If substantiated, the paper would offer a parsimonious statistical-physics description of a robust regularity in multiparty elections, connecting voting outcomes to constraint-driven condensation and providing compact formulas for Lorenz/Pareto curves. The empirical coverage is nontrivial, and the use of public election data together with explicit Lorenz/Pareto construction is a strength. However, the current evidence is essentially qualitative: epsilon is calibrated to reproduce the Gini index, the RJS versus RJE choice is made after seeing the data, and no goodness-of-fit statistic or null model is provided. The significance is therefore conditional on quantitative validation of the distributional claim.","major_comments":[{"comment":"The central claim that the TTV/RJ model 'describes very well' the vote distribution is not supported by any quantitative goodness-of-fit measure. The text states that epsilon is determined such that the Gini coefficient of the RJ model coincides with the real-data value. Hence equality of G is enforced by construction, and the only evidence for the distributional claim is visual agreement of the remaining Lorenz/Pareto shape. No KS/AD statistic, residual distance, confidence band, or null model is reported for any of the 28 elections. With N_p between 13 and 41 (and 6-16 for the French presidential elections), smooth one-parameter curves may look acceptable even when the data are not RJ-distributed. A Monte Carlo or bootstrap test of the fitted RJ distribution at the observed N_p is required before the central claim can be assessed.","section":"RJC and Lorenz, Pareto curves construction; Table 1"},{"comment":"For PL, RO, BE, and SE the paper replaces the RJS model with the RJE model, and the additional spectral parameter a is fitted by minimizing the geometric Lorenz-curve distance to the real data. This is a second free parameter chosen post hoc for exactly those countries where RJS shows visible deviations. The paper provides no model-selection criterion (e.g., AIC/BIC or cross-validation) and no null distribution for the minimized distance. Given N_p = 13-22 in these cases, the extra parameter may simply absorb sampling fluctuations; the claimed improvement over RJS is therefore unquantified.","section":"Fig. 3 and Supplementary Material"},{"comment":"The theoretical curves are computed in the continuous limit N -> infinity, while the empirical Lorenz/Pareto curves are step functions based on a small number of parties/candidates. The statement that finite-N model curves with N = 10^4 are graphically identical addresses only the model's own discretization, not the finite-sample nature of the data. For the smallest samples (N_p = 6 in the 1965 French presidential election; N_p = 13-15 for IT and RO in 2024), visual comparison with a smooth continuous curve cannot establish distributional agreement. The comparison should be performed at the observed N_p, for example by simulating N_p draws from the fitted RJ distribution and constructing confidence bands for the Lorenz/Pareto curves.","section":"RJC and Lorenz, Pareto curves construction; Results"}],"minor_comments":[{"comment":"The manuscript contains several typos and formatting remnants: '*** Missing PACS ***' in the header, 'woth' in the RJC section, 'thenmalization' in the Introduction, 'Elsvier' in reference [11], and 'K,M,' in reference [13]. A full copyedit is needed.","section":"Throughout"},{"comment":"The x-axis label 'country (arb. unit.)' in the right panel is unhelpful. The ordering of the ten countries should be stated explicitly, for example by listing the ISO codes in the caption.","section":"Fig. 4"},{"comment":"The paper uses 'Pareto curve' for a general cumulative distribution function even when no power-law tail is being fitted or claimed. Please state this convention explicitly at first use, otherwise readers may expect a Pareto exponent analysis.","section":"Pareto curve definition"},{"comment":"References [1]-[3] are Wikipedia and EU web pages accessed in 2026. For reproducibility, the authors should archive the downloaded datasets or provide a stable data repository.","section":"Data references"}],"recommendation":"major_revision","confidential_remarks":"The paper has a clear construction and a broad empirical coverage, but its central assertion is currently supported only by visual fits with parameters matched to the data. This is fixable: the authors should add a quantitative fit diagnostic, a bootstrap/Monte Carlo null, and a model-selection protocol for RJS versus RJE. I do not see a fundamental internal inconsistency; the physical analogy is a phenomenological assumption rather than a derived mechanism, which is acceptable if stated carefully. The word 'explains' in the Discussion should be moderated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is a clean, honest application of the authors' Rayleigh-Jeans thermalization model to a new dataset (election vote shares), and the figures look nice. But the paper's central claim—that the model 'describes very well' the distribution of votes—rests on visual agreement alone, and the one free parameter ε is chosen by matching the Gini coefficient, so the first summary statistic of the Lorenz curve is matched by construction. With 13 to 41 data points, that is not enough to establish the fit.\n\nWhat's genuinely new is the domain shift: none of the prior work applied RJ condensation to vote shares. The paper assembles a consistent dataset, gives a clear parameter table, and is transparent about the fitting procedure—ε from the Gini coefficient, and for four countries an additional spectral parameter a chosen to minimize a geometric Lorenz distance. That transparency is good. The underlying RJ distribution is standard statistical mechanics, so there is no hidden circularity in the thermodynamics itself.\n\nWhere it softens: the evidence for 'good description' is weak. No goodness-of-fit test, no comparison against alternative one-parameter distributions, no bootstrap or uncertainty quantification. With Np around 13–41, many smooth curves will look acceptable by eye, and the RJS vs RJE choice is post hoc. The microfoundation is also assumed rather than derived—the authors say couplings and nonlinear interactions between electors 'are assumed,' which is fair, but it makes the theory an analogy, not a mechanism. They do note that the model doesn't predict who wins, which I count as honesty, but it underlines that this is descriptive.\n\nWho is this for: sociophysics and econophysics readers interested in statistical regularities in voting, and maybe political scientists who want to see how far such analogies go. It deserves a serious referee because the idea is plausible and the data are public, but I would not accept the current claim as stated.\n\nMy recommendation: send it to peer review, but ask for substantial revision—add a quantitative goodness-of-fit test with a null distribution, report uncertainty in ε, and retract the phrase 'describes very well' unless the numbers back it up.","headline":"Applying their RJ thermalization model to vote shares is a genuine extension, but the main fit claim is only visual and partly by construction.","tokens_in":14472,"tokens_out":2373,"would_cite":false,"duration_ms":26734,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Multi-party vote shares follow the same thermal-equilibrium curve family that fits wealth inequality; the paper shows this for EU elections 1994–2024 and French presidential first rounds since 1965.","keywords":["Rayleigh-Jeans distribution","voting theory","Lorenz curve","Pareto curve","Gini coefficient","condensation","EU elections","statistical mechanics of society"],"falsifier":"Use an out-of-sample multi-party election, fix ε from its Gini coefficient, and compare the predicted RJ Lorenz and Pareto curves with the measured ones; the theory fails if the curves deviate beyond the scatter seen in the paper's figures. A sharper falsifier is a measured cumulative vote distribution with a genuine power-law tail over more than a decade of vote share, since the RJ curve has a different, non-power-law shape.","tokens_in":13675,"feed_emoji":"🗳️","tokens_out":9214,"duration_ms":91609,"temperature":0.7,"pith_summary":"This paper claims that the uneven distribution of votes among parties and candidates is not a collection of accidents but a statistical regularity with the same mathematical shape as thermal equilibrium. The authors introduce the Thermodynamic Theory of Voting (TTV), in which each party's vote share plays the role of the energy of a mode in a system of coupled nonlinear oscillators, and interactions among electors thermalize the system into a Rayleigh-Jeans distribution. A single parameter — the rescaled energy, chosen so the model's Gini coefficient equals the real election's Gini — then fixes the entire Lorenz and Pareto curves, and those curves consistently match German and French EU elections over 30 years, the 2024 EU elections of ten EU countries, and first-round French presidential elections since 1965. If this is right, extreme vote inequality (many parties near zero, a few top parties taking most votes) is the expected condensed state of a constrained statistical system, not an anomaly. The theory constrains the shape of the field; it does not say which party or candidate will win.","feed_headline":"One physics law fits 30 years of EU vote results","feed_subtitle":"The Rayleigh-Jeans curve that models wealth inequality also reproduces real election Lorenz and Pareto curves.","key_machinery":"The central object is the Rayleigh-Jeans (RJ) distribution ρ_m = T/(E_m − μ), applied to vote shares E_m = V_m. It is the equilibrium occupation of modes in a classical system conserving total energy E = Σ E_m ρ_m and total probability Σρ_m = 1; T and μ are fixed by those two constraints. The paper tests this law with two model spectra — uniform ('RJS') and exponentially modified ('RJE') — and builds Lorenz and Pareto curves from the RJ probabilities. The single free parameter, rescaled energy ε, is set by matching Gini coefficients; the whole curve then follows. Low ε yields RJ condensation: most probability piles onto the lowest states, matching the observed pattern of many tiny parties an","core_discovery":"Vote distributions are thermal: when enough parties or candidates compete, vote fractions follow the Rayleigh-Jeans law ρ_m = T/(E_m − μ), the equilibrium of a classical oscillator system with conserved total energy and norm. Fitting one parameter, the rescaled energy ε set by the Gini coefficient, reproduces the full Lorenz and Pareto curves of seven German and seven French EU elections (1994–2024), eight other 2024 EU elections, and eleven French presidential first rounds (1965–2022). The fits imply vote condensation: in 2024 EU elections the bottom half of parties collect about 1% of votes while the top 10% take 57–68%. The same curves describe 6–16 candidate presidential fields, with low","pith_inferences":["If the paper is right, vote inequality and wealth inequality are two instances of the same constraint-driven condensation; one could test this by checking whether countries with similar wealth Gini coefficients also have similar RJ parameters in their elections.","The paper's practice of fixing ε by Gini means the full Lorenz curve is a falsifiable prediction; one can test it on future elections or on other multi-candidate contests (e.g., primary elections, party leadership races) before any fit.","The decrease of Gini with increasing number of candidates is a quantitative prediction: for small N_p the condensation regime should weaken; comparing many elections with N_p from 6 to 40 could map the boundary of the RJ regime."],"forward_implications":["For EU elections, the RJ model implies that high Gini values (~0.8) are the statistical norm, with vote distributions stable over 30 years; deviations (e.g., France 1994/1999, G≈0.6) coincide with fewer parties.","The near-linear relation G ≈ 1 − 2ε for ε ≤ 0.2 means a single inequality number directly fixes the temperature-like parameter of the election's vote distribution.","The TTV description extends beyond parties to individual candidates whenever the field is not too small, so the same curves describe first-round presidential elections with 6–16 candidates.","The theory sets a general statistical constraint on typical vote distributions, not a prediction of winners; any specific outcome is outside its scope."],"fun_headline_variants":["Vote counts obey physics law in EU elections","Same law explains wealth gap and EU votes","EU vote shares thermalize like heated particles","Rayleigh-Jeans law fits three decades of EU votes","Top 10% of EU parties take 68% of votes"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing assumption is that voter interactions actually thermalize vote shares like a classical oscillator system, so the conserved-energy picture — not a specific model of voter psychology — determines the vote distribution.","fun_headline_variants_meta":{"raw":{"variants":["Vote counts obey physics law in EU elections","Same law explains wealth gap and EU votes","EU vote shares thermalize like heated particles","Rayleigh-Jeans law fits three decades of EU votes","Top 10% of EU parties take 68% of votes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000626,"raw_usage":{"total_tokens":2686,"prompt_tokens":652,"completion_tokens":2034,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":1972}},"tokens_in":396,"tokens_out":2034,"duration_ms":18016,"temperature":1.0,"reasoning_tokens":1972,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T00:04:06.172208+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Use an out-of-sample multi-party election, fix ε from its Gini coefficient, and compare the predicted RJ Lorenz and Pareto curves with the measured ones; the theory fails if the curves deviate beyond the scatter seen in the paper's figures. A sharper falsifier is a measured cumulative vote distribution with a genuine power-law tail over more than a decade of vote share, since the RJ curve has a different, non-power-law shape.","supporting_citations":[],"review_version":1}