1% ============================================================================2% Related work and contribution3% ============================================================================4\section{Related work and contribution}5\label{sec:literature}67Every reference in this section was checked against Crossref by DOI or bibliographic query on 1~November 20268(\texttt{research/prior\_art.csv}, \texttt{research/bib\_overrides.csv}); the abstracts and key results of the9eighteen works closest to our hypothesis were read from primary or repository sources in a separate verification pass10(\texttt{docs/novelty\_search\_report.md}). Works that we know only through reviews or citing articles are flagged in11footnotes and are cited for their general thesis only.1213\subsection{Scalar stress and information-processing hierarchies}14\label{sec:lit-scalar}1516The oldest form of our idea is \citeauthor{johnson1982}'s \emph{scalar stress}:\footnote{Known to us through17library records and through \citet{alberti2014}; the chapter itself was not read.} as the number of units that must be18coordinated grows, the communication and decision load grows faster than linearly, and groups respond either by fission19or by adding a hierarchical level, with a modal span of about six units per level \citep{johnson1982}. \citet{alberti2014}20estimate a related fission threshold on Hutterite colonies. \citet{wright1975} define the state as a specialized21administrative hierarchy with at least three decision-making levels and reject population growth and exchange as single22sufficient causes of its emergence in Susiana; \citet{johnson1978} extends the information-processing reading of23decision-making organizations. \citet{carneiro1967} documents the log-log regularity between population and the number24of organizational traits, \citet{naroll1956} proposes an early index of social development, \citet{feinman2011} reviews the25size--complexity relation, and \citet{flannery1972} frames institutional evolution as a change in information-processing26architecture. None of this strand builds a demand/capacity \emph{ratio}, and none tests it against instability.2728\subsection{Declining returns and collapse}29\label{sec:lit-collapse}3031\citet{tainter1988} argues that complexity is a problem-solving investment with declining marginal returns, so that32collapse (simplification) eventually becomes rational and the society loses its capacity to absorb shocks.\footnote{Content33known from reviews and secondary sources; the book was not re-read for this paper.} \citet{middleton2017},34\citet{butzer2012} and \citet{cline2014} review the collapse literature and warn against mono-causal accounts.35Structural-demographic theory \citep{turchin2003,turchin2009,turchin2023} models secular cycles driven by population36pressure, elite overproduction and fiscal distress; \citet{goldstone1991} is its historical origin. \citet{sng2014} is37the closest economic precedent: in late-imperial China the cost of transmitting information over distance creates a38principal--agent problem whose severity rises with territorial size, so that fiscal capacity decays during the39prosperous eighteenth century---a demand (territory) versus capacity (monitoring) argument, tested within a single40polity with prefecture-level data. Using the Seshat CrisisDB, \citet{hoyer2024} show that crisis consequences are41heterogeneous, largely uncorrelated with one another, and not well predicted by territory, religion or administrative42complexity taken one at a time (``no typical societal crisis'').4344\subsection{Seshat-based quantitative history}45\label{sec:lit-seshat}4647The Seshat Global History Databank \citep{turchin2015,francois2016} underlies a growing quantitative literature.48\citet{turchin2018} found that nine complexity characteristics (population, territory, capital population, hierarchical49levels, government, infrastructure, writing, texts, money) load on a single principal component explaining about three50quarters of the variance across 414 polity-centuries. \citet{shin2020} decomposed the remaining variance and identified a51second axis opposing \emph{scale} characteristics to \emph{information-processing} characteristics, defined a ``scale52threshold'' and an ``information threshold'' along the first component, and wrote---explicitly as a conjecture left for53future research---that ``mismatch among aspects of social development should be considered as a generic cause of54instability''. \citet{miranda2020} clustered Seshat polities into two super-types with an unstable transitional cluster55in which polities dwell only a few centuries. \citet{turchin2022} compared more than 100{,}000 predictor combinations for56the \emph{growth} of complexity and found agricultural productivity interacting with military technology best supported;57they did not model instability. Related Seshat-based work includes \citet{currie2010} on the rise and fall of political58complexity in Austronesia, \citet{sheehan2018} on the coevolution of intensive agriculture and hierarchy,59\citet{kohler2017,kohler2018} on long-run inequality, \citet{peregrine2017,peregrine2019} on resilience to disasters,60and \citet{hoyer2023} on the long-run factors that shape responses to changing climate.61\citet{whitehouse2019} was retracted \citep{whitehouse2021retraction} after \citet{beheim2021} showed that recoding62missing moralizing-god data as ``absent'' drove the result; the Seshat team's ``retake'' \citep{whitehouse2023} reports63qualitatively similar conclusions with corrected coding, and \citet{slingerland2020} record the historians' objections.64The methodological lesson---never treat unknown as absent, never forward-fill, report results on extant data---is65written into our pre-registration.6667\subsection{State capacity and its informational foundations}68\label{sec:lit-capacity}6970\citet{mann1984,mann1986} distinguish despotic from \emph{infrastructural} power; \citet{soifer2008} and71\citet{hanson2021} discuss its measurement, the latter with a Bayesian latent model over 21 indicators for 1960--2015.72\citet{scott1998} describes \emph{legibility}---standardized names, cadastres, censuses, weights and measures---as the73state's way of making society readable, and \citet{scott2017} traces it to the earliest states. \citet{lee2017}74operationalize legibility through age-heaping in censuses and show that it predicts tax contributions;75\citet{brambor2020} build an ``information capacity'' index (census, statistical agency, yearbooks, civil and population76registers) for 85 polities since 1789; \citet{darcy2017} use cadastral records. \citet{besley2009,besley2011},77\citet{dincecco2017}, \citet{johnson2017}, \citet{mayshar2022}, \citet{borcan2018} and \citet{acemoglu2001} study the78origins and consequences of fiscal and legal capacity; \citet{gennaioli2015} link capacity to military competition, as do79\citet{tilly1992}, \citet{mcneill1982} and \citet{hoffman2015}. \citet{stasavage2010} shows that geographic scale limited80representative assemblies in Europe---a direct instance of distance-driven coordination cost---and81\citet{stasavage2020,ahmed2020} and \citet{blaydes2013} trace the long-run divergence of political institutions;82\citet{ko2018}, \citet{fernandezvillaverde2023}, \citet{scheidel2019} and \citet{dincecco2018} compare China and Europe.83Classic comparative treatments of empires and bureaucracy are \citet{weber1922}, \citet{eisenstadt1963},84\citet{trigger2003}, \citet{yoffee2005}, \citet{blanton2008}, \citet{fukuyama2011} and \citet{graeber2021}. None of these85operationalize a \emph{ratio} of coordination demand to capacity, nor test it against instability across pre-modern86polities.8788\subsection{Transaction costs, codes and communication in organizations}89\label{sec:lit-codes}9091\citet{coase1937}, \citet{williamson1985} and \citet{north1990,north1991} treat institutions as devices that economize on92transaction costs; \citet{hayek1945} on dispersed knowledge, \citet{simon1955} on bounded rationality and93\citet{simon1962} on near-decomposable hierarchies are the classic information-processing views. \citet{arrow1974}94introduced organizational \emph{codes};\footnote{Verified only through secondary citations.} \citet{cremer2007} model the95firm's language as a code of bounded size in which precise words go to frequent events---literally lossy compression that96trades scope against precision. \citet{marschak1972}, \citet{radner1993}, \citet{bolton1994}, \citet{vanzandt1999} and97\citet{garicano2000} model hierarchies as information-processing networks with explicit communication costs;98\citet{dessein2002} and \citet{dewatripont2005} study authority and modes of communication; \citet{crawford1982},99\citet{hagenbach2010}, \citet{galeotti2013} and \citet{calvoarmengol2015} analyze strategic communication on networks.100Mechanism-design results bound the \emph{message-space size} needed to implement allocations101\citep{mount1974,segal2007,nisan2006}, and communication complexity \citep{yao1979,kushilevitz1997} gives the102computer-science counterpart. The business-history literature documents the same economizing in practice:103\citet{chandler1977}, \citet{yates1989}, \citet{beniger1986} and, earlier, \citet{innis1950}. These results are the formal104basis for Section~\ref{sec:theory}, but they concern firms, markets or abstract protocols, not polities.105106\subsection{Coordination, common knowledge and conventions}107\label{sec:lit-coord}108109\citet{lewis1969} and \citet{schelling1960} define conventions and focal points; \citet{young1993},110\citet{kandori1993} and \citet{ellison1993} show how conventions are selected and how local interaction speeds111convergence; \citet{morris2000} on contagion and \citet{skyrms2010} on signalling complete the picture.112\citet{chwe2000,chwe2001} argues that coordination requires \emph{common knowledge} and that rituals and public113communication structures exist to manufacture it cheaply; \citet{rubinstein1989} shows that almost-common knowledge is114not enough. \citet{morris2002} and \citet{angeletos2007} quantify the social value of public signals, and \citet{dewan2008}115model leaders as coordination devices. \citet{ostrom1990,ostrom2010} and \citet{olson1965} are the collective-action116anchors; \citet{boyd1985,boyd1988}, \citet{henrich2004}, \citet{richerson2016} and \citet{turchin2016} supply the117cultural-evolution mechanisms for norm maintenance in large groups, \citet{gavrilets2008} and \citet{turchin2009gavrilets}118the evolutionary dynamics of hierarchy, and \citet{stanish2017} the role of ritual in stateless societies;119\citet{greif1993,greif2006} and \citet{milgrom1990} are the historical cases of reputation- and record-based enforcement.120121\subsection{Scaling and information theory}122\label{sec:lit-scaling}123124\citet{bettencourt2007,bettencourt2013}, \citet{ortman2014} and \citet{lobo2020} show superlinear scaling of125socioeconomic outputs with settlement population, ancient and modern; \citet{dunbar1993} and \citet{hamilton2007}126document scaling constraints in small-scale societies. Scaling theory implies that interaction demand grows faster than127population, but it assumes that infrastructure and institutions co-scale; it predicts benefits, not instability.128\citet{dedeo2017} is the closest explicit statement that transitions in political order are driven by129information-processing innovations and that coarse-graining (lossy compression) is the central mechanism;130\citet{wolpert2017} surveys state-space compression, and \citet{marzen2017} give a rate-distortion model with a131high-fidelity regime (cost linear in environmental complexity) and a low-fidelity regime. \citet{flack2012,flack2017},132\citet{daniels2017} and \citet{krakauer2020} develop coarse-graining and collective computation in animal societies;133\citet{hidalgo2015} popularizes the information-growth view of economies. \citet{shannon1948} and \citet{cover2006} are134cited only for the definitions used in the simulations, where messages and distributions are actually defined.135To our knowledge no published work uses the name ``Social Compression Hypothesis'' (search of 1~November 2026), but the136idea has the clear ancestors listed above, which this paper credits.137138\subsection{Instability forecasting and methods}139\label{sec:lit-methods}140141\citet{goldstone2010} forecast instability onsets for 1955--2003 with regime type as the dominant predictor;142\citet{fearon2003} and \citet{hegre2001} are the civil-war baselines, and \citet{cederman2003} models the size143distribution of wars. For binary time-series--cross-section data we follow \citet{beck1998} and \citet{carter2010};144for event history, \citet{cox1972} and \citet{boxsteffensmeier2004}; for logistic calibration, \citet{hosmer2013}; for145clustered and spatial inference, \citet{cameron2015} and \citet{conley1999}; for sensitivity to unobservables,146\citet{oster2019} and \citet{cinelli2020}; for Galton's problem, \citet{mace1994} and \citet{dow2007}; for latent147measurement, \citet{bollen1989}, \citet{samejima1969}, \citet{treier2008} and \citet{pemstein2018}; for agent-based148modelling, \citet{epstein1996}, \citet{axtell2002}, \citet{macy2002}, \citet{watts1998} and \citet{jackson2008}; and for149pre-registration and the statistical crisis, \citet{nosek2018} and \citet{gelman2014}. Data sources are documented by150\citet{seshat2026}, \citet{kirby2016} and \citet{murdock1967}, \citet{bolt2024}, \citet{coppedge2020} and151\citet{vdem2026}, and \citet{glottolog2026}.152153\subsection{Novelty matrix}154\label{sec:novelty}155156Table~\ref{tab:novelty} summarizes which elements of the Social Compression Hypothesis already exist in the closest157works. The ranking by overlap is: \citet{shin2020} (a verbal mismatch--instability conjecture, declared untested);158\citet{johnson1982} (scalar stress); \citet{sng2014} (distance-driven information cost, one polity); \citet{tainter1988};159the codes-as-compression micro-foundation of \citet{arrow1974}, \citet{cremer2007} and \citet{dedeo2017};160\citet{miranda2020}; the CrisisDB results of \citet{hoyer2024}; and the settlement-scaling literature as a foil.161162\begin{sidewaystable}163\centering164\footnotesize165\caption{Novelty matrix: elements of the Social Compression Hypothesis in the closest prior work}166\label{tab:novelty}167\begin{tabular}{@{}lllllllll@{}}168\toprule169Work & Demand/scale & Capacity & Mismatch & Instability & Cross-polity & Comm.\ cost & Simul. \\170\midrule171\citet{johnson1982} & group size & hierarchy levels & implicit & fission (ethnogr.) & -- & verbal & -- \\172\citet{wright1975} & settlement hier. & admin.\ levels & -- & -- & one region & -- & -- \\173\citet{tainter1988} & -- & complexity & -- & collapse (cases) & -- & verbal & -- \\174\citet{sng2014} & territory, distance & monitoring & implicit & fiscal decay & one polity & \checkmark & -- \\175\citet{turchin2018} & pop., territory & govt., writing & -- & -- & \checkmark & -- & -- \\176\citet{shin2020} & PC2 scale side & PC2 info.\ side & verbal & conjectured & \checkmark & verbal & -- \\177\citet{miranda2020} & cluster features & cluster features & -- & dwell time & \checkmark & -- & -- \\178\citet{turchin2022} & complexity (outcome) & -- & -- & -- & \checkmark & -- & -- \\179\citet{hoyer2023,hoyer2024} & territory (covariate) & admin.\ (covariate) & -- & CrisisDB & \checkmark & -- & -- \\180\citet{lee2017,brambor2020} & -- & legibility, info.\ capacity & -- & -- & modern & -- & -- \\181\citet{hanson2021} & -- & latent capacity & -- & -- & modern & -- & -- \\182\citet{goldstone2010} & pop.\ threshold & regime type & -- & onsets & modern & -- & -- \\183\citet{cremer2007,arrow1974} & org.\ scope & code size & trade-off & -- & -- & \checkmark & -- \\184\citet{chwe2000,chwe2001} & network & common knowledge & -- & -- & -- & \checkmark & -- \\185\citet{bettencourt2013,ortman2014} & settlement pop. & infrastructure & -- & -- & \checkmark & interaction & -- \\186\citet{dedeo2017,marzen2017} & -- & coarse-graining & -- & -- & -- & rate--distortion & -- \\187\midrule188This paper & $\log$ pop., territory, largest settlement, hierarchy & 4 disjoint blocks (mean, IRT, PC) & $P=z(\log D)-z(K)$ & CrisisDB severity; hazard; modern onsets & Seshat, EA, V-Dem & propositions & factorial ABM \\189\bottomrule190\end{tabular}191\end{sidewaystable}192193\subsection{Precise contributions and non-claims}194\label{sec:contributions}195196\begin{enumerate}197\item \emph{Empirical.} The first pre-registered test of the scale/information mismatch conjecture of \citet{shin2020}198 against an independently coded instability outcome (the Seshat CrisisDB), with a transparent and a latent measurement199 model, propagated uncertainty, and out-of-sample validation, replicated in form on a modern country-year panel. The200 result is negative in the pre-registered sense: the mismatch index $P$ is associated with crisis severity and with201 modern instability onsets, but it does not predict either outcome beyond its own components (scale in the historical202 panel, capacity in the modern panel), fails the out-of-sample rule against those baselines, and performs no better than203 a placebo index in the historical panel. What survives is that scale predicts crisis severity among crisis cases and204 that capacity predicts fewer onsets in modern states---neither of which is new, and neither of which is the Social205 Compression Hypothesis.206\item \emph{Theoretical.} Formal propositions stating when a shared code or record system lowers the expected207 communication cost per coordinated task as a function of scale, when fixed institutional overhead makes it worse, by208 how much a stale code raises cost after a shock, and when hierarchies saturate; each is stated as a theorem with proof209 or as a conjecture, never as an empirical fact.210\item \emph{Simulated.} A factorial agent-based comparison of five imposed institutional regimes with directly measured211 messages, bits, success rates and overhead, including an explicit search for regimes in which institutions reduce212 efficiency or add fragility (set-up costs that are never repaid, saturated hierarchies, weakly concentrated plan spaces).213\item \emph{Not claimed.} No causal identification in either panel; no literal Shannon compression of historical214 institutions; no universal theory. The name ``Social Compression Hypothesis'' has no prior academic use that we could215 find, but the idea has clear ancestors (Table~\ref{tab:novelty}), which this paper credits, and the data do not support216 the universal framing under which the project was mandated.217\end{enumerate}218