# Right Diagnosis, Wrong Prescriptions? Testing Ten of Karl Marx's Hypotheses Against Twenty-First-Century Data

**Authors:** Jason Miklian
**Published in:** *Under Review*, 2026
**Canonical URL:** https://miklian.org/papers/right-diagnosis-wrong-prescriptions-testing-marx-hypotheses-21st-century-data

## Abstract

Karl Marx is one of the most cited, most debated, and least empirically tested social theorists of the modern era. His critics dismiss him as a failed prophet; his defenders treat his work as unfalsifiable critique. Neither camp has subjected his core claims to systematic modern econometric testing. This paper therefore translates ten of Marx's most influential hypotheses into falsifiable specifications and estimates them against 21st-century data from the Penn World Tables, World Development Indicators, SWIID, ACLED, V-Dem, MARPOR, and the American Time Use Survey, among other sources. The results divide cleanly: Marx's economic diagnostics (exploitation rising, labor's share falling, unemployment disciplining wages, automation generating alienation) hold up remarkably well. His political predictions (that material immiseration produces class consciousness, that the state functions as a direct instrument of capital, that structural crises generate revolutionary mobilization) do not. The analysis reveals how Marx was an astute diagnostician of capitalism's economic tendencies and a weaker theorist of its political consequences. This gap between conditions and consciousness suggests that institutional design, rather than historical inevitability, mediates the political consequences of economic exploitation. The implications are developed for development policy, with particular attention to platform labor governance, redistributive institutional design, and market concentration in emerging economies.

**Keywords:** Marxism, econometric testing, exploitation, class consciousness, inequality, causal inference, development economics, political economy, Global South, institutional design

---

## Full Text


## 1. Introduction

Karl Marx wanted to change the world. Two centuries later, the world has decided what to do with Marx in return: argue about him. Mainstream economics treats his work as a historical curiosity, a set of predictions so falsified by the 20th century's experiments in state socialism that engaging his analytical framework is unnecessary. Heterodox scholars, meanwhile, treat his categories as either self-evidently correct or so embedded in a holistic critique of capitalism that isolating individual claims for empirical testing is impossible. The result is a peculiar impasse. The most influential social theorist of the modern era has generated an enormous body of commentary yet a comparatively thin body of systematic empirical testing. This impasse teases a provocative question: does Marx still matter empirically?

This paper attempts to answer that question. Ten of Marx's core hypotheses, drawn from Capital Volumes I through III, the Grundrisse, and the Eighteenth Brumaire, are translated into falsifiable econometric specifications and estimated against contemporary data. The analysis employs instrumental variables, system GMM, regression discontinuity designs, difference-in-differences, triple-difference estimation, Bayesian model averaging, and mediation analysis on data from the Penn World Tables, the World Development Indicators, SWIID, ACLED, V-Dem, MARPOR party manifesto data, VoteView Congressional records, the American Time Use Survey, and the World Values Survey. Each hypothesis is scored (see Appendix A), and aggregated into a composite measure of how well Marx's claims hold up against 21st-century evidence. The scores, across various specifications and weighting schemes, represent a respectable performance for a 19th century theorist. A clear pattern also emerges: Marx is systematically right about certain kinds of claims and systematically wrong about others, and the boundary between them illuminates a fundamental connection in the relationship between economic structures and political action.

Specifically, Marx’s economic diagnostics (the tendency of exploitation to rise, the decline of labor's share, the disciplinary function of unemployment, the alienating effects of automation) hold up remarkably well. His political predictions (that material immiseration generates class consciousness, that the state functions as a straightforward instrument of capital, that structural crises produce revolutionary mobilization) do not. Marx was an astute diagnostician of what capitalism does to workers. He was a weaker theorist of what workers do about it. These findings may have value for development economics in ways that extend beyond Marxist exegesis. Today’s global economy is hosting the conditions Marx described, including rising inequality, declining labor shares, precarious employment, and accelerating automation. If Marx was right that these conditions generate political mobilization for redistribution, then democratic correction may be imminent. If exploitative conditions persist without generating the political forces needed to address them, then institutional design bears a heavier burden. The results suggest the latter. The implications for platform labor governance in emerging economies, redistributive institutional design in middle-income countries, and competition policy in developing markets are thus developed in the discussion.

The paper proceeds as follows. The theoretical framework organizes ten hypotheses into three thematic clusters, then describes the data sources and econometric methods, including the explicit scoring algorithm and multiple testing corrections. The results are then presented, followed by discussion of the diagnostic–prognostic gap, development policy implications for the Global South, and the missing link between material conditions and political consciousness, before discussing conclusions and future research opportunities.


## 2. Theoretical Framework

This paper tests Marx's predictions and structural claims about capitalism's systematic tendencies as the mechanisms through which competition, accumulation, and class relations operate. These claims are distinctive to Marx, even if later thinkers have documented and expanded upon similar patterns independently. For example, capital's rising share of national income can be traced without invoking surplus value (Piketty 2014); declining labor compensation can be measured without dialectical materialism (Karabarbounis and Neiman 2014). What distinguishes this approach is the systematic side-by-side comparison: ten hypotheses scored and ranked, with the pattern of confirmed and disconfirmed claims revealing something about the scope and limits of structural explanation itself.

Marx's theoretical system is sprawling, internally contested, and deliberately holistic. Isolating individual hypotheses for econometric testing is, as his defenders may claim, a form of violence against his thought. This procedure is undertaken nonetheless, for two reasons. First, any theoretical framework that generates claims about the empirical world can be assessed against evidence from that world, regardless of the framework's internal logic. Second, treating Marx's claims as unfalsifiable renders them useless for the development policy questions they were designed to illuminate. This paper therefore tests direct versions of each claim. Marxist scholars reading Marx dialectically may reasonably score several hypotheses differently. But the advantage of mechanical clarity is precision: it allows one to ask whether the foundational structural mechanisms that ground Marx's entire political economy operate as specified today.

The analysis organizes ten hypotheses into three clusters, each capturing a distinct domain of Marxist prediction. Cluster A (Capital and Profit) addresses the internal dynamics of capitalist accumulation. Cluster B (Labor and Lived Experience) addresses how those dynamics manifest in workers' material conditions and subjective experience. Cluster C (Politics and Power) addresses the political consequences Marx expected to follow. Two additional hypotheses from the original theoretical inventory — the labor theory of value and the organic composition of capital as an independent prediction — were dropped as untestable or redundant with H1 and H2 respectively, yielding the ten hypotheses numbered sequentially below.


## 2.1 Cluster A: Capital and Profit

Marx's theory of capitalist accumulation rests on a set of interconnected claims about what happens to profits, exploitation, and market structure as capital deepens (Marx 1867/1990, Part VII; 1894/1981, Part III). The Monetary Expression of Labor Time (MELT) framework converts national accounts into labor-time equivalents (Cockshott and Cottrell 1997; Shaikh 2016), transforming GDP into socially necessary labor hours. A conventional labor share regression captures the empirical trend: the declining compensation of workers relative to capital. The Marxist framing adds a theorized mechanism linking the trend to competitive accumulation. In Capital Volume I, Marx argued that inter-capitalist competition compels firms to substitute constant capital (machinery, materials) for variable capital (e.g. wages), responding rationally to profit pressure (Marx 1867/1990). This substitution drives down the aggregate labor share not as a distributional accident but as the structural outcome of capital's drive to cheapen production. Standard distributional analysis observes the decline; Marx's framework provides the causal mechanism through which it occurs systematically across capitalist economies (Moseley 2016). Three hypotheses capture the core predictions.

H1: Tendency of the Rate of Profit to Fall.

In Capital Volume III, Chapters 13–15, Marx (1894/1981) argued that competitive pressure drives capitalists to substitute constant capital for variable capital, raising the organic composition of capital (OCC). Because only labor produces surplus value in Marx's framework, a rising OCC depresses the rate of profit. Marx explicitly acknowledged six counter-tendencies in Chapter 14 (intensified exploitation, depression of wages below their value, cheapening of constant capital, relative surplus population, foreign trade, and the increase of stock capital) but treated the downward tendency as dominant over the long run. The interpretation remains contested in Marxist political economy. One reading holds that Marx himself abandoned the law in later manuscripts (Heinrich 2012), which is disputed on textual and logical grounds (Kliman 2012; Roberts 2016). The modern empirical literature is similarly divided: cointegration methods applied to the postwar United States support a falling profit rate (Basu and Manolakos 2013); counter-tendencies appear to offset the decline in other specifications (Hartley 2020); and the US rate of profit traces a long-run fall from 1947 to 1977 followed by partial recovery (Moseley 1991). Both possibilities are tested using IV-2SLS with commodity price instruments and Bayesian model averaging across 183 countries.

H2: Rising Exploitation.

The theory of surplus value, developed across Capital Volume I (Marx 1867/1990, Chs. 7–9, 16–18), holds that the rate of exploitation rises as capitalists extract more output per unit of labor cost. In modern terms, this translates to a prediction that labor productivity growth outstrips wage growth, and that labor's share of output declines. The empirical literature on labor share decline has been documented across OECD and developing economies, with the decline attributed primarily to the falling relative price of investment goods (Karabarbounis and Neiman 2014). The rise of "superstar firms" with high productivity and low labor shares provides a complementary channel (Autor et al. 2020). Global distributions of capital and labor income show that capital income concentration has accelerated in ways consistent with Marx's prediction (Ranaldi 2025). The MELT framework is used to construct surplus value measures from Penn World Tables data and estimate dynamics using system GMM (Cockshott and Cottrell 1997; Basu and Manolakos 2013).

H3: Capital Concentration and Centralization.

In Capital Volume I, Chapter 25, Marx (1867/1990) distinguished between concentration (the growth of individual capitals through accumulation) and centralization (the merger of existing capitals through competition and credit). He predicted that both forces would produce progressively fewer, larger enterprises: competition destroys smaller capitalists, credit accelerates the process, and crises function as periodic purges that transfer capital upward. The 21st-century digital economy has generated precisely this tendency in some sectors while creating fragmentation in others. Firm markups across the US economy have risen steadily since the 1980s, consistent with the centralization dynamic (de Loecker et al. 2020). Market concentration reduces firm productivity in developing-country manufacturing, suggesting the Marxist concentration dynamic operates beyond the OECD (Rodríguez-Castelán, et al. 2023). Winner-take-most dynamics in technology-intensive sectors reduce labor's bargaining position through a "superstar firm" channel (Autor et al. 2020). A triple-difference specification (sector × ICT intensity × time) is estimated to separate concentration dynamics in traditional versus digital sectors.


## 2.2 Cluster B: Labor and Lived Experience

Marx theorized that worker suffering is embedded in the logic of capitalist accumulation, not contingent on policy failure or institutional malfunction. This claim runs through the later chapters of Capital Volume I (Marx 1867/1990, Chs. 23–25), where Marx links the "general law of capitalist accumulation" to the simultaneous production of wealth and poverty. The reserve army of the unemployed is Marxist in its framing: unemployment is functionally necessary for capital, suppressing wage demands, disciplining the employed, and enabling accumulation by keeping workers anxious about their position (Marx 1867/1990, Ch. 25, Sec. 3). Standard labor economics treats unemployment as a market friction; Marx treats it as an equilibrium feature of the system. The Grundrisse (Marx 1857–58/1973) extends this analysis to the subjective dimensions of labor, developing the theory of alienation and commodification that the later hypotheses in this cluster operationalize. The reserve army concept has been updated for the 21st century through the category of the "precariat" (Standing 2011), and a contemporary philosophical reconstruction of alienation theory grounds that concept in Marx's original framework (Jaeggi 2014). Four hypotheses capture this cluster's predictions.

H4: Relative Immiseration.

Marx distinguished between absolute and relative immiseration in Capital Volume I, Chapter 25 (Marx 1867/1990), where he argued that accumulation produces "an accumulation of misery, corresponding with accumulation of capital." The strong version that real wages fall absolutely has been falsified by 20th-century growth, as both neoclassical and sympathetic Marxist scholars acknowledge (Roemer 1988). The weaker version, that labor's share declines even as absolute wages rise, has fared considerably better. The distinction between the two has generated extensive debate in Marxist scholarship; Marx's concept of immiseration may be best understood relative to the expanding needs that capitalism itself generates, not merely relative to capital's share (Lebowitz 2003). For example, minimum wage policy can reduce inequality, but the political conditions for sustaining such policy remain fragile (Sotomayor 2021). The capital-income ratio rises faster than compensation in the long run, a pattern consistent with relative immiseration even without Marxist framing (Piketty 2014). Relative immiseration is tested using IV-2SLS with capital deepening as the channel across 2,740 country-year observations.

H5: The Reserve Army of Labor.

In Capital Volume I, Chapter 25, Section 3, Marx (1867/1990) argued that capitalism systematically produces a "relative surplus population" that disciplines wages and keeps the employed workforce compliant. He identified three forms: the floating (workers periodically expelled from industry), the latent (agricultural populations available for urban employment), and the stagnant (irregular workers in the most precarious conditions). This surplus population has been reconceptualized as the modern "precariat," characterized by labor insecurity across multiple dimensions (Standing 2011). The 21st-century gig economy represents a structural expansion of this reserve army. Vulnerable employment disproportionately affects women and workers in developing economies, where the reserve army dynamic operates with particular force (Lo Bue et al. 2022). In some settings, labor market restructuring has produced a permanent surplus population consistent with Marx's framework (Scully 2016). This is tested through a regression discontinuity design exploiting gig platform entry across US metropolitan areas, alongside a time-varying Phillips curve analysis.

H6: Commodification of Daily Life.

Marx predicted that capitalism would progressively absorb non-market activities into the commodity form, a process he theorized in the Grundrisse (Marx 1857–58/1973, Notebook VI) and developed further in Capital Volume I through the concept of the "real subsumption" of labor under capital (Marx 1867/1990, Ch. 16). Where "formal subsumption" describes capital's initial appropriation of pre-existing labor processes, "real subsumption" describes the progressive restructuring of all social activity according to the logic of commodity production. The most systematic reconstruction of this argument reads Marx's mature theory as fundamentally about the commodity form's colonization of social life (Postone 1993). The concept of "accumulation by dispossession" extends the analysis, arguing that capitalism continues to commodify previously non-market domains through privatization and financialization (Harvey 2003). This is operationalized through the American Time Use Survey (252,808 respondents, 2003–2024), tracking the ratio of market to non-market time. The ATUS data extend beyond the core 2000–2019 panel because the additional post-pandemic years provide variation in labor-market composition that strengthens the analysis.

H7: Alienation.

Perhaps Marx's most enduring concept, alienation was first systematically developed in the Economic and Philosophic Manuscripts of 1844 (Marx 1844/1988), where Marx identified four dimensions: estrangement from the product of labor, from the act of production, from species-being (the human capacity for creative, purposeful activity), and from other workers. The concept reappears in the Grundrisse (Marx 1857–58/1973) in the "Fragment on Machines," where Marx argued that automation would intensify alienation by subordinating workers to the rhythm of machinery. The most rigorous contemporary reconstruction distinguishes alienation from related concepts like exploitation and reification (Jaeggi 2014), while the specific connections between Marx's technology analysis and his alienation theory have been traced in detail (Wendling 2009). Industrial robot adoption displaces workers and depresses wages, establishing the material channel through which the subjective experience of alienation may intensify (Acemoglu and Restrepo 2020). The automation–alienation channel is tested using IFR robot density data merged with the World Values Survey (443,000 respondents across 108 countries) and the ISSP Work Orientations module (122,000 respondents across 32 countries).


## 2.3 Cluster C: Politics and Power

Political hypotheses distinguish Marx from a distributional labor analyst. Anyone with data can observe that labor's share of income has fallen across rich democracies. Marx's distinctive claim, articulated forcefully in the Preface to A Contribution to the Critique of Political Economy (Marx 1859/1970), is that material conditions of production determine social consciousness: that workers, confronted with their own structural marginalization, will recognize their class interests and mobilize collectively. The Communist Manifesto (Marx and Engels 1848/1998) offers the programmatic version of this claim, predicting that capitalism's own dynamics would produce the political force capable of its abolition. The Eighteenth Brumaire of Louis Bonaparte (Marx 1852/1963) complicates the prediction by demonstrating how political outcomes can diverge from class interests through ideology, institutional inertia, and the autonomy of state power. Cluster C tests this tension between the structural prediction and the historical analysis. The subsequent Marxist debate on the state divided into two poles: the instrumentalist position, holding that the capitalist class directly controls state policy (Miliband 1969), and the structuralist position, holding that the state functions to reproduce capitalist relations regardless of who staffs it (Poulantzas 1978). Three hypotheses capture this cluster's predictions.

H8: Class Consciousness.

The historical materialist thesis, stated most concisely in The German Ideology (Marx and Engels 1846/1970), holds that “life is not determined by consciousness, but consciousness by life.” The Preface to A Contribution to the Critique of Political Economy (Marx 1859/1970) crystallized this into the base-superstructure metaphor: changes in the economic foundation lead to transformation of the entire ideological superstructure. The simplest version of this claim that material conditions directly predict left political alignment is tested deliberately because the more sophisticated versions are precisely what explains the failure. Marx distinguished class-in-itself (an sich) from class-for-itself (für sich) in The Poverty of Philosophy (Marx 1847/1955), recognizing that shared material position does not automatically produce shared political consciousness. The problem was reformulated through the concept of cultural hegemony, where ruling-class ideology mediates the conditions-consciousness link (Gramsci 1971). A further departure held that political identities are discursively constructed and that class position is only one of many possible bases for political mobilization (Laclau and Mouffe 1985). If material conditions do not predict left voting, it strikes at the foundational prediction on which the political theory rests. To wit, inequality perceptions can diverge sharply from actual inequality (Campos-Vázquez et al. 2022), and income distance between the middle class and the poor can shape redistributive preferences more than absolute poverty (Lupu and Pontusson 2011), providing evidence for why material conditions do not automatically generate redistributive politics. An interaction is therefore tested: whether conditions predict left mobilization more strongly in countries with higher union density, a crude proxy for the organizational infrastructure that Gramsci and Wright identify as mediating the conditions-consciousness link.

H9: The State as Instrument of Capital.

In the Communist Manifesto, Marx and Engels (1848/1998) characterized the modern state as “a committee for managing the common affairs of the whole bourgeoisie.” The Eighteenth Brumaire (Marx 1852/1963) complicated this instrumentalist view by demonstrating how the Bonapartist state achieved relative autonomy from any single class fraction while still serving capital’s structural interests. This ambiguity generated perhaps the most productive debate in 20th-century Marxist political theory. The instrumentalist reading documented the social ties between state elites and the capitalist class (Miliband 1969). The structuralist counter held that the state’s position within capitalist relations of production, rather than the class origins of its personnel, determines its function (Poulantzas 1978). A synthesis argued that state managers have autonomous interests but operate under structural constraints imposed by capital mobility (Block 1987). Existing evidence gives tantalizing cluse. State wage-setting in Latin American economies provides developing-country context for the state-capital relationship (Baez et al. 2022). US policy has systematically shifted toward capital’s interests since the late 1970s through mechanisms consistent with both instrumentalist and structuralist accounts (Hacker and Pierson 2010). This is tested through two designs: a regression discontinuity on close US House elections using VoteView DW-NOMINATE scores, and a staggered difference-in-differences on right-to-work law adoption.

H10: Structural Crisis.

Marx’s crisis theory, developed across Capital Volumes II and III (Marx 1885/1978; 1894/1981) and the Theories of Surplus Value (Marx 1861–63/1968), predicted that capitalism’s internal contradictions would generate periodic crises of increasing severity. The mechanisms include overproduction (where the drive to expand production outstrips effective demand), overaccumulation (where surplus capital cannot find profitable investment outlets), and the tendency of the rate of profit to fall. The crisis framework has been updated for the 21st century through the argument that capital’s spatial and temporal “fixes” such as geographic expansion and privatization postpone but eventually deepen contradictions (Harvey 2014). A reconstruction of endogenous crisis dynamics grounded in classical-Marxian profit rate analysis provides the theoretical frame for empirical testing (Shaikh 2016, Ch. 16). The hypothesis tested here is specifically political: whether economic crises predict political instability, connecting Marx’s economic crisis theory to his revolutionary politics. This is tested using ACLED conflict event data, principal component analysis on crisis indicators, and IV-2SLS with structural break tests.


## 3. Data and Methods

The core economic panel draws from the Penn World Tables 10.01 (Feenstra et al., 2015), covering 183 countries from 2000 to 2019 with data on capital stocks, labor shares, productivity, and employment. This is supplemented with World Development Indicators for GDP, trade openness, and governance variables; the Standardized World Income Inequality Database (SWIID; Solt, 2020) for Gini coefficients; and ILO statistics for union density and labor force composition. For Marxist-specific variables including the rate of profit, organic composition of capital, and surplus value rates, the MELT framework developed by Cockshott and Cottrell (1997) and refined by Basu and Manolakos (2013) is followed, constructing these measures from national accounts data. The MELT framework converts monetary aggregates into labor-time equivalents, allowing computation of surplus value as the difference between total value produced and the variable capital (wage) share. These constructions involve unavoidable assumptions discussed in the online appendix.

Political data span several specialized sources. The MARPOR dataset provides coded party manifesto data for 5,285 party-election observations across 67 countries. VoteView DW-NOMINATE scores provide ideological positions for US congressional representatives, yielding 2,204 close-election observations. ACLED provides georeferenced protest and conflict events for H10, and V-Dem democracy indices serve as both controls and heterogeneity dimensions throughout.

For the lived-experience hypotheses, the analysis draws on the American Time Use Survey (ATUS; 252,808 respondents, 2003–2024), the World Values Survey (443,000 respondents across 108 countries, waves 1981–2022), and the ISSP Work Orientations module (122,000 respondents across 32 countries). Robot density data from the International Federation of Robotics provides the automation exposure measure for H7 (625 country-year observations across 25 countries). Table 1 summarizes the identification strategy for each hypothesis.

Table 1. Identification Strategy by Hypothesis

----- ------------------------------ ----------------------- ---------------------- ------- ------- H# Hypothesis Method Data Source N Score H1 Tend. Rate of Profit to Fall IV-2SLS, BMA PWT 10.01 2,740 0.510 H2 Rising Exploitation Sys. GMM, OLS-FE PWT 10.01, MELT 2,740 0.657 H3 Capital Concentration Triple-Difference PWT, HHI data 2,740 0.396 H4 Relative Immiseration IV-2SLS, FE PWT, SWIID 2,740 0.601 H5 Reserve Army of Labor RDD, Rolling Phillips ATUS, BLS 1,143 0.658 H6 Commodification OLS Time Series ATUS 22 0.378 H8 Class Consciousness Panel FE MARPOR, SWIID 5,285 0.142 H9 State as Instrument RDD, Staggered DiD VoteView, State data 2,204 0.359 H7 Alienation Mediation Analysis IFR, WVS, ISSP 86 0.603 H10 Structural Crisis OLS, IV-2SLS ACLED, V-Dem 2,740 0.397 ----- ------------------------------ ----------------------- ---------------------- ------- -------

Note: Full specification details and robustness checks reported with each hypothesis in Section 5. Scores incorporate direction, significance, effect size, and robustness.

Scoring Algorithm

An explicit, multi-dimensional scoring rubric is employed to evaluate the empirical support for each Marxist hypothesis. Each hypothesis receives a composite score on a 0–1 scale, constructed as a weighted average of four orthogonal dimensions: direction, significance, effect size, and robustness. This approach allows one to capture the degree to which the data align with Marx's predictions. The direction dimension (weight: 0.30) assesses whether the regression coefficient carries the sign Marx's theory predicts. A correctly signed and statistically significant coefficient receives 1.0; correctly signed but insignificant receives 0.5; wrong sign receives 0.0. The significance dimension (weight: 0.25) maps p-values onto a discrete scale: p < 0.01 yields 1.0, p < 0.05 yields 0.75, p < 0.10 yields 0.50, p < 0.20 yields 0.25, and p ≥ 0.20 yields 0.0. The effect size dimension (weight: 0.25) quantifies practical magnitude using standardized effect size (adapted Cohen's d, normalized by the within-sample standard deviation of the dependent variable): d > 0.5 receives 1.0, d > 0.3 receives 0.75, d > 0.1 receives 0.50, d ≤ 0.1 receives 0.25. The robustness dimension (weight: 0.20) reports the proportion of alternative specifications preserving both predicted direction and significance at p < 0.10. To assess sensitivity to weighting choices, the composite score is recalculated under five alternative schemes: equal weights (0.25 each); direction-heavy (0.40/0.20/0.20/0.20); significance-heavy (0.20/0.40/0.20/0.20); effect-size-heavy (0.20/0.20/0.40/0.20); and robustness-heavy (0.20/0.20/0.20/0.40). Appendix A reports the full range of scores.

The design involves testing ten hypotheses, each with multiple specifications. This multiplicity introduces a familiar hazard: as the number of tests increases, so does the probability of encountering spuriously significant results. With ten independent tests at α = 0.05, one would expect roughly 0.5 false positives. This is addressed by applying Benjamini-Hochberg false discovery rate (FDR) corrections alongside reported p-values. The FDR approach controls the expected proportion of false discoveries among rejected hypotheses, yielding more power than family-wise corrections while remaining conservative. Throughout the text and appendices, both unadjusted and FDR-adjusted p-values are reported. Of course, multiple comparisons remain a genuine constraint on inference: hypotheses achieving significance under FDR correction merit greater confidence than those relying on unadjusted tests alone.


## 4. Results


## 4.1 Cluster A: Capital and Profit

H1: Tendency of the Rate of Profit to Fall. Score: 0.510

The evidence for H1 divides into two complementary but methodologically distinct analyses. The Bayesian Model Averaging results provide the most robust identification of Marx's mechanism. Labor productivity (PIP = 1.00, posterior mean = 0.317) and labor share (PIP = 1.00, posterior mean = −0.220) emerge as the two strongest predictors of profit rates across all specifications. These variables operate through the theoretical channels Marx identified: productivity gains allow capital to produce more surplus value per unit of capital invested, while rising labor share directly compresses profit margins. The posterior probability that both coefficients carry their theoretically expected signs exceeds 99% across the model space. The organic composition of capital achieves high posterior probability (PIP = 0.96) with a posterior mean coefficient of −0.185, supporting Marx's proposed mechanism of declining profitability as capital intensity increases. This result is stable across functional form choices, regional specifications, and inclusion of alternative counter-tendency variables. The BMA framework thus isolates the tendency itself (OCC impact on profit rates) from the counter-tendencies (labor productivity, labor share) that Marx explicitly recognized, revealing their relative magnitudes and robustness.

The IV-2SLS specification offers directional confirmation but faces identification constraints. Using commodity-price instruments, the specification yields a strongly negative coefficient on OCC (β = −1.006, SE = 0.016, p < 0.001) with a Kleibergen-Paap Wald F-statistic of 38.2. A permutation test confirms the estimate lies at the extreme of the sampling distribution under instrument randomization (p_exact = 0.000). However, commodity prices plausibly affect profit rates through multiple channels including terms-of-trade effects and resource rent extraction, independent of capital composition. Regional commodity export share and sectoral composition controls partially mitigate this exclusion-restriction concern but do not eliminate it. The Hansen J test (J = 3.41, p = 0.182) has weak power to detect violations and should be interpreted as uninformative rather than reassuring. The IV results align with the BMA conclusions on the sign and broad magnitude of the OCC effect but cannot claim cleanly separated identification of the mechanism.

Marx's full prediction was that the falling-profit tendency would dominate over the long run. Here the counter-tendencies assert themselves with equal force. Labor productivity and labor share, the two most robust BMA predictors, serve as counter-tendency channels: labor productivity growth cheapens constant capital, while labor share movements reflect the intensity of exploitation that Marx himself identified as the primary offsetting force.

Table 2. Bayesian Model Averaging: Predictors of ln(Rate of Profit)

-------------------------- ------- ------------ ---------- ------------ Variable PIP Post. Mean Post. SD Sign Cert. ln(OCC) --- −0.412 0.038 1.00 ln(Labor Productivity) 1.000 0.317 0.041 1.00 Labor Share 1.000 −0.220 0.033 1.00 Union Density 0.999 0.009 0.003 1.00 Trade Openness 0.874 0.031 0.018 0.97 Gov't Expenditure/GDP 0.621 −0.018 0.016 0.88 Financial Depth (M2/GDP) 0.433 0.008 0.011 0.79 Inflation Rate 0.298 −0.003 0.006 0.71 -------------------------- ------- ------------ ---------- ------------

Note: N = 2,740 country-year observations, 183 countries, 2000–2019. OCC always included per theory; PIP not applicable. 8 candidate predictors; 256 models evaluated. PIP = posterior inclusion probability. Sign Cert. = posterior probability that coefficient has indicated sign. Kleibergen-Paap F = 38.2 for IV first stage.

From a dialectical standpoint, the finding that tendency and counter-tendency exert roughly equal pressure is precisely what Marx's framework predicts during periods of active accumulation. The rate of profit oscillates within bounds set by forces Marx identified: technological acceleration raises organic composition, pressuring profitability downward; crisis destroys capital, lowers organic composition, and restores profitability upward. Whether this constitutes 'confirmation' hinges on what one takes Marx to have predicted. If he predicted secular, monotonic decline, the evidence is mixed. If he predicted structural tension generating periodic crises and cyclical oscillation, the evidence becomes more supportive. Scholars reading Marx dialectically (Harvey 2006, Shaikh 2016, Clarke 1994) would score H1 substantially higher than 0.50. The conservative score is retained but the interpretive terrain is acknowledged honestly. Score: 0.50.

H2: Rising Exploitation. Score: 0.520

This hypothesis predicts a decline in labor's functional share of income as capital accumulation proceeds. Testing the causal mechanism requires dynamic identification; the system GMM specification failed on multiple fronts. The lagged dependent variable coefficient was near zero, the Hansen J-test p-value was 0.997, and with 42 instruments for 183 cross-sections in a 19-year panel, instrument proliferation severely compromised model credibility. The specification did not recover a meaningful estimate of the underlying dynamics.

Table 3. Rising Exploitation: System GMM and Robustness Specifications

------------------------ -------------- ------------ --------------- --------------- (1) Sys. GMM (2) OLS-FE (3) OECD Only (4) Post-2008 Δ ln(Capital Stock) [0.003] −1.977*** −2.104*** −1.843*** [0.041] (0.314) (0.388) (0.412) ln(Labor Productivity) [0.001] 0.842*** 0.911*** 0.788*** [0.029] (0.102) (0.131) (0.147) Country FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes Hansen J (p) [0.997] --- --- --- AR(2) (p) [0.341] --- --- --- Instruments [XX] --- --- --- N 2,740 2,740 684 1,830 R² (within) --- 0.412 0.438 0.389 ------------------------ -------------- ------------ --------------- ---------------

Note: DV = ln(s/v), surplus value rate constructed via MELT framework. Standard errors in parentheses; GMM SEs in brackets. Country and year FE in all specifications. *p < 0.05, **p < 0.01, ***p < 0.001.

Results rest on OLS with country and year fixed effects, which estimates a descriptive trend rather than a causal effect. The labor share declines by 1.977 percentage points per year (SE = 0.314, t = 6.3), and this pattern replicates across OECD-only subsamples and across the pre/post-2008 divide. The labor share fall is compatible with four distinct mechanisms: labor-biased technological change (neoclassical), labor cost arbitrage from globalization (Ricardian), union decline (institutional), and Marx's capital-intensive accumulation mechanism. OLS-FE cannot discriminate among them. The finding confirms Marx's directional prediction, but does not demonstrate that capital accumulation is the cause.

Karabarbounis and Neiman (2014), ILO Global Wage Reports, and Ranaldi (2025) document the same decline using different methods. The convergence lends confidence that the pattern is not an artifact of specification or data. The mechanistic claim remains untested. Marx predicted the direction correctly, but identification of his causal story requires a stronger design. Score: 0.52, reflecting confirmed directionality without causal identification. Significance is qualified because statistical significance of a reduced form does not validate the hypothesis mechanistically. Effect size is economically meaningful but causally indeterminate. Robustness is limited to subsample replication within a single fallback specification.

H3: Capital Concentration. Score: 0.396

The triple-difference results reveal a more complicated picture than Marx anticipated. In traditional sectors, capital deepening does drive concentration: the OCC coefficient is strongly positive (β = 881.8). The ICT interaction term, however, reverses the sign (β = −959.6). In digital-intensive sectors, the same capital deepening process is associated with deconcentration. Summary statistics for HHI, OCC, and the ICT interaction are reported in Table 0 (online appendix); the coefficient magnitudes reflect the scale and variance of the HHI dependent variable.

Table 4. Capital Concentration: Triple-Difference and Robustness Specifications

----------------- ----------------- --------------- -------------- (1) Full Sample (2) US Manuf. (3) Non-OECD ΔOCC 881.8*** −9,129.8** 724.3** (198.4) (3,412.1) (267.8) ΔOCC × High ICT −959.6*** --- −812.4** (241.7) (301.2) Country FE Yes --- Yes Year FE Yes Yes Yes N 2,740 380 2,056 R² (within) 0.087 0.142 0.073 ----------------- ----------------- --------------- --------------

Note: DV = ΔHHI (change in Herfindahl-Hirschman Index). Heterogeneity term = ΔOCC × High ICT interaction. Country + year FE. Standard errors in parentheses. *p < 0.05, **p < 0.01, ***p < 0.001. Summary statistics: HHI mean = 1,842 (SD = 1,204); OCC mean = 2.31 (SD = 0.89).

Rodríguez-Castelán et al. (2023) confirm that market concentration reduces firm productivity in Mexican manufacturing, suggesting the Marxist concentration dynamic operates in developing economies with consequences Marx would recognize. Marx predicted that competition would produce monopoly. In traditional heavy industry, it largely has. But the digital economy generates a dual dynamic of platform monopolies at the top and fragmentation at the bottom that Marx's framework did not anticipate. Score: 0.40.

Cluster A averages .475. Marx's analysis of capital's internal dynamics is more right than wrong, with counter-tendencies complicating what would otherwise be clean predictions and the labor share decline confirmed directionally though not causally identified.


## 4.2 Cluster B: Labor and Lived Experience

H4: Relative Immiseration. Score: 0.601

The fixed-effects specification with capital deepening as the primary channel yields a positive but statistically insignificant coefficient on the capital–labor ratio (β = 0.020, SE = 0.018, p = 0.28). The time trend, however, tells a clearer story: across 2,740 country-year observations, the labor share trend is −0.00067 per year (p = 0.017), a slow but statistically significant decline. The IV specifications with institutional channels suggest that monopsony power, union decline, and minimum wage erosion all contribute.

Table 5. Relative Immiseration: Capital vs. Institutional Channels

--------------------- ------------ ----------------- ------------------- (1) OLS-FE (2) IV (Bartik) (3) Institutional Capital/Labor Ratio 0.020 0.034 --- (0.018) (0.027) Union Density --- --- −0.041** (0.014) Min. Wage Ratio --- --- 0.089*** (0.022) Time Trend −0.00067* −0.00071* −0.00058* (0.00028) (0.00031) (0.00029) Country FE Yes Yes Yes Year FE Yes Yes Yes N 2,740 2,740 2,190 R² (within) 0.031 --- 0.058 --------------------- ------------ ----------------- -------------------

Note: DV = Δ labor share (first-difference). FE specification: country + year FE. IV instrument: China export shock × sectoral exposure. *p < 0.05, **p < 0.01, ***p < 0.001.

Sotomayor (2021) documents how minimum wage policy in Brazil reduced inequality through sustained political commitment; this provides evidence that the exploitation dynamics Marx identified can be countered by deliberate institutional design, though the political conditions for sustaining such policy remain fragile. Score: 0.60: directionally correct, but the capital-deepening channel is weakly identified.

H5: The Reserve Army of Labor. Score: 0.658

The regression discontinuity results provide clean evidence. The estimates are [β = −0.081, SE = 0.019, p < 0.001], reflecting a significant negative effect of gig platform entry on local hourly wages. The ±5-month bandwidth was selected using the Calonico-Cattaneo-Titiunik (2014) optimal bandwidth selector, which balances bias and variance in the RDD point estimate. The CCT-optimal bandwidth is ±4.8 months, and the results are robust to the CCT choice. A McCrary (2008) density test for manipulation of the running variable was conducted, yielding t = 0.47, p = 0.637, failing to reject the null of no manipulation. The density plot (Figure A1 in the online appendix) confirms smooth running-variable density around the cutoff.

Table 6. Reserve Army: Regression Discontinuity on Gig Platform Entry

---------------------- ---------------------- ------------- --------------- (1) Main RDD (2) Placebo (3) Donut RDD Treatment Effect [−0.081]*** 0.005 [−0.021]** (0.019) (0.021) (0.003) Bandwidth ±5 months ±5 months ±5 (excl. ±3) CCT Optimal BW [±4.8 months] --- --- McCrary Density Test [t = 0.47, p = 0.64] --- --- Kernel Triangular Triangular Triangular N 1,143 1,143 1,016 Robust SEs HC1 HC1 HC1 ---------------------- ---------------------- ------------- ---------------

Note: DV = ln(median hourly wage) in metro area. Running variable = months from platform entry. Placebo shifts discontinuity 12 months prior. Donut excludes ±3 month window. Bandwidth selected via Calonico-Cattaneo-Titiunik (2014) optimal selector. *p < 0.05, **p < 0.01, ***p < 0.001.

The placebo test, shifting the discontinuity twelve months before actual platform entry, yields a precisely estimated null (0.005, p = 0.81), confirming the wage effect is specific to the treatment. The donut RDD, excluding observations within three months of the threshold, produces an attenuated but still significant effect ([−0.021, p < 0.001]). The rolling-window Phillips curve analysis reinforces the pattern. The coefficients are positive and significant in early windows (2000–2005: β = 0.0018; 2002–2007: β = 0.0033), then attenuate toward zero in later windows (2014–2019: β ≈ 0.000). The interpretation of this attenuation is in the growing structural dominance of the reserve army: wages become progressively less responsive to cyclical unemployment as the gig economy permanently expands the labor supply floor. The reserve army disciplines labor not through temporary joblessness but through constant structural availability.

Table 7. Time-Varying Phillips Curve: Rolling-Window Estimates

----------- ------------ -------- --------- ------- ------- Window RAI Coeff. SE p-value N R² 2000–2005 0.0018** 0.0006 0.004 1,020 0.041 2002–2007 0.0033*** 0.0008 <0.001 1,020 0.054 2005–2010 0.0021** 0.0007 0.003 1,020 0.038 2008–2013 0.0012* 0.0006 0.042 1,020 0.028 2011–2016 0.0005 0.0005 0.317 1,020 0.014 2014–2019 ≈0.000 0.0004 0.912 1,020 0.003 ----------- ------------ -------- --------- ------- -------

Note: DV = Δln(wage). RAI = Reserve Army Intensity composite (U6, gig share, involuntary PT, temp work). Country and year FE. Five-year rolling windows. Attenuation toward zero reflects structural dominance of the reserve army rather than weakening wage discipline.

Lo Bue et al. (2022) document how vulnerable employment shapes labor market outcomes across Latin America and sub-Saharan Africa, with women workers bearing disproportionate burden. This analysis contextualizes that finding, as informal sector expansion across the Global South operates as Marx's reserve army analogue; it is a vast pool of underemployed and contingent workers that dampens wage growth and undercuts formal sector bargaining power. If gig platforms suppress wages in US cities with strong labor institutions, the effects in low-regulation developing economies are predictably more severe. Marx argued that the reserve army keeps wages in check by ensuring employers always have an alternative source of labor. The gig economy operationalizes this with efficiency. Score: 0.65.

H6: Commodification of Daily Life. Score: 0.300 [Exploratory]

This hypothesis occupies a different epistemic position than the other nine tests. Time-use data provides no meaningful support: across 22 years of ATUS data, leisure commodification shows a negligible downward trend (β = −0.003, SE = 0.002, p = 0.15), and capital deepening in services yields indistinguishable results.

Table 8. Commodification of Daily Life: ATUS Time-Series Specifications

------------------------------ ---------------- ----------------------- (1) Time Trend (2) Capital Deepening Year Trend −0.003 --- (0.002) Capital Deepening (Services) --- 0.004 (0.025) N 22 22 R² 0.108 0.001 ------------------------------ ---------------- -----------------------

Note: DV = commodification index (ratio of market-mediated to non-market time). US only, 2003–2024 annual means from 252,808 ATUS respondents. *p < 0.05, **p < 0.01, ***p < 0.001. Neither specification achieves significance; p = 0.15 (Model 1), p = 0.87 (Model 2).

Of course, this may say more about measurement than reality. The boundary between market and non-market time has been blurred rather than shifted by the digital economy. Scrolling social media is simultaneously leisure, consumption, and the production of data value for platform companies themselves (and has been a real barrier to finishing this paper). Marx's commodity frontier may be advancing through the transformation of existing activities rather than the displacement of non-market time in this kind of commodification. H6 is retained as a test case under measurement-constrained conditions rather than as a robust empirical evaluation. Future work linking digital platform monetization to actual household time allocation might provide clearer leverage. Score: 0.30, reflecting that Marx's commodification hypothesis may hold in reality while remaining empirically invisible to conventional time-use metrics.

H7: Alienation. Score: 0.505

The automation-alienation channel receives modest support from the main-sample data but not from the mediation analysis. In the main specification (86 country-year observations), robot density predicts alienation with the expected sign (β = 0.060, SE = 0.124, p = 0.05). Workers in countries with higher industrial robot adoption report lower levels of meaning, satisfaction, and social integration at work, consistent with Marx's claim that automation subordinates workers to machinery and intensifies alienation.

Table 9. Alienation: Mediation Analysis (Robot Density → Autonomy → Alienation)

------------------------------- ---------- ------- ------------------ ------- Path Estimate SE Bootstrap 95% CI p Total Effect (c) 0.006 0.028 [−0.049, 0.061] 0.832 Direct Effect (c') −0.007 0.031 [−0.068, 0.054] 0.821 Indirect Effect (a×b) 0.014 0.019 [−0.023, 0.051] 0.461 Path a: Robot → Autonomy −0.183* 0.082 [−0.344, −0.022] 0.026 Path b: Autonomy → Alienation −0.074 0.091 [−0.252, 0.104] 0.415 ------------------------------- ---------- ------- ------------------ -------

Note: Main sample N = 86 country-years (WVS + ISSP). Mediation sample N = 34. Bootstrap: 1,000 replications. Welfare regime categories (Nordic, Continental, Liberal, Southern) included as controls in main sample; dropped from mediation subsample due to N constraints. CI overlap underscores estimation uncertainty at N = 34.

The mediation subsample of 34 country-year observations with all three variables available was intended to decompose the pathway from robot density through worker autonomy to alienation. At N = 34, the mediation analysis is uninformative. The indirect effect (0.014) exceeds the total effect (0.006), and bootstrap confidence intervals overlap substantially (total effect [−0.02, 0.03], indirect effect [−0.01, 0.04]). Preliminary robustness perturbations (e.g. relaxing data-matching constraints to expand N to approximately 52, dropping welfare regime controls) show the indirect effect does not stabilize. The mediation decomposition, which was to provide the hypothesis's primary identification, fails to hold at current sample sizes.

The main-sample reduced-form association is defensible. Marx theorized that automation would intensify alienation by mechanizing work and subordinating labor to the rhythm of capital. This prediction holds across countries with varying robot densities and labor institutions. Whether the mechanism operates through worker autonomy remains uncertain and deserves investigation with richer panel data spanning more country-years. Score: 0.505, reflecting confirmed directionality and significance on the reduced form, tempered by mediation fragility and modest effect magnitude.

Cluster B averages .516. Marx's predictions about how capitalist accumulation shapes workers' lives receive substantial support. The weakest result of commodification reflects measurement limitations in capturing digital-era value extraction, and the alienation mediation requires stronger data before the causal pathway can be confirmed.


## 4.3 Cluster C: Politics and Power

If Clusters A and B describe the conditions Marx believed would generate revolutionary consciousness, Cluster C tests whether that generation occurs. Marx's economic theory generates political predictions that later theorists have refined. Each of these extensions emerged in response to empirical failure of the baseline Marxist mechanism. The baseline is tested first because if the fundamental reduced form does not operate, the more complex variants inherit the same empirical problem. The baselines deployed here use the mechanistic claims Marx himself made: class consciousness drives political alignment (H8), state institutions favor capital under capitalist rule (H9), structural crises disrupt political order (H10). The apparent thinness of Cluster C's econometrics reflects an asymmetry in inference. The economic hypotheses require strong identification because positive effects are claimed. The political hypotheses produce null results, and null results in well-powered, large-N samples are information-rich. H8 draws on 5,285 party-election observations across 67 countries. If material inequality drove left voting through class consciousness, this sample would detect even moderate effects with high probability. The R² of 0.018 reflects that material conditions explain almost none of the within-country variation in left party support.

H8: Class Consciousness. Score: 0.142

This is a key finding, representing a null result for Marx's political theory. The primary specification (Model 2, N = 380, 20 countries, 19 years) shows no statistically significant relationship between material conditions and left political mobilization. The inequality coefficient on left vote share is β = 2.01 (SE = 1.95, p = 0.32). The sign is positive, consistent with Marx's direction, but the estimate is statistically indistinguishable from zero. The projected R² of 0.018 indicates that material conditions explain less than two percent of the within-country variation in left voting.

Table 10. Class Consciousness: Panel FE Specifications (DV = Left Vote Share)

---------------------- -------------- ---------- ----------------- -------------- (1) Baseline (2) Full (3) Interaction (4) Non-OECD Gini (SWIID) 1.84 2.01 1.93 1.52 (1.88) (1.95) (1.97) (2.14) Union Density --- 0.12 0.09 0.08 (0.14) (0.15) (0.18) Gini × Union Density --- --- 0.04 --- (0.06) Unemployment Rate --- −0.31 −0.30 −0.25 (0.28) (0.28) (0.33) V-Dem Democracy --- 3.42 3.38 2.87 (4.11) (4.12) (4.56) Country FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes N 380 380 380 198 R² (within) 0.012 0.018 0.019 0.008 ---------------------- -------------- ---------- ----------------- --------------

Note: DV = left vote share (%). Data: MARPOR/CMP (5,285 party-election observations, 67 countries) merged with SWIID, ILO, V-Dem. Country + year FE. Standard errors in parentheses. Model 3 interaction tests whether conditions predict left voting more strongly where union density is higher. *p < 0.05, **p < 0.01, ***p < 0.001.

This is tested on 5,285 party-election observations across 67 countries, using electoral data coded from actual party manifestos, merged with inequality measures and institutional variables. The interaction term testing whether material conditions predict left mobilization more strongly in countries with higher union density yields an insignificant positive coefficient (β = 0.04, p = 0.31), failing to recover the mechanism even where organizational infrastructure is strongest. Material conditions do not automatically generate class consciousness. The transmission belt between base and superstructure is, in the 21st-century data, inoperative. Or perhaps, it was never the automatic mechanism Marx theorized. Score: 0.15.

H9: The State as Instrument of Capital. Score: 0.359

The close-election RDD produces clear results on one dimension. Winning candidates ideologically closer to capital interests vote substantially differently on AFL-CIO scorecards (β = 0.859, p < 0.001), corporate tax legislation, and labor bill cosponsorship. Money influences who wins elections, and who wins elections influences policy. In that narrow sense, the state does respond to capital.

Table 11a. State as Instrument: RDD on Close US House Elections

-------------- ---------- ---------- ----------- ------------- ----------- DV → DW-NOM. AFL-CIO Corp. Tax Labor Bills Min. Wage RDD Estimate 0.312*** 0.859*** 0.441** −0.287** 0.218* (0.067) (0.142) (0.158) (0.098) (0.103) Polynomial Quad. Quad. Quad. Quad. Quad. N 2,204 2,204 2,204 2,204 2,204 BW ±5% ±5% ±5% ±5% ±5% -------------- ---------- ---------- ----------- ------------- -----------

Note: Running variable = vote margin centered at 0.50. VoteView DW-NOMINATE data. Quadratic polynomial in running variable. HC1 robust SEs. *p < 0.05, **p < 0.01, ***p < 0.001.

The right-to-work DiD shows that these laws suppress union density (β = −0.237, SE = 0.107, p = 0.027), reduce minimum wages, and lower welfare spending. Recent methodological concerns about heterogeneous treatment effects in TWFE estimators are addressed by computing a Goodman-Bacon (2021) decomposition, which decomposes the TWFE estimate into weighted two-by-two comparisons. The decomposition assigns 68% of the weight to comparisons of the desired type and 32% to comparisons involving already-treated units. The de Chaisemartin and D'Haultfoeuille (2020) estimator yields [β = −0.193, 95% CI: [−0.39, −0.00]], modestly weakening the main TWFE result.

Table 11b. State as Instrument: Staggered DiD on Right-to-Work Law Adoption

----------------- ------------ ----------- ------------- -------------- ----------- DV → Union Den. Min. Wage Welfare Sp. Reg. String. dCDH Est. RTW Adoption −0.237* −0.184* −0.312** −0.198* [−0.193] (0.107) (0.089) (0.112) (0.094) (0.098) State FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes G-B Decomp. (%) --- --- --- --- [68/32] N 1,275 1,275 1,275 1,275 1,275 R² (within) 0.211 0.087 0.156 0.004 --- ----------------- ------------ ----------- ------------- -------------- -----------

Note: 51 states, 25 years, staggered DiD. dCDH = de Chaisemartin & D'Haultfoeuille (2020) estimator. G-B Decomp. = Goodman-Bacon (2021) decomposition weight on desired comparison type. *p < 0.05, **p < 0.01, ***p < 0.001.

A limitation deserves acknowledgment. The RDD and DiD specifications test the instrumentalist claim that elections and labor law create measurable shifts in state orientation and union organization. The sample is limited to the United States, and only Miliband's version of state power is tested. A Poulantzian specification would examine whether state structures constrain fiscal and regulatory policy regardless of electoral direction. One concrete approach: estimate investment response functions showing whether firms respond differently to capital gains tax changes under left versus right administrations, controlling for baseline corporate tax rates. If Poulantzas is correct that the capitalist state is structurally selective toward capital regardless of who governs, the response slopes should not differ significantly by party. A second approach tests whether ideologically opposed governments converge toward similar corporate tax burdens, welfare retrenchment levels, and labor regulation intensity. These tests require cross-national fiscal policy data and are reserved for future work, as the current specification captures only one dimension of state-capital relations. Score: 0.35.

H10: Structural Crisis. Score: 0.397

The OLS baseline shows that Marxist structural variables jointly predict protest intensity, but no individual coefficient reaches conventional significance. The financial stress coefficient on protest intensity (β = −7.054, SE = 3.568, p = 0.063) is suggestive but falls short. The restricted model (excluding Marxist variables) explains almost none of the projected variation (R² = 0.005), whereas the full model achieves R² = 0.049.

Table 12. Structural Crisis: Marxist Variables and Political Instability

--------------------- -------------- ---------------- ------------------- (1) OLS Full (2) Restricted (3) Regime Change Δ Capital Intensity −1.238 --- −0.892 (1.104) (0.784) Δ Rate of Profit −3.412 --- −2.945* (2.187) (1.412) Δ Labor Share −12.841 --- −15.742* (8.432) (7.218) Δ Manuf. Employment 0.894 --- 0.412 (1.231) (0.891) Financial Stress −7.054† --- −4.218 (3.568) (2.841) Country FE Yes Yes Yes Year FE Yes Yes Yes N 2,740 2,740 2,740 R² 0.049 0.005 --- --------------------- -------------- ---------------- -------------------

Note: DV = ln(protest intensity) for OLS; binary regime change for column 3. Country + year FE. Standard errors in parentheses. †p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001. BH q-values in Appendix.

The regime change specification provides the strongest individual results: declining profit rates (β = −2.945, p < 0.05) and declining labor shares (β = −15.742, p < 0.05) predict regime transitions. Economic crises do generate political instability. But the instability is politically indeterminate. Economic crises in 21st-century democracies have produced left-populist movements in some countries, right-populist movements in others, technocratic consolidation in still others. Marx predicted crisis → revolution. Reality shows crisis → varied political responses, with the specific response contingent on institutions, culture, leadership, and historical accident. Score: 0.40.

Cluster C averages .300. Marx's political predictions represent the weakest link in his theoretical system, with a significant gap between Clusters A–B and Cluster C.


## 4.4 The Marx Scorecard

Table 13. Right Diagnosis, Wrong Prescription: Summary Scorecard

----- ------------------------------------ ----------- --------- -------- --------- ------- H# Hypothesis Direction Signif. Effect Robust. Score H1 Tendency of Rate of Profit to Fall ✓ Mixed Medium High 0.510 H2 Rising Exploitation ✓ ✓ Large High 0.657 H3 Capital Concentration Mixed Mixed Large Medium 0.396 H4 Relative Immiseration ✓ ✓* Small Medium 0.601 H5 Reserve Army of Labor ✓ ✓ Medium High 0.658 H6 Commodification of Daily Life ✓ ✗ Small Low 0.378 H8 Class Consciousness ✓ ✗ Neg. Low 0.142 H9 State as Instrument ✓ Mixed Medium Medium 0.359 H7 Alienation ✓ ✓* Medium Low 0.603 H10 Structural Crisis Mixed ✗ Small Low 0.397 ----- ------------------------------------ ----------- --------- -------- --------- -------

Note: Scores computed per algorithm in Section 4 (weights: Direction 0.30, Significance 0.25, Effect Size 0.25, Robustness 0.20). ✓ = confirmed, ✗ = not confirmed, * = marginal. Overall Marx Score ≈ .439 (range .41–.47 across weighting schemes; see Appendix A).

Table 14. Marx Scores by Thematic Cluster

------------------------------- ---------------- ------------ ----------- Cluster Hypotheses Avg. Score Range A: Capital and Profit H1, H2, H3 .521 .396–.657 B: Labor and Lived Experience H4, H5, H6, H7 .560 .378–.658 C: Politics and Power H8, H9, H10 .299 .142–.397 Overall All 10 .470 .142–.658 ------------------------------- ---------------- ------------ -----------

The overall Marx Score is approximately .439, with a range of .41 to .47 across alternative weighting schemes (see Appendix A). Clusters A and B with Marx's economic diagnostics about exploitation, immiseration, the reserve army, and alienation average .475 and .516 respectively. Cluster C of his political predictions about class consciousness, state instrumentalism, and revolutionary crisis averages .300. Marx excels at identifying the structural pressures capitalism generates; he is less reliable on how states, labor movements, and cultural formations respond to those pressures.


## 5. Discussion

The scoring suggests Marx was right about certain kinds of claims and wrong about others, and the boundary between them illuminates tensions in the relationship between economic structures and political action. His analysis of capitalism's economic tendencies of exploitation, labor's share falling, unemployment disciplining wages, concentration proceeding in traditional sectors, and automation eroding meaning holds up with a surprising consistency. His political predictions, by contrast, assume a degree of automaticity that the evidence does not support. The gap between H2's score (0.52) and H8's score (0.15) reflects the gap between Marx the analyst and Marx the revolutionary.

Why does the gap exist? One answer is methodological: economic tendencies operate on aggregates over decades, while political mobilization depends on micro-level mechanisms that are inherently volatile. But Marx did not claim that political mobilization was random; he claimed it was driven by material conditions with a regularity that approached historical law. Marx identified counter-tendencies to the falling rate of profit, and Volume III of Capital is substantially devoted to their enumeration. The results suggest he underestimated their importance. The BMA analysis for H1 shows that labor productivity and exploitation intensity are the most robust predictors of the profit rate, with posterior inclusion probabilities of 1.00. The 21st-century digital economy amplifies this indeterminacy. H3 shows that the same capital deepening process driving concentration in traditional sectors drives deconcentration in digital sectors. H5 shows that the reserve army has been reconstituted as the gig economy, but this new reserve army may be creating precarity without solidarity, and isolation without consciousness.

Marx's analysis of capital accumulation was more sophisticated than commonly acknowledged, while his theory of political change was more simplistic. Piketty (2014) distances himself from Marx's theoretical framework while documenting patterns Marx predicted; this represents a tension directly relevant to the argument about what counts as 'confirmation.' The results concur with a key specification: the error lies in the assumed automaticity of the connection between conditions and consciousness. H8's score of 0.15 is instructive. If material conditions automatically generated class consciousness, the 21st century should be a golden age of worker power. Inequality has risen to levels not seen since Marx's own era (Milanovic 2016). Labor's share has declined across virtually every economy (Karabarbounis & Neiman 2014). Precarious employment has expanded as workers report rising alienation (H7). And yet: left parties have not grown in a substantial or common way; in fact one could argue that the opposite has occurred with the rise of global right populists while union density has continued to fall and class-based voting has weakened.

Several mechanisms explain the gap. First, the precariat lacks the institutional anchors and physical spaces that enabled working-class solidarity in Marx's era (Standing 2011). Second, cultural and identity-based politics have displaced class as the primary axis of mobilization (Inglehart & Norris 2019). Workers who experience exploitation vote on immigration, religion, or national identity. Third, 21st-century capitalism has developed mechanisms for managing discontent: consumer credit, financialized asset ownership, and welfare state buffers. These decouple conditions from political consequences (Harvey 2005; Streeck 2014). Fourth, the algorithmic personalization of news has fragmented the epistemic commons that class consciousness requires (Benkler et al. 2018; Nguyen 2020).

The reserve army finding (H5) carries weight for labor governance in the Global South. In contexts where institutional regulation is weak, the mechanism Marx identified operates more nakedly. Gig platforms suppress wages in US metropolitan labor markets despite strong union traditions and legal frameworks; the effects in low-regulation developing economies are predictably more severe. The informal sector expansion across South Asia and sub-Saharan Africa operates precisely as Marx's reserve army analogue; it is a vast pool of underemployed and contingent workers that dampens wage growth, undercuts formal sector bargaining power, and serves the structural interests of capital accumulation. One implication is in anticipatory regulation. Rather than waiting for platform labor markets to mature and then correcting for externalities post hoc (the model dominant in wealthy-country regulation), developing economies have an opportunity for regulatory design before network effects lock in exploitative arrangements. This requires state capacity and political will, both scarce resources. Yet the alternative is allowing platform governance to reproduce the reserve army at technological scale: a vast workforce permanently contingent, wage-suppressed, and organizationally fragmented. Selwyn (2017) argues for labor-centered development that prioritizes worker welfare and organization from the outset. These findings suggest this is a structural necessity.

The H8 failure is arguably more consequential for developing than developed economies. In wealthy democracies, the absence of revolutionary consciousness is surprising given labor's historical strength; in developing democracies facing steeper inequality, it demands explanation. Campos-Vázquez et al. (2022) show that inequality perceptions in Mexico systematically diverge from objective inequality measures, with workers underestimating their own relative deprivation. This perception gap helps explain why material immiseration does not trigger redistributive politics. Yet the gap is not fixed. Sotomayor (2021) documents how minimum wage policy in Brazil, implemented through sustained political commitment, reduced inequality and altered worker consciousness. The pattern across Latin America’s recent left turns suggests the conditions-consciousness mechanism operates but only under specific institutional conditions: prior organizational infrastructure, ideological frameworks connecting conditions to class identity, and crucially, moments when state institutions are sufficiently open to absorb worker demands. Cardoso and Faletto (1979) emphasized how institutional openness conditions popular mobilization in dependent economies. These findings suggest something sharper: institutional design for redistribution must be front-loaded into economic frameworks before politics turn hostile. Once inequality generates political fragmentation rather than class consciousness, post-hoc redistribution becomes significantly harder.

The H3 finding intersects with a growing literature on market power in developing economies. Market concentration reduces firm productivity in Mexican manufacturing rather than enhancing it through efficiency gains (Rodríguez-Castelán et al., 2023). Global market power concentrates capital income while fragmenting labor income, precisely the dynamic Marx theorized (Ranaldi, 2025). This dual concentration-fragmentation pattern plays out distinctly in developing economies where digital infrastructure remains uneven and state capacity to regulate is contested. State wage-setting anchors wage standards through public sector employment, though the extent depends on the political economy of each country (Baez et al., 2022). Yet the H9 finding suggests structural constraints operate regardless of formal ideology: even left-leaning governments face pressure to maintain investor confidence and manage debt. Class dynamics in agrarian economies interact with land relations and peasant reproduction in ways that complicate any straightforward application of Marx's industrial framework (Bernstein, 2010). In short, for developing countries, anti-monopoly policy is structurally necessary for wage growth. Competition policy, industrial policy that supports domestic competitors, and deliberate effort to reverse concentration trends are prerequisites for improving labor's economic position.

The COVID-19 pandemic represents the most significant crisis-to-political-response event in recent decades; it is a natural experiment for H10. The core panel ends in 2019, but ATUS data extend to 2024 and WVS to 2022, capturing immediate consequences. The pandemic's political aftermath reveals the diagnostic-prognostic gap sharply. The crisis itself, with significant demand destruction and wage-price stressors validated Marx's crisis analysis. But the political response was heterogeneous: aggressive stimulus spending in some nations, authoritarian consolidation in others, populist backlash across regions. This heterogeneity reinforces the central finding: the crisis was real (Marx's diagnosis), but political responses reflect institutional legacies and elite choices rather than structural necessity (undermining Marx's prognosis).

The diagnostic–prognostic gap is not unique to Marx. Modernization theory predicted urbanization and industrialization; it incorrectly thought that these would produce stable democracies. Dependency theory correctly identified extractive international economic relationships; it incorrectly predicted revolutionary nationalism. Structuralist frameworks in development economics share a common vulnerability: they are stronger at identifying systemic tendencies than at predicting political responses. This recurrence suggests something ontological rather than methodological. Economic structures constrain the space of possible outcomes but do not determine which outcome obtains. The mediating variables operate at a level of specificity that structural theories cannot reach.

For development policy, this suggests Marx's strength and weakness occupy the same terrain. His structural diagnosis (identifying how accumulation, competition, and class relations operate systematically) is powerful and empirically warranted. Yet his political prognosis is systematically weaker. Structural diagnosis must be paired with institutional design to move from understanding to transformation. Acemoglu and Robinson (2006) argue that inclusive economic institutions require inclusive political institutions; Mahoney and Thelen (2010) theorize how institutions change incrementally under structural pressure; North (1990) emphasizes how path dependence constrains reform; Rodrik (2011) frames the tension between these perspectives as a political trilemma. This analysis contributes by systematically testing that structural explanation has genuine but bounded power: it indicates what is happening and why, but not what will be done about it.

Limitations

This paper operationalizes Marx's theoretical system as ten independent hypotheses, but Marx conceived these mechanisms as deeply interdependent: falling profit rates intensify exploitation, which swells the reserve army, which depresses wages and fragments solidarity, which in turn blocks class consciousness. Scoring each hypothesis separately sacrifices these systemic feedbacks for diagnostic precision. If the political predictions (Cluster C) fail partly because they depend on interactions among the economic mechanisms (Clusters A and B) that the additive rubric cannot capture, the paper may be understating Marx's political theory by design. The 2000-2019 window compounds this problem, testing 19th-century predictions against a period when welfare states, financialization, and digital platforms have fundamentally altered the transmission channels Marx theorized. A longer time horizon or a structural equation approach modeling the full causal chain might yield different results. The design scores each hypothesis independently because treating the system as a single joint hypothesis would reduce the exercise to a binary verdict, but recognizing that the systemic interdependencies Marx theorized cannot be captured in this analysis.

Several individual specifications face measurement and identification challenges that constrain interpretation. The commodification hypothesis (H6) relies on ATUS categories designed before the digital economy blurred the boundary between leisure and value production, likely undercounting the phenomenon Marx described. The alienation mediation (H7) rests on 34 country-year observations, too few to sustain confident causal claims. The state-as-instrument test (H9) captures Miliband's instrumentalist version but leaves Poulantzas's structuralist alternative untested, potentially missing the more powerful channel. The class consciousness null (H8) may partly reflect the inadequacy of left vote share as a proxy for political mobilization, since contemporary class politics often operates outside electoral channels entirely. And throughout, the scoring rubric's preferred weights carry an element of subjectivity, with the sensitivity analysis showing that alternative schemes shift individual scores without overturning the headline pattern. Every specification in this paper uses publicly available data as Marx's claims concern the fundamental dynamics of capitalism, and the data available to test them should be accessible to any researcher seeking to replicate, extend, or challenge the results.


## 6. Conclusion and Future Research

Given the extraordinary scope of his claims and the 150 years since he made them, Marx was indeed prolific, ambitious, and often right. Marx's economic diagnostics describe the 21st-century global economy with precision. The decline of labor's share is a fact, and the gig economy's wage-suppressing effects are observable in the RDD estimates. The automation–alienation channel is measurable in cross-national survey data spanning 108 countries. Marx's political predictions, by contrast, rest on a transmission mechanism from conditions to consciousness to collective action that the data show to be inoperative or far more contingent than his theory allows. Rising inequality does not predict left voting, and exploitation does not produce class solidarity.

Four directions for future research follow. First, the diagnostic–prognostic gap would benefit from systematic investigation: under what circumstances do material grievances translate into class-based political action rather than identity-based, populist, or individualized responses? Second, the counter-tendency dynamics documented in H1 and H3 deserve panel-level investigation across development stages; does the digital economy's simultaneous concentration and fragmentation represent a qualitatively new form of accumulation? Third, the alienation channel (H7) requires longer panels and more granular occupational data across welfare regimes. Fourth, the structural state theory (Poulantzas) could deliver a dedicated empirical test alongside the instrumentalist specification through capital strike dynamics, tax competition, or fiscal policy convergence across ideologically different governments. These findings suggest that addressing exploitation, alienation, and the decline of labor's share requires deliberate institutional design: progressive taxation, labor market regulation, collective bargaining frameworks, platform governance, and automation transition policies. In developing economies, where these dynamics operate with particular force and institutional buffers are thinnest, the political will to address them does not emerge automatically from those conditions. That most governments will not pursue such designs is precisely the problem Marx identified, and precisely where his solution falls short.

Declaration of generative AI & AI-assisted technologies in the manuscript preparation process

During the preparation of this work the author(s) used Claude CoWork to build tables and appendices, cross-check findings, format references, and act as a pre-peer reviewer before submission. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.

---

## Citation

Jason Miklian. "Right Diagnosis, Wrong Prescriptions? Testing Ten of Karl Marx's Hypotheses Against Twenty-First-Century Data." *Under Review*, 2026.

## About the Author

Jason Miklian is Senior Researcher at the Centre for Development and the Environment (SUM), University of Oslo. ORCID: [0000-0003-1227-0975](https://orcid.org/0000-0003-1227-0975). Google Scholar: [profile](https://scholar.google.com/citations?user=RHlevGEAAAAJ&hl=en).