On the Mean-Entropy Frontier: Profit Rate Equalization

Technical Summary: On the Mean-Entropy Frontier: Profit Rate Equalization

Problem Statement
The paper addresses the difficulty of empirically assessing the functioning of capitalist competition, particularly in the context of rising market concentration, intellectual property rights, and declining investment despite high profitability. While existing literature offers conflicting views—ranging from the "monopoly" hypothesis (declining competition) to the "superior performance" hypothesis (Demsetz, 1973)—robust evidence is hindered by a lack of a broadly accepted theoretical framework for competition.

Specifically, the paper critiques the limitations of traditional mean-variance approaches (e.g., Glick and Ehrbar, 1990) which rely on variance as a risk measure, noting conceptual and implementation challenges. It also highlights that while information-theoretic approaches (e.g., Farjoun and Machover, 1983; Scharfenaker and Foley, 2017) offer inferential simplicity, they often abstract from industry classification or focus on firm-level data grouped into bins, potentially misaligning the statistical object with the theoretical level at which profit-rate equalization operates (industry/sector level). Furthermore, previous probabilistic models often lack transparent mechanisms to evaluate how far observed allocations deviate from an efficient benchmark.

Methodology
The author develops a maximum-entropy model of profit-rate equalization, treating capital allocation as an information-theoretic inference problem.

1. Theoretical Framework:

   • The model defines the weighted average profit rate ($\mu$) as the sum of industry profit rates ($r_i$) weighted by capital shares ($x_i$).
   • It utilizes the principle of maximum entropy to infer the equilibrium distribution of capital shares. The objective is to maximize the entropy of weighted profit-rate shares relative to the weighted average, subject to constraints:
     • Normalization: $\sum x_i = 1$.
     • Expected Return: $\sum x_i r_i = \mu$.
     • Prior Information: The distribution should be closest to a prior distribution ($\pi_i$), represented by industry sales shares.
   • The solution yields an allocation rule where capital shares depend on observed profit rates, prior sales shares, and a scalar allocation parameter ($\alpha$) that acts as a return-dependent tilt. The rule is given by:
     $$x_i \propto \pi_i r_i^{-1} \exp(-\alpha/r_i)$$
   • This formulation minimizes the Kullback-Leibler divergence between the implied returns distribution and the prior, rather than minimizing variance.

2. Data:

   • The study uses U.S. firm-level data from the Compustat/CRSP Annual North American Fundamentals database (1962–2024).
   • Industries are classified by three-digit SIC codes (excluding regulated utilities, financials, and government).
   • Profit rates are calculated as Operating Income Before Depreciation (OIBDP) divided by Total Assets (AT). Industries with negative profit rates are excluded.
   • Capital shares are the proportion of industry total assets to aggregate assets; the prior distribution is constructed from industry sales shares.

3. Estimation:

   • The model is estimated using Bayesian inference via Hamiltonian Monte Carlo (NUTS sampler in Stan).
   • Observed and predicted capital shares are transformed using additive log-ratio (ALR) coordinates to handle the compositional nature of the data.
   • The allocation parameter $\alpha_t$ is estimated separately for each year, while the error scale $\sigma$ is pooled across years.
   • The posterior distribution of $\alpha_t$ generates a distribution over model-implied capital allocations, weighted average profit rates, and normalized entropy values.

4. The Mean-Entropy Frontier:

   • The paper constructs a "mean-entropy frontier," analogous to the Markowitz mean-variance frontier. This frontier represents the set of attainable combinations of weighted average profitability ($\mu$) and normalized entropy ($H$) given the observed profit rates and priors.
   • An "equalization gap" is defined as the distance between the observed allocation point ($H_{obs}, \mu_{obs}$) and the maximum attainable entropy on the frontier at a comparable profitability level.

Key Results

• Secular Trends: The analysis reveals a long-run decline in the model-implied weighted average profitability ($\mu_{imp}$) from the 1960s through the early 2000s, with no subsequent recovery to earlier levels.
• Equalization Dynamics: Concurrent with the decline in profitability, there is a secular increase in the normalized entropy measure ($H_{imp}$), indicating a trend toward greater equalization of weighted profit rates across industries. This trend stabilizes after the early 2000s.
• The Equalization Gap: The estimated gap between observed and model-implied entropy narrows substantially over time. In recent decades, the observed allocation lies closer to the model-implied frontier, suggesting limited room for further gains from additional equalization at comparable levels of profitability.
• Sectoral Deviations:
  • Industries with the largest positive deviations (observed capital shares > predicted) are asset-intensive sectors such as communications, pipelines, health services, and equipment rental.
  • Industries with the largest negative deviations (observed capital shares < predicted) are concentrated in wholesale and retail distribution sectors (e.g., groceries, drugs), where sales shares exceed capital shares.
  • These deviations reflect persistent heterogeneity in asset requirements across production, service, and distribution sectors.

Significance and Contributions
The paper claims three primary contributions to the literature on profit-rate equalization and capitalist competition:

1. Analytical Framework: It provides a tractable maximum-entropy model that replaces the variance-minimization logic of portfolio theory with entropy maximization (or KL-divergence minimization). This approach incorporates prior information (sales shares) and treats capital allocation as a probabilistic inference problem, avoiding high-dimensional mean-variance structures.
2. Empirical Object Alignment: By shifting the focus from firm-level profit rates to industry-level rates, the analysis aligns the statistical object with the theoretical level at which profit-rate equalization is expected to operate, addressing a gap noted by Basu (2018).
3. Diagnostic Tool: The construction of the "mean-entropy frontier" offers a new metric for evaluating competition. It allows researchers to measure the "equalization gap," quantifying the extent to which observed capital allocations fall short of the efficient benchmark implied by the model. This provides a probabilistic framework to study the relationship between profitability and the degree of competition without relying on restrictive equilibrium assumptions.

The paper concludes that the observed decline in U.S. profitability is closely associated with the equalization process, and that the economy has moved closer to the model-implied efficient frontier in recent decades, implying that low profitability may be a persistent structural feature rather than a temporary deviation.