Emergence of Statistical Financial Factors by a Diffusion Process
Jose Negrete Jr, Jaime Joel Ramos · 2026
Read Emergence of Statistical Financial Factors by a Diffusion Process by Jose Negrete Jr, Jaime Joel Ramos (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-12197.txt.
Introduction
Negrete and Ramos build a network model in which financial factors emerge from asset interactions rather than being imposed by statistics. Assets follow coupled iterated maps tied to past returns of their own and of related assets, with links growing from isolated clusters to a fully connected network through a transformed Laplacian matrix. Stable co-movement patterns then read as factors, and one regime explains asset variance well. The link between initial clustering and factor count matches a center manifold reduction.
Catalog record
| CREATOR | Jose Negrete Jr, Jaime Joel Ramos |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30151 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.12197
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Emergence of Statistical Financial Factors by a Diffusion Process Jose Negrete Jr, Jaime Joel Ramos Abstract Factor models characterize the joint behavior of large sets of financial assets through a smaller number of underlying drivers. We develop a network-based framework in which factors emerge naturally from the structure of interactions among assets rather than being imposed statistically. The market is modeled as a system of coupled iterated maps, where assets’ return depends on its own past returns and those of related assets. Effectively modeling the influence of irrational traders whose decisions are based on the past movements of a collection of stocks. The interaction structure between stock returns is defined by a coupling matrix derived from an orthogonal transformation of a Laplacian matrix that gradually links initially isolated clusters into a fully connected network. Within this structure, stable patterns of co-movement arise and can be interpreted as financial factors. The relationship between the initial clustering and the number of observed factors is consistent with a center manifold reduction. We identify an optimal regime in which assets’ variance is effectively explained by the set of factors produced by the network. Our framework offers a structural perspective based on interaction-based factor formation and dimension reduction in financial markets. Factors corresponds to a reduced number of signals that describes the time evolution of a large cross-section of stocks. Here we wonder what might be their possible origin. One possibility is that stocks reflect the movements dictated by exogenous signals, this would be in agreement with the efficient market hypothesis (EMH) where movements in the stock prices reflect the random movements in the economy. An opposing view of the EMH is that traders are irrational and make their decisions given the previous movements of a stock i.e. behaving as a dynamical system. We use a combination of dynamical systems and network theory to propose that factors might emerge from the activity given by irrational traders. I Introduction Factor models are central to empirical finance. They summarize co-movements across many assets through a small set of underlying drivers. These underlying drivers determine the systemic risk behind an investment portfolio i.e. the noise that cannot be mitigated by diversification 32 . A cross-section of K stocks with returns r_{t}=\left[r_{t}^{\left(1\right)},r_{t}^{\left(2\right)},\cdots,r_{t}^{\left(K\right)}\right] is reduced to M common factors f_{t}=\left[f_{t}^{\left(1\right)},f_{t}^{\left(2\right)},\cdots,f_{t}^{\left(M\right)}\right] . The relationship between stock returns and latent factors is given by r_{t}=\mathbf{B}f_{t}+\xi_{t}. (1) Here M\ll K , and \mathbf{B} is an M\ \times K loading matrix whose elements \beta_{mk} quantify the exposure of each stock to a given factor. The vector \xi_{t} accounts for any time deviations (or noise) not explained by the factors. Factors are divided into two classes: economic and statistical factors. The first economic factor model proposed was the Capital Asset Pricing Model (CAPM) ( 42 ) , where stock returns depend on the market factor. This model gives a mathematical definition for the risk premium, which is the excess return a stock has as a function of risk. However, independent tests with empirical data does not fulfill the CAPM predictions between risk and return ( 23 ) . Extensions to CAPM have been proposed where different fundamentals like firm size and book-to-market equity are taken to construct different factors ( 18 ; 19 ) . Alternatively, statistical factors correspond to a set of common orthogonal signals that are intrinsic to a cross section of stocks ( 13 ; 16 ) . Techniques like principal components analysis (PCA) and variations of it are used to infer factors from observed returns without explaining their structural origin. Several authors have used random matrix theory to determine the number M of factors behind a given cross section of stocks ( 27 ; 39 ; 38 ) . Interestingly, the CAPM market factor displays a strong correlation with the first component extracted via PCA ( 4 ) . Here we ponder on why factors emerge in the first place? According to the efficient market hypothesis (EMH) stock prices depend only on the fundamentals behind a firm. In EMH the stock prices are only driven by random exogenous signals defined by the economy 32 . Still, deviations from fundamental values are observed in trading markets which are driven by investors exhibiting herding behavior or by the use of investment strategies like hedging 44 . These instances creates feedback mechanisms between investors and the market that are well described by dynamical systems. Several works have explored the effect of different feedback mechanisms such as trend following 36 , adaptation of expectations 3 ; 11 and leverage 47 . This perspective motivates a shift: rather than treating factors as externally imposed or purely statistical artifacts, we treat and understand them as emergent properties of a dynamical system. Modern financial markets are densely interconnected via shared information channels, overlapping investor bases, sectoral relationships, and common economic exposures. If assets influence one another over time, coherent return patterns (i.e. factors) can arise naturally from these interactions. Matrices can be used to define the structure of the interconnections between stocks. Previous works suggest the existence of self-organization in finance ( 8 ; 10 ) , as transitions to a highly coordinated states were observed in both the dot-com and housing crisis. These are instances of dimensionality reduction that can be achieved by two different connectivity structures. In one case the structure is described by a low-rank matrix 46 where the rank equals the reduced dimension ( 33 ) . The alternative case is a center manifold reduction ( 20 ) where the matrix nullity equals to the reduced dimension. The specific nonlinearities of a system determines if the rank or the nullity of a matrix determines the reduced dimension 33 ; 20 . We propose a network-based model that formalizes this structure and offers insight into why certain factors appear from the perspective of dimensionality reduction. II The Model II.1 A Chaotic Model For Asset Returns Inter-day returns time series typically exhibit an absence of autocorrelation and heavy-tailed distributions ( 17 ) . Also there is evidence that financial time series evolve with some degree of chaotic dynamics ( 22 ; 26 ) . Motivated by these, we model the evolution of r_{t} using coupled iterative maps, which provide a suitable structural approach to deterministic dynamics in interacting systems. A general coupled iterative map 37 ; 48 takes the form r_{t+1}=\left(1-\varepsilon\right)g\left(r_{t}\right)+\frac{\varepsilon}{N}\mathbf{C}g\left(r_{t}\right) (2) where \mathbf{C} is a K\ \times K connectivity matrix that determines how the K components of r_{t} influence each component of r_{t+1} . The parameter N denotes the number of elements within each cluster (defined below). The coupling parameter \varepsilon controls the components’ degree of independence/dependence: when \varepsilon=\ 0 each asset evolves independently according to r_{t+1}^{\left(k\right)}=g\left(r_{t}^{\left(k\right)}\right), (3) whereas for \varepsilon=\ 1 , every component is fully influenced by all other through a given weight and the following mapping r_{t+1}=\frac{\mathbf{C}}{N}g\left(r_{t}\right). (4) II.2 The Local Map g\left(r_{t}\right) The mapping g\left(r_{t}\right) is defined by the following steps: u_{t}=h\left(r_{t}\right) (5) u_{t+1}=\frac{u_{t}}{\delta}-\left\lfloor\frac{u_{t}}{\delta}\right\rfloor (6) r_{t+1}=h^{-1}\left(u_{t+1}\right). (7) In this construction, Eq.( 6 ) corresponds to the Bernoulli map, a well-known source of chaotic dynamics ( 35 ) . The variable u_{t} is uniformly distributed on \left[0,1\right] , and therefore its probability density function (pdf) is \rho\left(u_{t}\right)=U\left[0,1\right] . This allows full control