Paper: arXiv 2509.10553
Authors: Erina Nanyonga, Juma Kasozi, Fred Mayambala, Hassan W. Kayondo, Matt Davison
Abstract
This study explores the behavioral dynamics of illiquid stock prices in a listed stock market. Illiquidity, characterized by wide bid and ask spreads affects price formation by decoupling prices from standard risk and return relationships and increasing sensitivity to market sentiment. We model the prices at the Uganda Securities Exchange (USE) which is illiquid in that the prices remain constant much of the time thus complicating price modelling. We circumvent this challenge by combining the Markov model (MM) with two models; the exponential Ornstein Uhlenbeck model (XOU) and geometric Brownian motion (gBm). In the combined models, the MM was used to capture the constant prices in the stock prices while the XOU and gBm captured the stochastic price dynamics. We modelled stock prices using the combined models, as well as XOU and gBm alone. We found that USE stocks appeared to have low correlation with one another. Using theoretical analysis, simulation study and empirical analysis, we conclude that this apparent low correlation is due to illiquidity. In particular data simulated from combined MM-gBm, in which the gBm portion were highly correlated resulted in a low measured correlation when the Markov chain had a higher transition from zero state to zero state.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 3.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper employs advanced stochastic modeling techniques (Markov chains combined with SDEs like XOU and gBm) requiring significant mathematical theory, but the empirical component is primarily theoretical and simulation-based, lacking the backtest-ready implementation details or heavy data processing typical of production-ready quant strategies.
Research Flowchart
flowchart TD
A["Research Goal<br>Model illiquid stock prices & explain<br>apparent low correlation at USE"] --> B{"Data & Inputs"}
B --> B1["USE historical stock data<br>(characterized by price constancy)"]
B --> B2["Simulation data<br>(controlled correlation)"]
subgraph C ["Key Methodology Steps"]
C1["Combine Models<br>Markov + Exponential O-U + gBm"]
C2["Compare Models<br>Combined vs. gBm alone"]
C3["Correlation Measurement<br>via simulation & empirical analysis"]
end
A --> C1
C1 --> C2
C2 --> C3
subgraph D ["Computational Processes"]
D1["Capture constant prices<br>Markov Chain Component"]
D2["Capture stochastic dynamics<br>XOU & gBm Component"]
D3["Measure correlation under<br>varying illiquidity parameters"]
end
C1 --> D1
C1 --> D2
C3 --> D3
D3 --> E["Key Findings & Outcomes"]
E --> E1["Illiquidity decouples prices from<br>risk/return & increases sentiment sensitivity"]
E --> E2["Apparent low correlation is<br>an artifact of illiquidity"]
E --> E3["Verified via simulation:<br>High correlation input yields low measured output<br>under high Markov persistence"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes combining a Markov model with exponential Ornstein‑Uhlenbeck and geometric Brownian motion models to capture constant and stochastic price movements in illiquid stocks.
Source sentence
We circumvent this challenge by combining the Markov model (MM) with two models; the exponential Ornstein Uhlenbeck model (XOU) and geometric Brownian motion (gBm).
Assumes 2
- The current state of the Markov chain is sufficient to predict the next state. model
Source sentence
A Markov chain is based on the assumption that the current state alone is sufficient to predict the next state.
- Illiquid stocks have long periods of constant prices, leading to low observed correlation. market
Source sentence
Illiquidity, characterized by wide bid and ask spreads affects price formation by decoupling prices from standard risk and return relationships and increasing sensitivity to market sentiment.
Data
- Stanbic Bank Uganda (SBU) historical prices market-prices public
Source sentence
Data for the listed stocks that we considered at USE was freely downloaded from the Wall Street Journal (WSJ) website; https://www.wsj.com/market-data/quotes/UG/XUGA/SBU/historical-prices for Stanbic Bank Uganda (SBU)
- Bank of Baroda Uganda (BOBU) historical prices market-prices public
Source sentence
https://www.wsj.com/market-data/quotes/UG/XUGA/BOBU/historical-prices for bank of Baroda Uganda (BOBU) historical prices
- Development Finance Company of Uganda (DFCU) historical prices market-prices public
Source sentence
https://www.wsj.com/market-data/quotes/UG/XUGA/DFCU/historical-prices for Development Finance Company of Uganda Ltd (DFCU) historical data
- Uganda All Share Index (ALSIUG) market-prices public
Source sentence
Data for the USE All Share Index (ALSIUG), the main stock market index in Uganda, was downloaded from investing.com on the following url; https://www.investing.com/indices/uganda-all-share-historical-data.
- Kenya Commercial Bank (KCB) historical prices market-prices public
Source sentence
and Kenya Commercial Bank (KCB); https://www.wsj.com/market-data/quotes/KE/XNAI/KCB/historical-prices
Costs
Not applicable
Finds 5
- The MM‑gBm model outperforms the other models, especially for short forecasting horizons, yielding the smallest MAPE values. performance
Source sentence
The MM-gBm outperforms the other models while modelling closing prices of securities at USE especially for a short time period.
- USE stocks exhibit low measured correlation, which the authors attribute to illiquidity and frequent constant price periods. statistical
Source sentence
We found that USE stocks appeared to have low correlation with one another. Using theoretical analysis, simulation study and empirical analysis, we conclude that this apparent low correlation is due to illiquidity.
- Simulation results show that when the Markov chain stays longer in the zero‑price state, even highly correlated gBm components appear weakly correlated in observed data. theoretical
Source sentence
In particular data simulated from combined MM-gBm, in which the gBm portion were highly correlated resulted in a low measured correlation when the Markov chain had a higher transition from zero state to zero state.
- Forecasting accuracy (MAPE) for the four models on SBU data before COVID‑19: 0.2413% (MM‑gBm), 0.8487% (gBm), 0.3202% (MM‑XOU), 0.4872% (XOU). performance
Source sentence
... with MAPE values; 0.2413%, 0.8487%, 0.3202% and 0.4872% for MM-gBm, gBm, MM-XOU and XOU respectively.
- For illiquid local stocks (SBU, BOBU) the observed correlation is very low and sometimes negative, despite being in the same sector. statistical
Source sentence
SBU and BOBU are in the same sector and were expected to be highly correlated but it is not the case in our results
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 58 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.