READ ME File For 'Essays on the Role of Memory in Financial Markets – Supporting Dataset'

Dataset DOI: 10.5258/SOTON/D3837

ReadMe Author: Dmitri Mustanen, University of Southampton ORCID ID https://orcid.org/0009-0008-7918-7790

This dataset supports the thesis entitled 'Essays on the Role of Memory in Financial Markets'
AWARDED BY: University of Southampton
DATE OF AWARD: 2026

Date of data collection: 
Chapter 2: 14-01-2023
Chapter 3: 10-10-2023
Chapter 4: 02-07-2025 

Information about geographic location of data collection: 
Not applicable in the conventional fieldwork sense. The datasets comprise secondary financial market time series obtained electronically from publicly available sources and commercial data providers, and processed for the thesis analyses.

Licence:
Creative Commons Attribution (CC BY). 

Related projects/Funders:
University of Southampton (Southampton Business School) – Doctoral College research scholarship. The research was not supported by any external grant funding.


--------------------
DATA & FILE OVERVIEW
--------------------

This dataset contains:

CH2_Data.xlsx
  Chapter 2 dataset for “Improving System Generalisation & Forecastability via Spillover-Based Variable Selection Approach.”
  Includes Tabs: (i) main analysis raw panel, (ii) z-score normalised panel used for modelling, and (iii) a robustness-check dataset used for comparison.

CH3_Data.xlsx
  Chapter 3 dataset for “Forward and Backward Memory in Fractionally Cointegrated Systems.”
  Includes Tabs: (i) oil-market system (Brent/WTI spot + PCA factors) and (ii) cryptocurrency system (crypto prices + macro/uncertainty/liquidity proxies) used for bidirectional memory analysis.

CH4_Data.xlsx
  Chapter 4 dataset for “A Memory-driven Multi-period Capital Asset Pricing Model”
  Includes US and Japan equity universes with (i) price panels, (ii) return panels, and (iii) memory (fractional integration parameter d) panels used in the empirical tests.


Relationship between files, if important for context:  
Each Excel file corresponds to a self-contained thesis chapter and contains the data required to reproduce the chapter’s empirical analysis (and its core summary tables/figures), including any derived panels (normalised variables, PCA factors, returns, and memory estimates). 
The files are independent but share the thesis-level methodological theme of modelling system dynamics and memory in financial time series.

Additional related data collected that was not included in the current data package:
Large intermediate data pulls (e.g., full Bloomberg futures-chain extractions) and intermediate transformation objects used during exploratory processing are not included. The deposited files contain the final analysis panels required for replication.

If data was derived from another source, list source:
Chapter 2:
  Commercial data (e.g., Bloomberg) for oil spot and futures-curve series; futures-curve derived measures (Expansion, Liquidity, Regime) and PCA factors are constructed as described in the thesis.
Chapter 3:
  Commercial data (e.g., Bloomberg) for oil and related financial variables; publicly available sources for cryptocurrency prices and uncertainty indices/proxies (as listed in the thesis chapter data sources).
Chapter 4:
  Publicly available and/or commercial index/asset series (as described in Chapter 4) used to build US and Japan equity universes; returns and memory estimates are computed from those series following the methods documented in the thesis.

If there are there multiple versions of the dataset, list the file updated, when and why update was made:
Not applicable. The deposited files correspond to the final thesis supporting dataset for deposit and DOI registration.


--------------------------
METHODOLOGICAL INFORMATION
--------------------------

Description of methods used for collection/generation of data:
The dataset is generated as part of the thesis empirical workflow. Data are assembled from market data providers and publicly available sources and transformed into analysis panels aligned to each chapter’s identification strategy:
  Chapter 2 constructs a system of oil-market variables combining observed spot prices with futures-curve derived measures and global PCA factors; the system is used for spillover estimation, variable selection, and medium-horizon forecasting frameworks.
  Chapter 3 constructs oil and crypto systems to study forward- and backward-looking memory in price formation using the chapter’s time-directional estimation framework and FCVAR-based system modelling documented in Chapter 3.
  Chapter 4 constructs US and Japan equity universes, computes returns, and estimates memory (fractional integration parameter d) used in the chapter’s multi-period asset pricing tests documented in Chapter 4.

Methods for processing the data:

Chapter 2: Data were aligned to a common daily trading calendar. Futures-curve information was transformed into term-structure measures capturing Expansion, Liquidity, and Regime effects, as described in the thesis. Global factors were extracted via principal component analysis (PCA). Variables were z-score normalised to ensure comparability prior to spillover estimation, variable selection, and system-based forecasting.
Chapter 3: Market data were aligned across oil and cryptocurrency systems. Futures-curve information was summarised using PCA-based factors. The resulting systems were used to study forward- and backward-looking memory in price formation using the bidirectional framework documented in the thesis.
Chapter 4: Equity price series were aligned by trading calendar and transformed into return series. Memory parameters (fractional integration parameter d) were estimated for each asset using the frequency-domain approach described in the thesis. These processed panels form the basis for the empirical evaluation of the memory-driven multi-period asset pricing framework.

Software- or Instrument-specific information needed to interpret the data, including software and hardware version numbers:
The deposited files are standard Microsoft Excel workbooks (.xlsx) and can be opened with Excel or compatible software (e.g., LibreOffice). No specialised instrument output formats are used.

Standards and calibration information, if appropriate:
Not applicable.

Environmental/experimental conditions:
Not applicable (secondary financial market datasets).

Describe any quality-assurance procedures performed on the data:
Quality assurance follows the thesis diagnostics and includes checks appropriate for time-series modelling (e.g., stationarity/unit-root diagnostics prior to cointegration-based modelling in Chapter 2; robustness and diagnostic checks referenced in Chapter 3; and summary-statistic validation and panel integrity checks for the equity universes in Chapter 4). Missingness arises primarily from market holidays, series availability, and estimation-window requirements.

People involved with sample collection, processing, analysis and/or submission:
Dmitri Mustanen (data assembly, processing, analysis, submission). Supervisory oversight and guidance are acknowledged in the thesis (Professor Taufiq Choudhry; Dr Ahmad Maaitah).


--------------------------
DATA-SPECIFIC INFORMATION
--------------------------

CH2_Data.xlsx (Chapter 2)

Number of variables:
Main Analysis - Raw Data: 12 columns (including DATE)
Main Analysis - Normalized Data: 12 columns (including DATE)
Robustness Check - Data: 12 columns (including DATE)

Number of cases/rows: 
Main Analysis - Raw Data: 5,014
Main Analysis - Normalized Data: 5,014
Robustness Check - Data: 1,121

Variable list, defining any abbreviations, units of measure, codes or symbols used:
Main Analysis - Raw Data / Normalized Data
  DATE  Trading date
  BSP   Brent spot price
  WSP   WTI spot price
  BE    Brent term-structure “Expansion” factor
  BL    Brent term-structure “Liquidity” factor
  BR    Brent term-structure “Regime” factor
  WE    WTI term-structure “Expansion” factor
  WL    WTI term-structure “Liquidity” factor
  WR    WTI term-structure “Regime” factor
  E     Global PCA “Expansion” factor E
  L     Global PCA “Liquidity” factor L
  R     Global PCA “Regime” factor R

Robustness Check - Data
  DATE  Trading date
  INE   Alternative indicator
  TR    Alternative indicator
  IR    Alternative indicator
  ER    Alternative indicator
  CF    Alternative indicator
  WTI   WTI spot price
  Brent Brent spot price
  SHCI  Shipping cost index
  EI    Energy index
  EPU   Economic Policy Uncertainty index

Missing data codes:
Blank cells indicate missing values. No imputation is applied in the deposited files.

Specialized formats or other abbreviations used:
Chapter-2 abbreviations follow the thesis notation. Constructed term-structure factors (Expansion/Liquidity/Regime) and PCA factors are documented in Chapter 2.

Date that the file was created: January, 2023

--------------------------

CH3_Data.xlsx (Chapter 3)

Number of variables:
OIL Market Data: 6 columns (including DATE)
Cryptocurrency Market Data: 8 columns (including Date)

Number of cases/rows: 
OIL Market Data: 1,324
Cryptocurrency Market Data: 1,786

Variable list, defining any abbreviations, units of measure, codes or symbols used:
OIL Market Data
  DATE  Trading date
  BSP   Brent spot price
  WSP   WTI spot price
  E     PCA-based Global PCA “Expansion” factor E
  L     PCA-based Global PCA “Liquidity” factor L
  R     PCA-based Global PCA “Regime” factor R

Cryptocurrency Market Data
  Date  Trading date
  BITC  Bitcoin price
  ETH   Ethereum price
  EGY   Energy proxy
  INFL  Inflation proxy
  LQT   Liquidity proxy
  HDG   Hedging-demand proxy
  UCY   Uncertainty proxy

Missing data codes:
Blank cells indicate missing values.

Specialized formats or other abbreviations used:
Abbreviations correspond to conceptual channels discussed in Chapter 3 (fundamental vs anticipatory drivers) and are described in the chapter’s data section.

Date that the file was created: October, 2023

--------------------------

CH4_Data.xlsx (Chapter 4)

Number of variables:
US - Asset Price Universe: 61 columns (Date + 60 assets)
US - Returns Universe: 61 columns (Date + 60 assets)
US - Memory Universe: 61 columns (Date + 60 assets)
JP - Asset Price Universe: 61 columns (Date + 60 assets)
JP - Returns Universe: 61 columns (Date + 60 assets)
JP - Memory Universe: 61 columns (Date + 60 assets)

Number of cases/rows: 
US - Asset Price Universe: 5,307
US - Returns Universe: 5,306
US - Memory Universe: 5,247
JP - Asset Price Universe: 3,789
JP - Returns Universe: 3,788
JP - Memory Universe: 3,669

Variable list, defining any abbreviations, units of measure, codes or symbols used:
Each sheet includes:
  Date   Trading date
  Asset columns labelled by ticker symbols (e.g., US tickers such as AAPL; Japanese tickers commonly include .T suffix).
Returns sheets contain daily returns computed from price series (standard log-return/return transformation as used in the thesis).
Memory sheets contain estimated fractional integration parameters (memory, d) for each asset, computed following the Chapter 4 methodology. Memory parameters are unit-free.

Missing data codes:
Blank cells indicate missing values, typically due to market holidays, series availability, and estimation-window requirements for memory estimation.

Specialized formats or other abbreviations used:
Tickers follow standard exchange identifiers. Memory is reported as parameter d (fractional integration).

Date that the file was created: July, 2025