Public repository: https://github.com/QuantLet/FDM.
Reproduction repository for the chapter Innovations in FinTech and Digital Markets, in the Springer volume Digital Finance: Reaching New Frontiers.
Daniel Traian Pele
- Department of Statistics and Econometrics, Bucharest University of Economic Studies, Piata Romana 6, Bucharest, Romania.
- Institute for Economic Forecasting, Romanian Academy, Casa Academiei, Calea 13 Septembrie 13, Sector 5, 050711 Bucharest, Romania.
- Institute for Digital Assets (IDA), Bucharest University of Economic Studies, Calea Dorobanților 15–17, 010552 Bucharest, Romania.
Prepared within the MSCA Industrial Doctoral Network on Digital Finance, funded by the European Union's Horizon Europe research and innovation programme under Marie Skłodowska-Curie grant agreement No. 101119635.
The empirical tables, figures and results in the chapter are produced by the four Quantlets below from cached public data. Reproduction from the complete cache requires no new API calls; fresh downloads depend on current access conditions.
| Quantlet | What it does | Key outputs |
|---|---|---|
FDMdata |
Downloads and aligns eight daily series (FRED + NASDAQ), builds the return panel, reports descriptives | data/prices.csv, data/returns.csv, 3 figures |
FDMbiblio |
Maps the research frontier from OpenAlex: six innovation domains, take-off dating, concept co-occurrence network | domain counts and shares, centrality, 3 figures |
FDMconnect |
Diebold–Yılmaz connectedness over the eight-asset system, static and rolling | spillover table, block matrix, 3 figures |
FDMtails |
Quantile connectedness and non-parametric lower tail dependence | quantile networks, tail dependence, 2 figures |
Each Quantlet folder contains a Metainfo.txt descriptor, the .py source, an
executed .ipynb notebook with stored outputs, and its result files.
- Three waves of innovation. Digital intermediation takes off in 2012, blockchain and crypto in 2014, DeFi and tokenisation in 2018, CBDC in 2020, generative AI in 2022. Louvain modularity of 0.485 indicates thematic clustering within the selected concept network; it does not validate the query boundaries or dates.
- Listed FinTech equity is closely linked to equity benchmarks. It imports 60.6% of its forecast error variance from traditional markets, against 8.0% from digital assets.
- Digital-asset connectedness increases at extreme quantiles. Bitcoin and Ether retain 79.9% of their variance within their own block at the median, but import 63.8% from the rest of the system at the 5% conditional quantile. Total connectedness rises by 25.8 percentage points from the median to the lower tail.
git clone https://github.com/QuantLet/FDM.git
cd FDM
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
make data # FDMdata -> data/prices.csv, data/returns.csv
make analysis # FDMbiblio, FDMconnect, FDMtails
make notebooks # re-execute all four notebooks with stored outputs
make all # data and all three analysesFDMdata must run first: the other market Quantlets read data/returns.csv.
| Source | Series | Access |
|---|---|---|
| FRED (St. Louis Fed) | CBBTCUSD, CBETHUSD, SP500, NASDAQCOM, VIXCLS, DTWEXBGS, DGS10 |
key-free CSV endpoint |
| NASDAQ | FINX (Global X FinTech ETF) |
public quote endpoint |
| OpenAlex | economics and management corpus, 2010–2025 | cached API responses |
All downloads are cached under data/raw. A second run performs no network
traffic and reproduces byte-identical inputs. Two transport quirks are handled
in FDMdata: FRED stalls when sent a spoofed browser User-Agent, while the
NASDAQ endpoint requires one.
Data vintage 27 July 2026; seeds fixed at 20260727. FRED and OpenAlex
revise their contents, so a later download may use different inputs. Re-running
against the committed data/raw preserves the reference data vintage.
The committed CSV files and executed notebooks preserve the chapter's reference
results. In a verification run with Python 3.12.13, NumPy 2.5.1, pandas 3.0.5,
SciPy 1.18.0, statsmodels 0.14.6, NetworkX 3.6.1 and Matplotlib 3.11.1, all four
modules ran with network downloads blocked. Of the 26 output CSV files, 23
matched within a numerical tolerance of 1e-8 absolute and 1e-7 relative;
three files (FDMtails_blocks_q05.csv, FDMtails_gfevd_q05.csv and
FDMtails_summary.csv) differed by at most 0.01 percentage points. Re-execution
in another numerical environment can therefore affect the last reported decimal.
Python 3.11+. Python dependencies are in requirements.txt; note that the
conditional-mean VAR and moving-average calculations are implemented in NumPy
and validated against statsmodels inside FDMconnect, so no specialist econometrics
package is required.
Pele, Daniel Traian (2026). FDMdata, FDMbiblio, FDMconnect, FDMtails: reproducible Quantlets for Innovations in FinTech and Digital Markets. QuantLet. https://github.com/QuantLet/FDM
The cached data retain the attribution and terms of their original providers: FRED, NASDAQ and OpenAlex. See the data-source table above and the four Metainfo.txt files for provenance.