Common_Risk_Factors_in_Cryptocurrency

A reproducible R-based research repository for analyzing cryptocurrency returns through a structured empirical pipeline. The project combines data cleaning, descriptive statistics, cross-sectional predictors, crypto factor construction, grouped market comparison, attention models, macro-sentiment-hashrate models, and blockchain network factor models into one transparent workflow.

This repository is designed not just as a code archive, but as a research handoff system: the outputs are organized, interpretable, reproducible, and ready for thesis writing, portfolio presentation, or methodological defense.

---

Research Focus

This project studies how cryptocurrency returns behave across several complementary dimensions:

• descriptive return dynamics
• cross-sectional return predictors
• factor-style portfolio construction
• meme vs conventional coin comparison
• Google Trends attention effects
• macro-financial, sentiment, and hash-rate drivers
• blockchain network activity factors

The central goal is to move from a raw crypto dataset to a complete analytical framework with exportable tables, figures, and model outputs.

---

What the Project Covers

1. Data ingestion and cleaning

The pipeline loads raw cryptocurrency data, validates variables, interpolates missing values where appropriate, merges a risk-free rate series, and constructs:

• daily panel
• weekly panel
• monthly panel
• market proxy series

2. Descriptive statistics

The project generates:

• summary statistics by frequency
• return distribution plots
• extreme return event tables
• benchmark comparisons against the crypto market and S&P 500
• panel-style market characteristic tables

3. Cross-sectional predictors

The repository evaluates characteristic-sorted portfolio effects using quintile-based analysis for:

• size-related characteristics
• momentum
• volume
• volatility

4. Crypto factor model

The project estimates crypto factor-style series such as:

• CMKT — crypto market factor
• CSMB — size-proxy factor
• CMOM — momentum-related factor

It also tests characteristic-sorted long-short portfolios against those factors.

5. Meme vs conventional comparison

The pipeline builds grouped return series for:

• Conventional coins: BTC, ETH, XRP
• Meme coins: DOGE, SHIB

and compares their performance, market-relative behavior, and drawdowns.

6. Attention models

The repository links cryptocurrency returns to Google Trends attention series across multiple keywords and target return groups.

7. Macro / sentiment / hash-rate models

The project incorporates:

• S&P 500
• Dow Jones
• 10Y Treasury yield
• Gold futures
• Fear & Greed Index
• Bitcoin hash rate

to study how external macro and sentiment conditions relate to crypto returns.

8. Network factor models

The repository builds Bitcoin blockchain network variables from raw JSON files, aggregates them to weekly frequency, constructs a PCA-based network factor, and tests whether on-chain activity helps explain crypto return dynamics.

---

Project Structure

Common_Risk_Factors_in_Cryptocurrency/
├── README.md
├── Common_Risk_Factors_in_Cryptocurrency.Rproj
├── addin.R
├── sync.R
│
├── assets/
├── code/
│   ├── 00_setup.R
│   ├── 01_load_and_clean_data.R
│   ├── 02_descriptive_statistics.R
│   ├── 03_cross_sectional_predictors.R
│   ├── 04_crypto_factor_model.R
│   ├── 05_memecoin_vs_conventional_analysis.R
│   ├── 06_google_trends_attention_models.R
│   ├── 07_macro_sentiment_hashrate_models.R
│   ├── 08_network_factors_models.R
│   ├── 09_export_tables_and_figures.R
│   └── 10_run_all.R
│
├── data/
│   ├── raw/
│   └── processed/
│
├── env/
├── memo/
├── paper/
│
└── outputs/
    ├── figures/
    │   ├── google_trends/
    │   ├── conventional_assets_vs_coin_market.png
    │   ├── crypto_factor_cumulative_returns.png
    │   ├── crypto_vs_coin_market.png
    │   ├── crypto_vs_coin_market_sp500.png
    │   ├── generic_long_short_returns.png
    │   ├── generic_quintile_returns.png
    │   ├── macro_sentiment_hashrate_correlation_heatmap.png
    │   ├── meme_vs_conventional_drawdowns.png
    │   ├── meme_vs_conventional_vs_market.png
    │   ├── network_factor_correlation_heatmap.png
    │   ├── network_pca_factor_vs_returns.png
    │   └── return_distributions.png
    │
    ├── logs/
    │   ├── pipeline_run_log.csv
    │   └── pipeline_run_summary.txt
    │
    ├── models/
    ├── tables/
    │   ├── crypto_factor_correlations.csv
    │   ├── crypto_factor_model_results.csv
    │   ├── crypto_factor_model_results.html
    │   ├── crypto_factor_returns.csv
    │   ├── extreme_return_events.csv
    │   ├── group_return_summary_statistics.csv
    │   ├── group_return_summary_statistics.html
    │   ├── macro_sentiment_hashrate_correlations.csv
    │   ├── macro_sentiment_hashrate_model_results.csv
    │   ├── macro_sentiment_hashrate_model_results.html
    │   ├── macro_sentiment_hashrate_summary.csv
    │   ├── macro_sentiment_hashrate_summary.html
    │   ├── memecoin_vs_conventional_summary_statistics.csv
    │   ├── memecoin_vs_conventional_summary_statistics.html
    │   ├── momentum_strategy_summary.csv
    │   ├── momentum_strategy_summary.html
    │   ├── network_factor_correlations.csv
    │   ├── network_factor_summary.csv
    │   ├── network_factor_summary.html
    │   ├── network_granger_results.csv
    │   ├── network_multivariate_model_results.csv
    │   ├── network_multivariate_model_results.html
    │   ├── network_pca_loadings.csv
    │   ├── network_univariate_model_results.csv
    │   ├── network_univariate_model_results.html
    │   ├── panel_a_market_characteristics.csv
    │   ├── panel_a_market_characteristics.html
    │   ├── panel_b_return_characteristics.csv
    │   ├── panel_b_return_characteristics.html
    │   ├── selected_assets_summary_statistics.csv
    │   ├── selected_assets_summary_statistics.html
    │   ├── size_strategy_quintiles.csv
    │   ├── size_strategy_quintiles.html
    │   ├── size_strategy_spreads.csv
    │   ├── size_strategy_spreads.html
    │   ├── strategy_test_portfolios.csv
    │   ├── summary_statistics_by_frequency.csv
    │   ├── summary_statistics_by_frequency.html
    │   ├── volatility_strategy_summary.csv
    │   ├── volatility_strategy_summary.html
    │   ├── volume_strategy_summary.csv
    │   └── volume_strategy_summary.html
    │
    ├── DELIVERABLES_INDEX.md
    ├── export_manifest.csv
    ├── export_manifest.html
    ├── portfolio_summary.md
    ├── project_inventory.csv
    └── reproducibility_summary.csv

Core Pipeline Modules

00_setup.R

Loads packages, creates project directories, sets global options, and defines lightweight helper utilities used throughout the project.

01_load_and_clean_data.R

Builds the cleaned daily, weekly, and monthly cryptocurrency datasets and prepares the basic market proxy and risk-free rate series.

02_descriptive_statistics.R

Generates descriptive tables and benchmark plots, including return distributions, market comparisons, and characteristic panels.

03_cross_sectional_predictors.R

Runs characteristic-sorted quintile analysis for size, momentum, volume, and volatility strategies.

04_crypto_factor_model.R

Constructs crypto factor returns, builds long-short test portfolios, and estimates factor regressions.

05_memecoin_vs_conventional_analysis.R

Creates grouped return baskets for conventional and meme coins, then compares performance and drawdowns.

06_google_trends_attention_models.R

Studies how Google search attention relates to cryptocurrency return movements using lag-augmented regression models.

07_macro_sentiment_hashrate_models.R

Combines weekly crypto returns with macro-financial, sentiment, and hash-rate variables and estimates baseline explanatory models.

08_network_factors_models.R

Builds and tests blockchain network factors from raw Bitcoin network JSON files, including a PCA-based composite factor.

09_export_tables_and_figures.R

Audits and indexes all generated outputs, producing a handoff-ready manifest and reproducibility summary.

10_run_all.R

Runs the full project pipeline sequentially and writes a complete execution log.

---

Raw Data Requirements

The repository expects source files in data/raw/. Depending on the module, these may include:

• raw cryptocurrency market data
• Google Trends exports
• Bitcoin network JSON files such as:
  • n-payments.json
  • n-transactions.json
  • n-unique-addresses.json
  • output-volume.json
• additional manually exported or downloaded series used by specific modules

Processed outputs are saved automatically to data/processed/.

---

Reproducibility

This project is designed as a reproducible analytical pipeline.

Run the full project

From the project root:

source("code/10_run_all.R")

Or run modules individually

For example:

source("code/01_load_and_clean_data.R")
source("code/02_descriptive_statistics.R")
source("code/03_cross_sectional_predictors.R")
source("code/04_crypto_factor_model.R")
source("code/05_memecoin_vs_conventional_analysis.R")
source("code/06_google_trends_attention_models.R")
source("code/07_macro_sentiment_hashrate_models.R")
source("code/08_network_factors_models.R")
source("code/09_export_tables_and_figures.R")

Execution logs

After a full run, logs are stored in:

• outputs/logs/pipeline_run_log.csv

• outputs/logs/pipeline_run_summary.txt

---

Main Outputs

The repository produces three main categories of deliverables.

1. Processed datasets

Examples:

• cleaned daily crypto panel

• weekly grouped return dataset

• macro/sentiment/hash-rate merged dataset

• network-crypto merged weekly dataset

2. Tables

Examples:

• descriptive summary statistics

• cross-sectional strategy tables

• factor model results

• grouped meme vs conventional summaries

• macro and network model outputs

• HTML-ready export tables for handoff

3. Figures

Examples:

• return distribution plots

• crypto vs market benchmark charts

• meme vs conventional performance comparison

• factor cumulative return plots

• correlation heatmaps

• PCA-based network factor comparison charts

---

Interpretation Notes

A few variables in the project require careful interpretation.

Market size proxy

In some modules, market size is approximated by:

Close × Volume

This is not true market capitalization. It should be interpreted as a size-liquidity proxy.

Attention data

Google Trends data are normalized search-intensity indices, not absolute search counts. They are best interpreted as relative attention measures.

Grouped portfolios

Value-weighted grouped series may be heavily influenced by the largest asset within a category. This matters especially for meme-coin baskets and should be acknowledged when interpreting results.

Factor interpretation

Crypto factors in this project are empirical constructions tailored to the available dataset and should not be treated as exact analogs of traditional equity factors without qualification.

---

Why This Repository Is Structured This Way

This project follows a simple principle:

A research project should be transferred as a transparent analytical system, not as an archive of disconnected files.

The final result is not only code, but a complete research handoff:

• structured processed data

• modular and readable scripts

• exportable tables and figures

• interpretable outputs for writing and defense

• logs and metadata for reproducibility

• portfolio-ready documentation

This makes the repository useful for:

• thesis and dissertation work

• empirical finance portfolio presentation

• research consulting showcase

• methodological handoff to supervisors or collaborators

• GitHub-based academic profile building

---

Suggested Next Improvements

Natural future extensions include:

• HAC / Newey-West standard errors

• rolling-window estimation

• out-of-sample testing

• regime-dependent models

• stronger robustness checks

• equal-weight vs value-weight grouped portfolios

• further modularization into dedicated utils/ files

---

Final Note

This repository is intended to demonstrate a professional standard of empirical research organization:

• structured

• transparent

• modular

• interpretable

• reproducible

• portfolio-ready

Citation

If you use this repository, please cite the Zenodo software record:

Azarov, Evgenii. Common_Risk_Factors_in_Cryptocurrency. Zenodo. https://doi.org/10.5281/zenodo.19015561

Code Organization

The codebase is structured as a staged research pipeline:

• 00_setup.R — package loading and project setup
• 01_load_and_clean_data.R — raw data import, cleaning, interpolation, and return construction
• 02_descriptive_statistics.R — summary statistics and descriptive plots
• 03_cross_sectional_predictors.R — size, momentum, volume, and volatility-based portfolio tests
• 04_crypto_factor_model.R — cryptocurrency factor construction and regression analysis
• 05_memecoin_vs_conventional_analysis.R — comparison of established cryptocurrencies and memecoins
• 06_google_trends_attention_models.R — investor attention analysis using Google Trends
• 07_macro_sentiment_hashrate_models.R — macro-financial and sentiment regressions
• 08_network_factors_models.R — blockchain network factor analysis
• 09_export_tables_and_figures.R — export of final tables and figures
• 10_run_all.R — full project pipeline