Introduction to didTEnets

The didTEnets package implements the Transfer Entropy-informed Difference-in-Differences (TE-DiD) framework for policy evaluation in interconnected systems. This methodology combines network analysis with causal inference to measure policy effects on financial contagion and systemic risk transmission.

Overview

Traditional difference-in-differences (DiD) approaches often use simple outcome variables like stock returns or volatility, which fail to capture the complex dynamics of systemic risk transmission in interconnected financial markets. The TE-DiD framework addresses this limitation by:

Network-Informed Outcomes: Using the Network-Informed Contagion Index (NICI) instead of simple returns
Directed Information Flow: Measuring transfer entropy to quantify directional risk transmission
Policy Spillovers: Accounting for spillover effects through the Network-Informed Policy Intensity (NIPI)
Crisis Analysis: Focusing on extreme events through quantile conditioning

Key Concepts

Transfer Entropy (TE)

Transfer entropy measures the directed flow of information between time series, quantifying how much knowing the past of series X reduces uncertainty about the future of series Y, beyond what we already know from Y’s own past.

Network-Informed Contagion Index (NICI)

NICI serves as the outcome variable in TE-DiD models. It measures the weighted sum of incoming transfer entropy flows to a country, representing how much systemic risk the country is absorbing from its network neighbors.

Network-Informed Policy Intensity (NIPI)

NIPI serves as the treatment variable, combining direct policy effects with spillover effects from neighboring countries in the network.

Installation

You can install the development version of didTEnets from GitHub:

# Install devtools if you haven't already
install.packages("devtools")

# Install didTEnets
devtools::install_github("avishekb9/didTEnets")

library(didTEnets)
#> Registered S3 method overwritten by 'quantmod':
#>   method            from
#>   as.zoo.data.frame zoo
library(ggplot2)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

Quick Start Example

Here’s a basic example using the package’s sample data:

# Load sample data
data(sample_financial_returns)
data(sample_policy_data)

# Quick analysis with built-in sample data
tickers <- c("^GSPC" = "USA", "^FTSE" = "GBR", "^GDAXI" = "DEU", 
             "^FCHI" = "FRA", "^N225" = "JPN", "^AXJO" = "AUS")

# Run complete TE-DiD analysis
results <- run_complete_tedid_analysis(
  tickers = tickers,
  start_date = "2019-01-01",
  end_date = "2021-12-31",
  create_plots = TRUE,
  verbose = TRUE
)

# View main results
summary(results$tedid$main_model)

Step-by-Step Workflow

1. Data Acquisition

# Define financial market tickers
tickers <- c("^GSPC" = "USA", "^FTSE" = "GBR", "^GDAXI" = "DEU")

# Download financial data
financial_data <- download_financial_data(tickers, "2019-01-01", "2021-12-31")

2. Policy Data Creation

# Create realistic policy data
policy_data <- create_policy_data(financial_data$countries, "2019-01-01", "2021-12-31")

3. Network Weights Calculation

# Calculate different types of network weights
network_weights <- calculate_network_weights(financial_data$returns)

4. Transfer Entropy Analysis

# Define periods for analysis
crisis_start <- as.Date("2020-03-01")
crisis_end <- as.Date("2020-12-31")

period_breaks <- list(
  "Pre_Crisis" = which(index(financial_data$returns) < crisis_start),
  "Crisis" = which(index(financial_data$returns) >= crisis_start & 
                  index(financial_data$returns) <= crisis_end),
  "Post_Crisis" = which(index(financial_data$returns) > crisis_end)
)

# Calculate TE matrices
te_matrices <- calculate_te_matrices_by_period(financial_data$returns, period_breaks)

5. NICI and NIPI Construction

# Calculate outcome variable (NICI)
nici_data <- calculate_nici(te_matrices, network_weights)

# Prepare monthly policy data
monthly_policy <- policy_data %>%
  mutate(year_month = format(Date, "%Y-%m")) %>%
  group_by(CountryCode, year_month) %>%
  summarise(StringencyIndex = mean(StringencyIndex, na.rm = TRUE), .groups = "drop")

# Calculate treatment variable (NIPI)
nipi_data <- calculate_nipi(monthly_policy, network_weights)

6. TE-DiD Model Estimation

# Prepare dataset for TE-DiD estimation
tedid_dataset <- prepare_tedid_dataset(nici_data, nipi_data, treatment_groups)

# Estimate TE-DiD model
tedid_model <- estimate_tedid_model(tedid_dataset)

# Extract and interpret results
results <- extract_tedid_results(tedid_model, tedid_dataset)

7. Visualization

# Create network comparison plots
network_plots <- create_network_comparison(te_matrices)

# Create TE-DiD result visualizations
tedid_plots <- visualize_tedid_results(tedid_dataset)

# Display plots
print(tedid_plots$timeseries)
print(tedid_plots$difference)

Interpreting Results

The key output of a TE-DiD analysis is the interaction coefficient β₁ from the model:

IncomingFlow_{it} = β₀ + β₁ · EarlyAdopter_i × CrisisPeriod_t + α_i + γ_t + ε_{it}

Where: - β₁ < 0: Early policy adoption reduced incoming financial contagion - β₁ > 0: Early policy adoption increased incoming financial contagion (possibly due to uncertainty) - β₁ ≈ 0: No significant policy effect

Advantages over Traditional DiD

Network-Aware: Captures interconnectedness and spillover effects
Crisis-Focused: Uses quantile conditioning to focus on extreme events
Directional: Measures directed information flows rather than simple correlations
Policy-Relevant: Provides clear implications for crisis management

Next Steps

See the Network Analysis vignette for detailed network construction methods
See the Policy Evaluation vignette for advanced TE-DiD specifications
Check the Case Studies vignette for real-world applications

References

Schreiber, T. (2000). Measuring information transfer. Physical Review Letters, 85(2), 461.
Angrist, J. D., & Pischke, J. S. (2009). Mostly harmless econometrics: An empiricist’s companion. Princeton University Press.
Hale, T., Angrist, N., Goldszmidt, R., et al. (2021). A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker). Nature Human Behaviour, 5, 529-538.