Financial Research Data Services

frds, Financial Research Data Services, is a Python library to simplify the complexities often encountered in financial research. It provides a collection of ready-to-use methods for computing a wide array of measures in the literature.

It is developed by Dr. Mingze Gao from the Macquarie University, initially started as as a personal project during his postdoctoral research fellowship at the University of Sydney.

Important

This project is under active development. Breaking changes may be expected.

If there’s any issue (likely), please contact me at mingze.gao@mq.edu.au.

Quick start

frds is available on PyPI and can be installed via pip.

pip install frds 

The structure of frds is simple:

• frds.algorithms provides a collection of algorithms.

• frds.measures provides a collection of measures.

• frds.datasets provides example datasets.

Read more

• Supported Measures
  • Absorption Ratio
  • Contingent Claim Analysis
  • Distress Insurance Premium
  • Kyle’s Lambda
  • Lerner Index (Banks)
  • Limit Order Book Slope
  • Long-Run Marginal Expected Shortfall (LRMES)
  • Marginal Expected Shortfall
  • Option Prices
  • Probability of Informed Trading (PIN)
  • Spread and Price Impact
  • SRISK
  • Systemic Expected Shortfall
  • Z-score
• Algorithms
  • GARCH(1,1)
  • GARCH(1,1) - CCC
  • GARCH(1,1) - DCC
  • GJR-GARCH(1,1)
  • GJR-GARCH(1,1) - DCC
• Datasets
  • Stock Returns

Examples

Some simple examples.

Measure

frds.measures.DistressInsurancePremium estimates Distress Insurance Premium, a systemic risk measure of a hypothetical insurance premium against a systemic financial distress, which is defined as total losses that exceed a given threshold, e.g., 15%, of total bank liabilities.

>>> import numpy as np >>> from frds.measures import DistressInsurancePremium >>> # hypothetical implied default probabilities of 6 banks >>> default_probabilities = np.array([0.02, 0.10, 0.03, 0.20, 0.50, 0.15]) >>> correlations = np.array( ...  [ ...  [ 1.000, -0.126, -0.637, 0.174, 0.469, 0.283], ...  [-0.126, 1.000, 0.294, 0.674, 0.150, 0.053], ...  [-0.637, 0.294, 1.000, 0.073, -0.658, -0.085], ...  [ 0.174, 0.674, 0.073, 1.000, 0.248, 0.508], ...  [ 0.469, 0.150, -0.658, 0.248, 1.000, -0.370], ...  [ 0.283, 0.053, -0.085, 0.508, -0.370, 1.000], ...  ] ... ) >>> dip = DistressInsurancePremium(default_probabilities, correlations) >>> dip.estimate() 0.2865733550799999 

Algorithm

Use frds.algorithms.GARCHModel to estimate a GARCH(1,1) model. The results are as good as those obtained from other software or libraries.

>>> import pandas as pd >>> from pprint import pprint >>> from frds.algorithms import GARCHModel >>> data_url = "https://www.stata-press.com/data/r18/stocks.dta" >>> df = pd.read_stata(data_url, convert_dates=["date"]) >>> nissan = df["nissan"].to_numpy() * 100 >>> model = GARCHModel(nissan) >>> res = model.fit() >>> pprint(res) Parameters(mu=0.019315543596552513,  omega=0.05701047522984261,  alpha=0.0904653253307871,  beta=0.8983752570013462,  loglikelihood=-4086.487358003049) 

Use frds.algorithms.GARCHModel_CCC to estimate a bivariate GARCH(1,1) - CCC model. The results are as good as those obtained in Stata, if not better (based on loglikelihood).

>>> from frds.algorithms import GARCHModel_CCC >>> toyota = df["toyota"].to_numpy() * 100 >>> model_ccc = GARCHModel_CCC(toyota, nissan) >>> res = model_ccc.fit() >>> pprint(res) Parameters(mu1=0.02745814255283541,  omega1=0.03401400758840226,  alpha1=0.06593379740524756,  beta1=0.9219575443861723,  mu2=0.009390068254041505,  omega2=0.058694325049554734,  alpha2=0.0830561828957614,  beta2=0.9040961791372522,  rho=0.6506770477876749,  loglikelihood=-7281.321453218112) 

Use frds.algorithms.GARCHModel_DCC to estimate a bivariate GARCH(1,1) - DCC model. The results are as good as those obtained in Stata/R, if not better (based on loglikelihood).

>>> from frds.algorithms import GARCHModel_DCC >>> model_dcc = GARCHModel_DCC(toyota, nissan) >>> res = model_dcc.fit() >>> from pprint import pprint >>> pprint(res) Parameters(mu1=0.039598837827953585,  omega1=0.027895534722110118,  alpha1=0.06942955278530698,  beta1=0.9216715294923623,  mu2=0.019315543596552513,  omega2=0.05701047522984261,  alpha2=0.0904653253307871,  beta2=0.8983752570013462,  a=0.04305972552559641,  b=0.894147940765443,  loglikelihood=-7256.572183143142)