FRDS - for better and easier finance research¶

What is FRDS?¶

frds is an open-sourced Python package for computing a collection of major academic measures used in the finance literature in a simple and straightforward way.

It is developed and maintained by Dr. Mingze Gao from the University of Sydney, as a personal project during postdoctoral research fellowship.

frds

LICENSE DOWNLOADS

Installation¶

Install via pipInstall from source

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

pip install frds -U

Sometimes new measures are added and available on GitHub but not yet published to PyPI.

In this case, it may be useful to install directly from source.

git clone https://github.com/mgao6767/frds.git

Build and install the package locally.

cd frds
python setup.py build_ext --inplace
pip install -e .

Note

On Windows, Microsoft Visual C++ Build Tools may be needed to compile the C/C++ extensions in the package.

Built-in measures¶

The primary purpose of frds is to offer ready-to-use functions used in researches.

Example 1Example 2

For example, Kritzman, Li, Page, and Rigobon (2010) propose an Absorption Ratio that measures the fraction of the total variance of a set of asset returns explained or absorbed by a fixed number of eigenvectors. It captures the extent to which markets are unified or tightly coupled.

Example: Absorption Ratio

>>> import numpy as np
>>> from frds.measures import absorption_ratio # (1)
>>> data = np.array( # (2)
...             [
...                 [0.015, 0.031, 0.007, 0.034, 0.014, 0.011],
...                 [0.012, 0.063, 0.027, 0.023, 0.073, 0.055],
...                 [0.072, 0.043, 0.097, 0.078, 0.036, 0.083],
...             ]
...         )
>>> absorption_ratio(data, fraction_eigenvectors=0.2)
0.7746543307660252

absorption_ratio function can also be imported using:
```
from frds.measures.bank import absorption_ratio
```
Tip: You can use Tab to navigate annotations.
Hypothetical 6 daily returns of 3 assets.

Another example, Distress Insurance Premium (DIP) proposed by Huang, Zhou, and Zhu (2009) as a systemic risk measure of a hypothetical insurance premium against a systemic financial distress, defined as total losses that exceed a given threshold, say 15%, of total bank liabilities.

Example: Distress Insurance Premium

>>> from frds.measures import distress_insurance_premium
>>> # 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],
...     ]
... )
>>> distress_insurance_premium(default_probabilities, correlations)       
0.28661995758

For a complete list of supported built-in measures, please check frds.io/measures/.

Data source integration¶

Additionally, frds provides an interface to load data from common data sources such as

Wharton Research Data Services (WRDS)
Refinitiv Tick History (formerly Thomson Reuters Tick History)
more to come...

WRDS¶

Example: Download WRDS DataExample: Use Download WRDS Data

As an example, let's say we want to download the Compustat Fundamentals Annual dataset.

Example: Download Compustat Fundamentals Annual

>>> from frds.data.wrds.comp import Funda
>>> from frds.io.wrds import load # (1)
>>> FUNDA = load(Funda, use_cache=True, obs=100) # (2)
>>> FUNDA.data.head()
                                    FYEAR INDFMT CONSOL POPSRC DATAFMT   TIC      CUSIP                   CONM  ... PRCL_F   ADJEX_F RANK    AU  AUOP  AUOPIC CEOSO CFOSO
GVKEY  DATADATE                                                                                                 ...
001000 1961-12-31 00:00:00.000000  1961.0   INDL      C      D     STD  AE.2  000032102  A & E PLASTIK PAK INC  ...    NaN  3.341831  NaN  None  None    None  None  None
       1962-12-31 00:00:00.000000  1962.0   INDL      C      D     STD  AE.2  000032102  A & E PLASTIK PAK INC  ...    NaN  3.341831  NaN  None  None    None  None  None
       1963-12-31 00:00:00.000000  1963.0   INDL      C      D     STD  AE.2  000032102  A & E PLASTIK PAK INC  ...    NaN  3.244497  NaN  None  None    None  None  None
       1964-12-31 00:00:00.000000  1964.0   INDL      C      D     STD  AE.2  000032102  A & E PLASTIK PAK INC  ...    NaN  3.089999  NaN  None  None    None  None  None
       1965-12-31 00:00:00.000000  1965.0   INDL      C      D     STD  AE.2  000032102  A & E PLASTIK PAK INC  ...    NaN  3.089999  NaN  None  None    None  None  None

[5 rows x 946 columns]

Here it skips the setup of WRDS login credentials. To do so, run the following script.
```
from frds.io.wrds import setup
setup(username="username", password="password", save_credentials=True)
```
If save_credentials=True, the username and password will be saved locally in credentials.json in the frds folder. Then in later uses, no more setup is required (not just current session).

The frds folder is created under the user's home directory to store downloaded data upon installation.
use_cache=True attempts to load the data from local cache instead of downloading it again.

We can then compute some measures on the go:

Example: Use Downloaded WRDS Data

>>> tangibility = FUNDA.PPENT / FUNDA.AT # (1)
>>> type(tangibility)
<class 'pandas.core.series.Series'>
>>> tangibility.sample(10).sort_index()
GVKEY   DATADATE
001000  1965-12-31 00:00:00.000000    0.604762
        1967-12-31 00:00:00.000000    0.539495
        1968-12-31 00:00:00.000000    0.654171
        1977-12-31 00:00:00.000000    0.452402
001001  1985-12-31 00:00:00.000000    0.567439
001003  1980-12-31 00:00:00.000000         NaN
        1988-01-31 00:00:00.000000    0.073495
001004  1967-05-31 00:00:00.000000    0.175518
        1980-05-31 00:00:00.000000    0.183682
        1982-05-31 00:00:00.000000    0.286231
dtype: float64

The frds.data.wrds.comp.Funda class has all the variables in the Fundamental Annual dataset as attributes with proper docstrings.

Hence, we can write much simpler expressions whenever possible.

Refinitiv Tick History¶

frds provides a dedicated command-line tool frds-mktstructure.

Use -h or --help to see the usage instruction:

frds-mktstructure can be used without programming

$ frds-mktstructure -h
usage: frds-mktstructure [OPTION]...

Download data from Refinitiv Tick History and compute some market microstructure measures.

optional arguments:
  -h, --help            show this help message and exit
  -v, --version         show program's version number and exit

Sub-commands:
  Choose one from the following. Use `frds-mktstructure subcommand -h` to see help for each sub-command.

  {download,clean,classify,compute}
    download            Download data from Refinitiv Tick History
    clean               Clean downloaded data
    classify            Classify ticks into buy and sell orders
    compute             Compute market microstructure measures

1. Download Data2. Clean Data3. Classify Trade Directions4. Compute Selected Measures

Let's download the tick history for all S&P500 component stocks from Jan 1, 2022, to Jan 31, 2022:

frds-mktstructure download -u {username} -p {password} --sp500 --parse --data_dir "./data" -b 2022-01-01 -e 2022-01-31

where {username} and {password} are the login credentials of Refinitiv DataScope Select.

Note that we set the --parse flag to parse the downloaded data (gzip) into csv files by stock and date into the ./data folder.

Then we clean the downloaded and parsed data in the ./data folder: sorting by time, removing duplicates, etc.

frds-mktstructure clean --all --data_dir "./data" --replace

The --replace flag, if set, asks the program to replace the data file with the cleaned one to save disk space.

Use the classify subcommand to classify trades into buys and sells by the Lee and Ready (1991) algorithm.

frds-mktstructure classify --all --data_dir "./data"

Lastly, use the compute subcommand to compute specified market microstructure measures:

frds-mktstructure compute --all --data_dir "./data" --out bidaskspread.csv --bid_ask_spread

Next step¶

Introduction of frds
Check out the notes to get started
Check out the documentation of built-in measures

Last update: April 17, 2022
Created: June 15, 2020