Python for Financial Research

Dr. Vincent Grégoire

Department of Finance

March 2018
 Why learn to code?

I It’s fun
I Saves you time
I Produce better research
I Looks good on your CV
I Likely to help your career down the road
 Why learn to code?

Automation is already here

“At its height back in 2000, the U.S. cash equities trading desk at
Goldman Sachs’s New York headquarters employed 600 traders,
buying and selling stock on the orders of the investment bank’s
large clients. Today there are just two equity traders left.

Automated trading programs have taken over the rest of the work,
supported by 200 computer engineers.”

–MIT Technology Review, February 2017

 Why Python?

I General purpose programming language

I Websites (i.e. Youtube, Google, Instagram)
I Games, full software
I Free, open source software
I Very large community
I Tons of online resources and textbooks
I Integrates nicely with R, C/C++ and other languages

I More general than domain-specific software:

I Matlab
I R
I SAS
I Stata
I EViews
I etc. . .
 Why Python?

I What is it very good for?

I Data manipulation
I Visualization
I Web scraping
I Text analysis
I Basic to somewhat advanced statistics and econometrics
I Linear algebra

I What is it less good for

I Very high-performance applications (additional libraries
needed)
I Advanced statistical analysis
 Main scientific Python libraries

NumPy
The main package for numerical analysis in Python. Includes
functions to deal with arrays of data, linear algebra, random
number generation, and much more.
 Main scientific Python libraries

NumPy
The main package for numerical analysis in Python. Includes
functions to deal with arrays of data, linear algebra, random
number generation, and much more.

pandas
Package for panel data analysis, built on top of NumPy. Great for
importing/exporting, merging, cleaning and analysing data.
Written by Wes McKinney while working at AQR (now at 2σ
Investments), so very good for dealing with financial data.
 Main scientific Python libraries

SciPy
A collection of packages that add a lot of nice features such as
more advanced linear algebra (above NumPy), numerical
integration and optimization.
 Main scientific Python libraries

SciPy
A collection of packages that add a lot of nice features such as
more advanced linear algebra (above NumPy), numerical
integration and optimization.

statsmodels
Statistical analysis, including OLS regressions with support for
fixed effects and clustering.
 Main scientific Python libraries

SciPy
A collection of packages that add a lot of nice features such as
more advanced linear algebra (above NumPy), numerical
integration and optimization.

statsmodels
Statistical analysis, including OLS regressions with support for
fixed effects and clustering.

matplotlib
Main package for visualization in Python. Can produce many type
of graphs, highly customizable.
 Sample Matplotlib plots

17.5
Intercept: 0.712***
Adj. R2 : 0.386
15.0 N. Obs.: 882
Regression Coefficient βδ,j

12.5

10.0

7.5

5.0

2.5

0.0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
j

Source: Boguth, Gregoire and Martineau, Shaping Expectations and Coordinating Attention: The Unintended
Consequences of FOMC Press Conferences, 2018.

Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2698477

 Sample Matplotlib plots

9,000,000
Group 1
8,000,000 Group 2
Group 3
7,000,000 Control

6,000,000
Volume ($)

5,000,000

4,000,000

3,000,000

2,000,000

1,000,000
6 6
016 6 016 016 6
201 201 2 201 42 82 201
09 23 t 07 t 21 v0 v1 c 02
Sep Sep Oc Oc No No De
Date

Source: Comerton-Forde, Gregoire and Zhong, Inverted Fee Venues and Market Quality, 2018.

Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2939012

 Sample Matplotlib plots

0.09
0.06

Order imbalance
0.03
0.00
-0.03
-0.05
-0.08
-0.11
-0.14
-0.17

ile
5
uint
4 s q
2 ise
4 3
r pr
Trade 6 2 s su
ing
bins 8
10 1 a rn
E
Source: Gregoire and Martineau, How is Earnings News Transmitted to Stock Prices?, 2018.

Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3060094

 Sample Matplotlib plots

0.25
Panel A: Log(Fund Size) 0.30
Panel B: Log(Fund Family Size)
Traditional Funds Traditional Funds
Performance Fee Funds 0.25 Performance Fee Funds
0.20

0.20
0.15
0.15
0.10
0.10

0.05
0.05

0.00 0.00
−6 −4 −2 0 2 4 6 8 10 −2 0 2 4 6 8 10 12
Log(Fund Size) Log(Fund Family Size)

0.10
Panel C: Fund Age 0.6
Panel D: Domicile
Traditional Funds Traditional Funds
Performance Fee Funds 0.5 Performance Fee Funds
0.08

0.4
0.06
0.3
0.04
0.2

0.02
0.1

0.00 0.0
0 10 20 30 40 50 60 70 80

UK
France

Germany

Ireland

Italy

Spain
Luxembourg

Switzerland
Fund Age

Source: Gregoire and Sotes-Paladino, Double Bonus? Implicit incentives in Mutual Funds with Explicit
Performance Fee, 2017. (Work in progress)
 Additional useful libraries

I seaborn, plotly, Bokeh: other plotting libraries

I SymPy: symbolic algebra (think Mathematica or Maple)
I NetworkX: network analysis
I scikit-learn: machine learning
I NLTK, textblob: textual analysis
I Beautiful Soup, Scrapy, Selenium: webscraping and
information extraction
I Plus many more! Just search for what you want to do,
chances are there is a Python package for it.
 Anaconda

I The easiest way to install Python with scientific packages is

with the Anaconda distribution:
I Mac OS, Windows, Linux
I Python 2 or 3 (we use 3.6 in this tutorial)
I https://www.anaconda.com/download/
I Includes hundreds of packages by default.
I Easily manage “environments” (sets of packages) for each
project.
I Includes Spyder and Jupyter, and R as an option.
Spyder
Jupyter (IPython notebook)
 Keep you code clean

I Write code that is easy to read, and document it well.

I Will save you time in the long run, easier to spot errors.
I Also easier to get feedback and help!
I Python style guide is called PEP8.
I https://www.python.org/dev/peps/pep-0008/
 Keep you code clean

I Write code that is easy to read, and document it well.

I Will save you time in the long run, easier to spot errors.
I Also easier to get feedback and help!
I Python style guide is called PEP8.
I https://www.python.org/dev/peps/pep-0008/

Credit: http://xkcd.com
 Online resources

I Slides and code from this workshop https:

//github.com/vgreg/python-finance-unimelb2018

I Python documentation https://docs.python.org/3/

I Tutorial:
https://docs.python.org/3/tutorial/index.html

I LearnPython.org http://www.learnpython.org/

I Package documentation
I pandas:
http://pandas.pydata.org/pandas-docs/stable/
I matplotlib: http://matplotlib.org/contents.html
I numpy: https://docs.scipy.org/doc/numpy/index.html
 Online resources

I Using Python with WRDS

https://wrds-www.wharton.upenn.edu/pages/support/
wrds-cloud/holding/python-programming-wrds/

I Kaggle https://www.kaggle.com/

I Quantopian lectures
https://www.quantopian.com/lectures

I Stack Overflow http://stackoverflow.com/

I Google / DuckDuckGo / Bing (?)

 Books

If there’s one book you need to get, it’s this one.

Note: Make sure to get the second edition (2017).

Books
 Other (free, open source) tools for research

I Markdown: Easy markup language that can be used in

Jupyter notebooks. Visual Studio Code is a great text editor
for Markdown and Python (now distributed by Anaconda.)
I http://daringfireball.net/projects/markdown/
I https://code.visualstudio.com/

I LATEX: Typesetting system to produce nice report papers (and

presentations such as this one). You can export pandas
datasets directly to LATEX tables, and import in LATEX the PDF
figures created in Matplotlib.
I https://www.tug.org/begin.html
 Other (free, open source) tools for research

I R: Statistical computing, more advanced functions than

currently in Python. You can install R through Anaconda, and
call R code from within Python. (Microsoft’s version is free as
well)
I https://www.r-project.org/
I https://mran.microsoft.com/

I git: Version control, think “track changes” for your code.

I https://git-scm.com/doc
I https://github.com/
 Outline

Four blocks of three hours:

1. Introduction to Python programming

We will discuss what is Python and you will learn the basic
structure of the language. You will also learn our way around the
programming environment, including the two main editors for
scientific Python, Spyder and Jupyter.
 Outline

Four blocks of three hours:

1. Introduction to Python programming

We will discuss what is Python and you will learn the basic
structure of the language. You will also learn our way around the
programming environment, including the two main editors for
scientific Python, Spyder and Jupyter.

2. Introduction to data analysis using pandas, matplotlib

You will learn how to import, export and transform data using
pandas, the panel data package for Python. You will also learn
how to explore the data by generating summary statistics and
plotting graphs using matplotlib.
 Outline

3. More data analysis using pandas and statsmodels

You will learn more advanced features of Python and pandas,
including dealing with timestamps and estimating measures from
daily and intraday data. You will also learn how to estimate OLS
and panel regressions using statsmodels.
 Outline

3. More data analysis using pandas and statsmodels

You will learn more advanced features of Python and pandas,
including dealing with timestamps and estimating measures from
daily and intraday data. You will also learn how to estimate OLS
and panel regressions using statsmodels.

4. Other topics
In this block, you will be introduced briefly to other Python
packages that can be helpful for research. We will look at an
example of web scraping with textual analysis.
Let’s get started!