Traditional empirical research workflows often involve manual data downloads, local data cleaning, independent analysis, and manual report compilation, leading to challenges in reproducibility, collaboration, and version control. This repository provides an infrastructure for an open science-oriented empirical project, specifically targeted at the empirical accounting research community. It features a project exploring the informativeness of quarterly earnings announcements and their contribution to annual share price movements using an event-study methodology. The project showcases a reproducible workflow integrating Python scripts and data analysis, requiring access to the research platform WRDS, which provides access to a variety of datasets.

Assignment III is an empirical replication and paper extension that follows open science principles, integrating programming skills and institutional knowledge gained throughout the CQA (Corporate Decision-Making and Quantitative Analysis) course. This project also builds on the methodological framework established in the course, which explored a range of empirical research methods—including archival analysis, field experiments, and survey-based approaches—to develop a comprehensive understanding of corporate decision-making and quantitative analysis. This repository equips users to explore the role of accounting in corporate practices while developing skills to gather, prepare, and analyze relevant data using tools and platforms essential for collaborative and reproducible research.

The task involves accessing and retrieving data from multiple databases through WRDS, including CRSP and Compustat, as well as Worldscope and Datastream from Thomson/Refinitiv. The use of multiple databases adds an additional layer of complexity, requiring not only a solid understanding of their individual structures but also the ability to integrate data from diverse sources. Reproducing tables and figures from a seminal paper necessitates a deep understanding of the paper’s methodology and meticulous attention to detail to match the results. Additionally, the project output includes documentation of the steps taken and explicit assumptions made. The paper (and presentation) output files present the findings, compare them with the key results from the paper, and discuss any observed differences.

Even if analyzing the informativeness of quarterly earnings announcements isn’t your usual area of interest (though why wouldn’t it be? 🤔) or you do not have access to WRDS Databases, the codebase provided in this repository will give you a clear understanding of how to structure a reproducible empirical project. The template and workflow used here are designed to ensure transparency and reproducibility, making it a valuable resource for any empirical accounting research project.

The default branch, only_python, is a stripped-down version of the template containing only the Python workflow. This branch was cloned from the TRR 266 Template for Reproducible Empirical Accounting Research (TREAT) repository, focusing solely on the Python workflow and utulizing the Python libraries listed in the requirements.txt file.

"Talk is cheap. Show me the code."
— Linus Torvalds

You start by setting up a few tools on your system:

• If you are new to Python, follow the Real Python installation guide that gives a good overview of how to set up Python on your system.

• Additionally, you will need to set up an Integrated Development Environment (IDE) or a code editor. We recommend using VS Code; please follow the Getting started with Python in VS Code Guide.

• You wll also need Quarto, a scientific and technical publishing system used for used for documenting this project. Please follow the Quarto installation guide to install Quarto on your system. I recommend downloading the Quarto Extension for enhanced functionality, which streamlines the workflow and ensures professional documentation quality for this project. You can find out more about the system here.

• Finally, you will also need to have make installed on your system, if you want to use it. It reads instructions from a Makefile and helps automate the execution of these tasks, ensuring that complex workflows are executed correctly and efficiently.

  • For Linux users this is usually already installed.
  • For MacOS users, you can install make by running brew install make in the terminal.
  • For Windows users, there are few options to install make and they are dependent on how you have setup your system. For example, if you have installed the Windows Subsystem for Linux (WSL), you can install make by running sudo apt-get install make in the terminal. If not, you are probably better off googling how to install make on Windows and follow a reliable source.

📂 Next, explore the repository to familiarize yourself with its folders and their contents:

• config: This directory holds configuration files that are being called by the program scripts in the code directory. We try to keep the configurations separate from the code to make it easier to adjust the workflow to your needs. In this project, pull_data_cfg.yaml file outlines the variables and settings needed to extract the necessary data from the WRDS databases. The prepare_data_cfg.yaml file specifies the configurations for preprocessing and cleaning the data before analysis, ensuring consistency and accuracy in the dataset and following the paper filtration requirements. The do_analysis_cfg.yaml file contains parameters and settings for performing the final analysis on the extracted earnings data.

• code: This directory holds program scripts used to pull data from WRDS directly using python, prepare the data, run the analysis and create the output files (a replicated (pickle) output). Using pickle instead of Excel is more preferable as it is a more Pythonic data format, enabling faster read and write operations, preserving data types more accurately, and providing better compatibility with Python data structures and libraries. The picture illustrates the process of serializing Python objects into a binary format (pickling) for storage in a file and deserializing them back into Python objects (unpickling) for reuse in analysis or other workflows.

• data: A directory where data is stored. It is used to organize and manage all data files involved in the project, ensuring clear separation between external, pulled, and generated data. Go through the sub-directories and a README file that explains their purpose.

• doc: This directory contains Quarto files (.qmd) that include text and program instructions for the paper and presentation ( not rendered in this project due to task instructions - however, feel free to use the presentation template and adjust it to your needs). These files are rendered through the Quarto process using Python and the VS Code extension, integrating code, results, and literal text seamlessly.

🌳 Tip: In order to present the project's directory structure in the presentation by simply copy and paste from within the VS Code, use the Extension.

You also see an output directory but it is empty. Why? Because the output paper (and presentation) are created locally on your computer.

Assuming that you have WRDS access, Python, VS Code, Quarto, and make installed, this should be relatively straightforward. Refer to the setup instructions in the section above.

1. Click on the Use this template button on the top right of the repository and choose Create a new repository. Provide the repository with a name, a description, and select whether it should be public or private. Then click Create repository.
2. Clone the repository to your local machine. Open the repository in VS Code and launch a new terminal.
3. It is advisable to create a virtual environment for the project:

python3 -m venv venv # Run this command in the terminal to create a virtual environment in the `venv` directory.
source venv/bin/activate # Activate the virtual environment on Linux or macOS.
# venv\Scripts\activate.bat # If you are using Windows Command Prompt.
# venv/Script/Activate.ps1 # If you are using Windows PowerShell and have allowed script execution.

You can deactivate the virtual environment by running deactivate.

4. With an active virtual environment, you can install the required packages by running pip install -r requirements.txt in the terminal. This will install the required packages for the project in the virtual environment.
5. Copy the file _secrets.env to secrets.env in the project main directory. Then edit the secret.env by adding your WRDS credentials.

6. Run make all in the terminal. I use the Makefile Tools extension in VS Code to run the makefile and create the necessary output files to the output directory. I highly recommend using the Makefile! Otherwise, you can run the following commands in the terminal:

python code/python/pull_wrds_data.py
python code/python/prepare_data.py
python code/python/do_analysis.py
quarto render doc/paper.qmd
mv doc/paper.pdf output
rm -f doc/paper.ttt doc/paper.fff
quarto render doc/presentation.qmd
mv doc/presentation.pdf output
rm -f doc/presentation.ttt doc/presentation.fff

7. Eventually, you will be greeted with the two files in the output directory: "paper.pdf" (and "presentation.pdf"). You have successfully used an open science resource and reproduced the analysis. Congratulations! 🚀

This code base, adapted from TREAT, should give you an overview on how the template is supposed to be used for my specific project and how to structure a reproducible empirical project.

📌 Note: This repository does not contain any pull requests because it is an individual project without ongoing code reviews or feature-based branching workflows. However, it includes a release after project finish, allowing future collaborative research to build upon this work.

To start a new reproducible project on the informativeness of quarterly earnings announcements and their impact on share price movements based on this repository, follow these steps:

1. Clone the repository by clicking “Use this Template” at the top of the file list on GitHub.
2. Remove any files that you don’t need for your specific project.
3. Over time, you can fork this repository and customize it to develop a personalized template that fits your workflow and preferences.

This project utilizes the template used in collaborative research center TRR 266 Accounting for Transparency, that is centered on workflows that are typical in the accounting and finance domain.

The repository is licensed under the MIT license. I would like to give the following credit:

This repository was built based on the ['treat' template for reproducible research](https://github.com/trr266/treat).

💡 If you’re new to collaborative workflows for scientific computing, here are some helpful texts:

• Christensen, Freese and Miguel (2019): Transparent and Reproducible Social Science Research, Chapter 11: https://www.ucpress.edu/book/9780520296954/transparent-and-reproducible-social-science-research
• Gentzkow and Shapiro (2014): Code and data for the social sciences: a practitioner’s guide, https://web.stanford.edu/~gentzkow/research/CodeAndData.pdf
• Wilson, Bryan, Cranston, Kitzes, Nederbragt and Teal (2017): Good enough practices in scientific computing, PLOS Computational Biology 13(6): 1-20, https://doi.org/10.1371/journal.pcbi.1005510