bdproject

This is the repository of the 2020 course project in ETH Zurich's Big Data and Public Policy course. The team members are Niklas Stolz and Felix Zaussinger. The project is being supervised by Malka Guillot and Elliott Ash, of ETH Zurich's Center for Law & Economics' Law, Economics and Data Science Group.

Abstract

Which policies are most effective in combating COVID-19? Leveraging panel regression and sentiment analysis in the United States.

Success in combating COVID-19 depends on multiple factors. Timing and extent of well-informed policies, guidelines and laws are important. Collaboration at practically all governance levels is needed. Lastly, societal sentiment with respect to the pandemic plays a key role, as it is crucial whether top-down imposed measures are responsibly obeyed or merely derided. To this end, the U.S. provides an interesting case as a federal country which government has been delaying broad interventions to combat the pandemic for a prolonged period of time. This led to sub-national governance level actors such as states, counties and municipalities taking the lead within their specific mandates. As a result, response strategies were likely heterogeneous over space and time. The goal of this project is to apply panel regression and sentiment analysis to disentangle the effect of state and national level policy interventions on reducing the number of infections and deaths due to COVID-19.

In a first step, we will manually code policy interventions such as guidelines, laws, and health infrastructure investments at the state-level over time. Second, we will collect state level data on demographic and socioeconomic characteristics, health-related infrastructure, weather, et cetera, which will be used as control variables in the regression framework. Third, we apply sentiment analysis to at least two major newspapers loosely representative of conservative and liberal agendas and reader bases. In this way we intend to get a hand on the societal sentiment with respect to the pandemic: we exploit how people talk about the topic, which might correlate with how they actually behave. Sentiment features will contribute to the model as additional explanatory variables.

State-level data on confirmed COVID-19 infections and deaths from the CSSE data base at John Hopkins University will be used as covariates. In a first run, we will formulate a fully parsimonious panel regression framework, i.e., by including single policies as dummy variables. If this doesn't yield plausible results, we will try to aggregate the policy interventions along a set of relevant dimensions, such as magnitude of restrictiveness and extent (to be defined).

Our analysis aims at answering questions such as: what would have happened if policy X had been applied earlier? What if the response had been better coordinated top-down? If meaningful results can be obtained, this could theoretically contribute to drawing well-informed conclusions out of the policy interventions of the past weeks and inform the policy response for the months to come.

Any suggestions?

To be honest, we are total newbies in this field. If you...

• have any suggestions on how we could improve the approach,
• know about interesting methodologies that might fit the project,
• know data sources that could fit our purpose,
• want to critique the approach in any constructive way

we would be really happy to learn about your ideas - simply shoot us an email. Thanks!

Project Organization

├── LICENSE
├── Makefile           <- Makefile with commands like `make data` or `make train`
├── README.md          <- The top-level README for developers using this project.
├── data
│   ├── external       <- Data from third party sources.
│   ├── interim        <- Intermediate data that has been transformed.
│   ├── processed      <- The final, canonical data sets for modeling.
│   └── raw            <- The original, immutable data dump.
│
├── docs               <- A default Sphinx project; see sphinx-doc.org for details
│
├── logs               <- Log files, for now of the main programme pipeline.
│
├── models             <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks          <- Jupyter notebooks. Naming convention is a number (for ordering),
│                         the creator's initials, and a short `-` delimited description, e.g.
│                         `1.0-jqp-initial-data-exploration`.
│
├── references         <- Data dictionaries, manuals, and all other explanatory materials.
│
├── reports            <- Generated analysis as HTML, PDF, LaTeX, etc.
│   └── figures        <- Generated graphics and figures to be used in reporting
│
├── requirements.txt   <- The requirements file for reproducing the analysis environment, e.g.
│                         generated with `pip freeze > requirements.txt`
│
├── setup.py           <- makes project pip installable (pip install -e .) so src can be imported
├── src                <- Source code for use in this project.
│   ├── __init__.py    <- Makes src a Python module
│   │
│   ├── data           <- Scripts to download or transform data
│   │   └── download.py
│   │   └── reader.py
│   │   └── transform.py
│   │   └── structures.py    
│   │
│   ├── features       <- Scripts to turn processed data into features for modeling
│   │   └── build_features.py
│   │
│   ├── models         <- Scripts to train models and then use trained models to make
│   │   │                 predictions
│   │   ├── predict_model.py
│   │   └── train_model.py
│   │
│   ├── utils       <- Utility functions to be used elsewhere.
│   │   └── paths.py
│   │
│   └── visualization  <- Scripts to create exploratory and results oriented visualizations
│       └── visualize.py
│
└── tox.ini            <- tox file with settings for running tox; see tox.testrun.org

Project based on the cookiecutter data science project template. #cookiecutterdatascience