Covid-USElection-Correlation

Contributors

This is a contribution by Rohan Sawahn and Lukas Laskowski.

Installation

You need python3 for using this dashboard. Please install the needed packages via: pip3 install -r requirements.txt.

Running the Code and viewing the Results

• Option1: We have hosted the current Version of our vizualization demo on https://lasklu.github.io/ Over there you can view our vizualizations and read about the different plots
• Option2: Building the html files: We have built several graphs with plotly. You can see the code of each graph in the mainfolder: They start with plot_. The plotting methods are also called in the main method. You have now two choices: To build the html files you use the method fig.write_html which are commented out. You can also start each plot via fig.show(). You will also need to use a ]mapbox-api-key](https://www.mapbox.com/) (which is free). Add the file config.iniin the following format:

[mapbox]
secret_token=MY_SECRET_KEY

• Option2: Building the html files: We have built several graphs with plotly. You can see the code of each graph in the mainfolder: They start with plot_. The plotting methods are also called in the main method. You have now two choices: To build the html files you use the method fig.write_html which are commented out. You can also start each plot via fig.show(). You will also need to use a ]mapbox-api-key](https://www.mapbox.com/) (which is free). Add it in the file config.iniin the following format:
• Option2: Building the html files: We have built several graphs with plotly. You can see the code of each graph in the mainfolder: They start with plot_. The plotting methods are also called in the main method. You have now two choices: To build the html files you use the method fig.write_html which are commented out. You can also start each plot via fig.show(). You will also need to use a ]mapbox-api-key](https://www.mapbox.com/) (which is free). Add it in the file config.iniin the following format:

We further built a dashboard (usage described above). If you want to start it locally, you will have to use serve from npm (which starts a local server). But we recommend using the already deployed website on https://lasklu.github.io/. You can find a description of each graph at the left side of the dashboard.

Tasks

Task 1 - Dataset acquisition

In our plotting we use two datasets:

• Dataset1 contains data on polls before the us-presidential election for every us-state We first searched for polls from different institutes and then discovered the webpage https://projects.fivethirtyeight.com/polls/president-general/wyoming/ . Over there we find data for every month beginnging from march 2020 and poll data that is merged from many different institutes. To acquire this data we discovered that there the page sends a request to the server which returns the data in Jsonformat for a specific state. So we then acquired a list of all available states from the website by using JSoup to save all possible states from a dropdown-menu. Then for every state we request the data in form of a Json. This Json is an Object that has different attributes such as the state name and the vote-trend for biden and trump.

  Now we have acquired the poll data beginning in march 2020 for every us-state. But we now have data from different dates and want work with a value for every month. In order to do that we parsed the date and then grouped by month(date) and state. We also added a column that directly saved who is leading in every state and a column that saves the monthly increase in predicted votes, in order to later on be able to plot this better. Finally, we merged the dataframes we acquired for every state into one big dataset which contains the polling results per month for every us-state. The dataset then contains information on the vote-trend for Joe Biden and Donald Trump of total votes in percent for every month since march 2020 and the increase of votes per month as an extra-column.

• Dataset2 contains data on the development of Covid-19 Infections for every-us state. We acquired this Dataset from https://www.kaggle.com/sudalairajkumar/covid19-in-usa. Overthere, we found a dataset in a csv format containing daily data on the development of covid-19 since the beginning of 2020. It contains data on positve cases, totalTests, hospitalized cases, cases on ventilator and deaths for different dates and states. The data is structured in rows that have these attributes and contain the according value. In order to properly be able to work with this data, we preprocessed this and grouped the data by month and state, so that we can work with the monthly increase of covid-numbers. Furthermore we added a third dataset which contains the population for every us-state. We need this to be able to normalize infections per inhabitans per state, so that we have a better value to compare. This population dataset was then merged via pandas into the covid dataset. We also removed states that contained bad data or that weren't present in the polls dataset.

Finally we merged the Dataset1 (Polls) and Dataset2 (Covid) by using the pandas merge function with state and date as parameters.

Task 2 - Vizualization

We will describe each graph and show which results we take out of the visualised data.

1. Deaths per state: This graph shows the development of the amount of deaths related to covid for each state from march to december. It visualises the strong influence of the "second wave", but also shows that especially states around New York were hit very strongly in the beginning of the pandemic.

2. Correlation between Trump's popularity and the covid cases: We want to analyze the influence of the pandemic on the popularity of Donald Trump. How do the citizens of the US evaluate Trumps handling of the crisis? For this we created two different lines. The red line shows the development of Trump's popularity, the green line the amount of positive cases in the US. You can see that especially the "first wave" influenced the popularity of Donald Trump. Many US-americans were not satisfied by Trump's covid-strategy. You can see that his popularity started falling in the beginning of the pandemic. In the months of summer his approval started stabilizing but could not reach its before-covid-level. It shows that the pandemic had an influence on Trumps approval and so, also, on the election in november.

3. Devlopment of Trump's approval rating in each state in relation to the amount of cases: The development of the graph before was very interesting: His approval has fallen very strongly, but only after approximately two months, it started rising continously again. We want to go a little bit more in detail: How does this look like for each state? Our first thought was to develop a similar graph to the second one for each state. But this is not very intuitive and does not help the user to see any relations. Because of that we decided to use an animated map. This map shows for each state of the united states a circle. The larger a circle, the higher is the average increase of new daily covid cases per month compared to the month before for every state. I recommend using the scrollbar at the bottom. Scroll between every month and look at the different developments in the covid-crisis compared. First of all the insights of the graph before seem to be confirmed. His approval has fallen in the first months of the pandemic and then started rising later on again. You can see that many states which were barely affected by the pandemic themselves also disapproved Trump. The high amount of infections at states around New York had influence on the whole country.
   Hint: The month december does not contain vote data anymore. But it shows that nearly any state is strongly hit by the "second wave". Because of that, we decided to leave this month in the map.

4. Trump vs. Biden: This is a snapshot of november, the month when the election took place. It shows for each state who is currently leading in the votes. Furthermore this information is put in relation to the rate of Death on the y-axis and to the rate of postive infections on the x-axis. It is now possible to see whether a state which would vote more likely democrats, has higher or lower death rates compared to states which would vote more likely republicans. It stands out that mainly states where Joe Biden leads have higher deathRates than states which would vote for Donald Trump. But we have also take into account that democratic states have more often a higher population density which makes it easier for the virus to spread.

In summary, the covid-pandemic had a massive impact on Donald Trumps popularity. It could stabilize to a certain level in the months of summer but could not reach its level as in the before-covid-months. Not every change can be attributed to covid, but it had a significant impact on the polls and thus on the election results in november

Sources

• create animations with plotly
• dataset - popularity
• dataset - covid
• pandas docs
• plotly, especially maps