This repository contains datasets of daily time-series data related to COVID-19 for 30+ countries around the world. For most countries, the data is at the spatial resolution of states/provinces, although for US, UK, NL and CO, it is at the finer resolution of county/municipality. All regions are assigned a unique key, which resolves discrepancies between ISO/ NUTS/ FIPS codes, etc.

There are multiple types of data:

• Outcome data Y(i,t), such as cases, deaths, tests, for regions i and time t
• Static covariate data X(i), such as population size, GDP, latitude/ longitude
• Dynamic covariate data X(i,t), such as mobility, weather
• Dynamic interventional data A(i,t), such as government lockdowns

The data is drawn from multiple sources, as listed below.

The data is stored in separate csv/ json files, which can be easily merged due to the use of consistent geographic (and temporal) keys.

¹ key is a unique string for the specific geographical region built from a combination of codes such as ISO 3166, NUTS, FIPS and other local equivalents.
² Refer to the data sources for specifics about each data source and the associated terms of use.
³ Datasets without a date column contain the most recently reported information for each datapoint to date.

For more information about how to use these files see the section about using the data, and for more details about each dataset see the section about understanding the data.

There are many other public COVID-19 datasets. However, we believe this dataset is unique in the way that it merges multiple global sources, at a fine spatial resolution, using a consistent set of region keys. We hope this will make it easier for researchers to use. We are also very transparent about the data sources, and the code for ingesting and merging the data is easy to understand and modify.

A simple visualization tool was built to explore the Open COVID-19 datasets, the Open COVID-19 Explorer:	If you want to see interactive charts with a unique UX, don't miss what @Mahks built using the Open COVID-19 dataset:
You can also check out the great work of @quixote79, a MapBox-powered interactive map site:	Experience clean, clear graphs with smooth animations thanks to the work of @jmullo:
Become an armchair epidemiologist with the COVID-19 timeline simulation tool built by @LeviticusMB:	Whether you want an interactive map, compare stats or look at charts, @saadmas has you covered with a COVID-19 Daily Tracking site:
Compare per-million data at Omnimodel thanks to @OmarJay1:

The data is available as CSV and JSON files, which are published in Github Pages so they can be served directly to Javascript applications without the need of a proxy to set the correct headers for CORS and content type. Even if you only want the CSV files, using the URL served by Github Pages is preferred in order to avoid caching issues and potential, future breaking changes.

For the purpose of making the data as easy to use as possible, there is a master table which contains the columns of all other tables joined by key and date. However, performance-wise, it may be better to download the data separately and join the tables locally.

Each table has a full version as well as subsets with only the last 30, 14, 7 and 1 days of data. The full version is accessible at the URL described in the table above. The subsets can be found by appending the number of days to the path. For example, the subsets of the master table are available at the following locations:

• Full version: https://open-covid-19.github.io/data/v2/master.csv
• Last 30 days: https://open-covid-19.github.io/data/v2/30/master.csv
• Last 14 days: https://open-covid-19.github.io/data/v2/14/master.csv
• Last 7 days: https://open-covid-19.github.io/data/v2/7/master.csv
• Latest: https://open-covid-19.github.io/data/v2/latest/master.csv

Note that the latest version contains the last non-null record for each key, whereas all others contain the last N days of data (all of which could be null for some keys).

If you are trying to use this data alongside your own datasets, then you can use the Index table to get access to the ISO 3166 / NUTS / FIPS code, although administrative subdivisions are not consistent among all reporting regions. For example, for the intra-country reporting, some EU countries use NUTS2, others NUTS3 and many ISO 3166-2 codes.

You can find several examples in the examples subfolder with code showcasing how to load and analyze the data for several programming environments. If you want the short version, here are a few snippets to get started.

You can use Google Colab if you want to run your analysis without having to install anything in your computer, simply go to this URL: https://colab.research.google.com/github/open-covid-19/data.

If you prefer R, then this is all you need to do to load the epidemiology data:

data <- read.csv("https://open-covid-19.github.io/data/v2/master.csv")

In Python, you need to have the package pandas installed to get started:

import pandas
data = pandas.read_csv("https://open-covid-19.github.io/data/v2/master.csv")

Loading the JSON file using jQuery can be done directly from the output folder, this code snippet loads the master table into the data variable:

$.getJSON("https://open-covid-19.github.io/data/v2/master.json", data => { ... }

You can also use Powershell to get the latest data for a country directly from the command line, for example to query the latest data for Australia:

Invoke-WebRequest 'https://open-covid-19.github.io/data/v2/latest/master.csv' | ConvertFrom-Csv | `
    where Key -eq 'AU' | select country_name,date,total_confirmed,total_deceased,total_recovered

Make sure that you are using the URL linked at the table above and not the raw GitHub file, the latter is subject to change at any moment in non-compatible ways, and due to the configuration of GitHub's raw file server you may run into potential caching issues.

Missing values will be represented as nulls, whereas zeroes are used when a true value of zero is reported.

Flat table with records from all other tables joined by key and date. See below for information about all the different tables and columns.

Non-temporal data related to countries and regions. It includes keys, codes and names for each region, which is helpful for displaying purposes or when merging with other data:

Information related to the population demographics for each region:

Information related to the economic development for each region:

Information related to the COVID-19 infections for each date-region pair:

*Values can be negative, typically indicating a correction or an adjustment in the way they were measured. For example, a case might have been incorrectly flagged as recovered one date so it will be subtracted from the following date.

**Total count will not always amount to the sum of daily counts, because many authorities make changes to criteria for counting cases, but not always make adjustments to the data. There is also potential missing data. All of that makes the total counts drift away from the sum of all daily counts over time, which is why the cumulative values, if reported, are kept in a separate column.

Information related to the geography for each region:

Health related indicators for each region:

Note that the majority of the health indicators are only available at the country level.

Google's and Apple's Mobility Reports] are presented in CSV form as mobility.csv with the following columns:

Summary of a government's response to the events, including a stringency index, collected from University of Oxford:

For more information about each field and how the overall stringency index is computed, see the Oxford COVID-19 government response tracker.

Daily weather information from nearest station reported by NOAA:

Most recent value for each indicator of the WorldBank Database.

Refer to the WorldBank documentation for more details, or refer to the worldbank_indicators.csv file for a short description of each indicator. Each column uses the indicator code as its name, and the rows are filled with the values for the corresponding key.

Note that WorldBank data is only available at the country level and it's not included in the master table. If no values are reported by WorldBank for the country since 2015, the row value will be null.

For countries where both country-level and subregion-level data is available, the entry which has a null value for the subregion level columns in the index table indicates upper-level aggregation. For example, if a data point has values {country_code: US, subregion1_code: CA, subregion2_code: null, ...} then that record will have data aggregated at the subregion1 (i.e. state/province) level. If subregion1_codewere null, then it would be data aggregated at the country level.

Another way to tell the level of aggregation is the aggregation_level of the index table, see the schema documentation for more details about how to interpret it.

Please note that, sometimes, the country-level data and the region-level data come from different sources so adding up all region-level values may not equal exactly to the reported country-level value. See the data loading tutorial for more information.

There is also a notices.csv file which is manually updated with quirks about the data. The goal is to be able to query by key and date, to get a list of applicable notices to the requested subset.

Please note that the following datasets are maintained only to preserve backwards compatibility, but shouldn't be used in any new projects:

• Data
• Latest
• Minimal
• Forecast
• Mobility
• Weather

The output data files are published under the CC BY-SA license. All data is subject to the terms of agreement individual to each data source, refer to the sources of data table for more details. All other code and assets are published under the Apache License 2.0.

All data in this repository is retrieved automatically. When possible, data is retrieved directly from the relevant authorities, like a country's ministry of health.

To update the contents of the output folder, first install the dependencies:

pip install -r requirements.txt

Then run the following script from the source folder to update all datasets:

cd src
python run.py

See the source documentation for more technical details.

If you spot an error in the data, or there's a country you would like to include, the best way to contribute to this project is by helping maintain the data on the relevant Wikipedia article. Not only can that data be parsed automatically by this project, but it will also help inform millions of others that receive their information from Wikipedia.

For code contributions, take a look at the source directory for more information.

If you do something cool with the data (e.g., visualization or analysis), please let us know!

The main creator of this project is Oscar Wahltinez. Other contributors will be listed here in the future.