| Major update, July 2022 |
|---|
| On 27 July 2022 the OxCGRT implemented data changes which incorporate different policies applying to vaccinated and non-vaccinated people. This results in breaking changes to the csv files published on Github, and the addition of new csv files. All of our documentation has been updated to reflect these changes, and we have also published a PDF summary of the changes separately. For those looking to make a quick update to a script using our main csv files (eg. OxCGRT_latest.csv), we recommend:
For those who wish to continue using the previous data structure and variables - we have established a legacy repo where we will continue to publish updated data in our old format. If you have been using our OxCGRT_latest.csv file, you can access the legacy version of the file at this URL: https://raw.githubusercontent.com/OxCGRT/covid-policy-tracker-legacy/main/legacy_data_202207/OxCGRT_latest.csv. For those wishing to explore our new expanded dataset, the key changes are:
We have a mailing list we use to provide technical users with detailed release notes and advance warning of changes to database structure. You can register for these email updates here: http://eepurl.com/hiMsdL |
The Oxford Covid-19 Government Response Tracker (OxCGRT) collects information on which governments have taken which measures, and when. This can help decision-makers and citizens understand governmental responses in a consistent way, aiding efforts to fight the pandemic. The OxCGRT systematically collects information on several different common policy responses governments have taken, records these policies on a scale to reflect the extent of government action, and aggregates these scores into a suite of policy indices. We also collect differentiated policies data where different policies apply to people who are vaccinated and non-vaccinated.
This is a project from the Blavatnik School of Government. More information on the OxCGRT is available on the school's website: https://www.bsg.ox.ac.uk/covidtracker. This README contains information about using the database.
Our data is made available free to use for any purpose under a Creative Commons CC BY 4.0 license (see: our license, and a summary of CC BY 4.0 at Creative Commons), this means you must give appropriate credit and link back to our original work. Here are three suggested ways to cite our work:
- Recommended reference for academic publications: Thomas Hale, Noam Angrist, Rafael Goldszmidt, Beatriz Kira, Anna Petherick, Toby Phillips, Samuel Webster, Emily Cameron-Blake, Laura Hallas, Saptarshi Majumdar, and Helen Tatlow. (2021). “A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker).” Nature Human Behaviour. https://doi.org/10.1038/s41562-021-01079-8
- Short credit for media use (CC BY 4.0 License): Oxford COVID-19 Government Response Tracker, Blavatnik School of Government, University of Oxford.
- Full credit for media use (CC BY 4.0 License): Thomas Hale, Anna Petherick, Jessica Anania, Bernardo Andretti de Mello, Noam Angrist, Roy Barnes, Thomas Boby, Emily Cameron-Blake, Alice Cavalieri, Martina Di Folco, Benjamin Edwards, Lucy Ellen, Jodie Elms, Rodrigo Furst, Liz Gomes Ribeiro, Kaitlyn Green, Rafael Goldszmidt, Laura Hallas, Beatriz Kira, Maria Luciano, Saptarshi Majumdar, Thayslene Marques Oliveira, Radhika Nagesh, Toby Phillips, Annalena Pott, Julia Sampaio, Helen Tatlow, Adam Wade, Samuel Webster, Andrew Wood, Hao Zha, Yuxi Zhang. Oxford COVID-19 Government Response Tracker, Blavatnik School of Government, University of Oxford.
OxCGRT collects publicly available information on 21 indicators of government response. This information is collected by a team of over 200 volunteers from the Oxford community and is updated continuously.
Alongside our own data, we also include statistics on the number of reported Covid-19 cases and deaths in each country. These are taken from the JHU CSSE data repository for all countries and the US States. We also include vaccination data in each jurisdiction – using national-level data from Our World in Data (the percentage of people fully vaccinated) and a range of other sources for subnational vaccination data in Australia, Brazil, Great Britain, and the United States.
Full descriptions of the policy indicators and their meaning can be found in our codebook. For more detailed guidance about the codebook and how we interpret the indicators, see the coding interpretation guide. This ensures consistency, and standardizes coding across the database.
Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement. Four of the indicators (E1-E4) record economic policies such as income support to citizens or provision of foreign aid. Eight indicators (H1-H8) record health system policies such as the Covid-19 testing regime or emergency investments into healthcare. Three indicators (V1-V3) record vaccination policies; a country/region/territory’s priority list, eligible groups, and the individual cost of vaccination.
We also record differentiated policies for vaccinated people and non-vaccinated people for 10 indicators: C1-C8, H6, and H8. We define differentiated policies to mean policies where vaccinated people can access greater freedoms due to their vaccination status, and are subject to less stringent restrictions.
Finally, we have a miscellaneous indicator (M1) for notes that do not fit elsewhere.
To help make sense of the data, we have produced four indices that aggregate the data into a single number. Each of these indices report a number between 0 to 100 that reflects the level of the governments response along certain dimensions. This is a measure of how many of the relevant indicators a government has acted upon, and to what degree. The index cannot say whether a government's policy has been implemented effectively.
- overall government response index (all indicators)
- containment and health index (all C and H indicators)
- stringency index (all C indicators, plus H1 which records public information campaigns)
- economic support index (all E indicators)
(Note: these only include indicators recorded on ordinal scales, so they all exclude E3, E4, H4, H5, and M1.)
We also produce different iterations of each index giving different weightings to differentiated policies that apply differently to people based on vaccination status. The documentation folder contains an index methodology explaining how the different indices are calculated and how they are reported for days with incomplete data. This also describes the methodology for the legacy stringency index which is based on the old database structure in place prior to 25 April 2020.
In addition to country-level data, our primary dataset additionally includes some subnational data. So far we have incorporated data for US states, Brazilian states, UK devolved nations, Canadian provinces and territories, Chinese provinces, Australian states and territories, and Indian states into the primary dataset on this covid-policy-tracker repository. These subnational entities are included in their own designated folders in the data folder.
Subnational data can be interpreted using the main codebook, with additional guidance on subnational-specific interpretation available in the documentation folder. The subnational data included in our primary dataset aims to describe the overall policy environment that applies to residents of the state or equivalent jurisdiction, and so includes policies set by the national government where those values are more stringent than state-level action.
The most up-to-date description of database components is here in the documentation folder of this GitHub repo, which contains a detailed codebook, index methodology, a coding interpretation guide, and notes on subnational interpretation.
We have also published a national working paper with our national methodology, data collection protocols, and description of the individual indicators.
Prior to 25 April 2020 the OxCGRT had a structure of 13 indicators (labelled S1-S13). And prior to July 2022, our data structure was simpler as we were not publishing policies that differentiated between people who are vaccinated vs. non-vaccinated. You can find data using our legacy database structures in our dedicated legacy repo.
The OxCGRT is updated continuously in real time. There are numerous ways you can access the raw data.
<-- status of connection to OxCGRT database
On 25 July 2022 the OxCGRT main Github repo was changed. It now includes data changes which incorporate different policies applying to vaccinated and non-vaccinated people, which resulted in new csv files. We will also keep our previous csvs updated in the new legacy repo.
Below is a summary of what is contained in each csv file, what each abbreviation represents, and a list of the contents of each folder. The codebook and index methodology documentation contains more information about what each indicator and index represents.
FROM TOBY: IT WOULD BE GOOD TO ADD A SIMPLE TABLE LIKE THIS:
| File designation | E | NV | V | M* |
|---|---|---|---|---|
latest |
x | x | x | yes |
latest_all_changes |
x | x | x | yes |
latest_combined* |
x | x | x | yes |
differentiated withnotes |
yes | yes | yes | yes |
timeseries |
x | x | x | yes |
*Note: In the latest_combined files, please note that as described in the codebook, many of our indicators are recorded across two variables: one that records the strictness of the policy, and one that records its scope.
- This is reported as a combination of the policy level (a number) and the scope flag (a letter: T for targeted policies or G for general policie; or F/A flags for indicator E1). For instance, for C3_Cancel public events we would have 0, 1T (recommend cancelling in some areas), 1G (recommend cancelling everywhere), 2T (require cancelling in some areas), 2G (require cancelling everywhere).
- We also include a numerical combination, using the same methodology to calculate compenents for our indices: a targeted policy is considered a half-step lower than a general jurisdiction-wide policy. For instance, for C3_Cancel public events we would have 0, 0.5 (recommend cancelling in some areas), 1(recommend cancelling everywhere), 1.5 (require cancelling in some areas), 2 (require cancelling everywhere).
You can also get some OxCGRT national-level data through an API, although this is out-of-date and does not reflect our new (post July 2022) data structure with policies that differentiate between vaccinated and non-vaccinated people. We do not recommend new users use this API. Documentation for this is published here.
It is important to understand the limitations of this dataset, most of which stem from the "live" nature of data collection. Our first goal is to publish a real-time dataset; but this carries risks. For instance, you may get a version of the database that was exported just as our team was half-way through entering new data, or that was exported in the window between an error being made and being fixed.
For details on how these issues around patchy or missing data affect our index calculations, please see our documentation on calculating indices.
- Be skeptical of reductions in index values. Incomplete or missing data will sometimes cause a dip in the calculated index (we conservatively treat some missing values as 0). For most countries, a reduction in index level that is recent, relatively small (less than 10 points), or only lasts a couple of days, is more likely the result of missing data rather than a legitimate reduction in the underlying policy.
- Not all countries are equally up to date. We try to ensure that all countries and relevant subnational units are updated at least once every two weeks, and many countries are updated more frequently. But there will inevitably be "patchiness" within the last month.
- For each country/territory, some indicators will be missing in some days. As our data collectors find information, they will update a country/territory in real time. This means a country/territory may only have up-to-date information for some indicators, but not all.
- Some indicators (and therefore, index values) will be changed retroactively. We aim to have a second pair of eyes review every data point in the OxCGRT. Inevitably, some things may be tweaked in this review process, leading to changes to past dates. We recommend you frequently download fresh data from OxCGRT, rather than relying on an old export.
- Null values are not the same as 0. The gaps described above – where countries are not up to date, or where some indicators are missing – will be represented as null values. These should not be interpreted as a 0, although for the purposes of calculating our indices, we conservatively treat them as such.
- Fiscal and monetary indicators are not evenly covered. We do not have comprehensive and high quality coverage of our indicators E3, E4, H4, and H5. You should check the data carefully before relying on these indicators, and do not assume they are perfectly comprehensive.
Here are several examples of the type of analysis enabled by OxCGRT:
An individual chart of each country/territory is in the /images/country charts folder.
