Governments’ stark daily figures on the spread of coronavirus are difficult to compare across countries, and may be significant undercounts. But the data needed to analyse the more reliable and comparable excess mortality metric are only available in a few jurisdictions, leaving these official case and death counts the best available data for much of the world.
- Oct 8 2021: The data source for Slovenia is now the Slovenian Ministry of Health and National Institute for Public Health.
- Sep 22 2021: The default population data source is now the World Bank 2020 population estimates. The 2020 US Census data is used for the US and its states. The data for Malta is the 2020 year-end estimate of the National Statistics Office of Malta.
- Sep 1 2021: Delaware added 130 deaths to the state total on July 30 following a review of death certificates. These deaths have been distributed between May 15 2020 and June 30 2021.
US states in detail
Since the start of the pandemic, lockdowns and social distancing procedures in the United States have been largely managed on a state-by-state basis. President Trump urged governors to use their latitude over reopenings, and several had raced to lift restrictions on business before meeting CDC guidelines on declining case counts as well as the need for widespread testing and contact tracers. But the majority are taking a phased approach to reopening.
Cases or deaths
Comparing the spread of coronavirus in different countries is difficult using the data being released by governments. Confirmed case counts depend heavily on the extent of countries’ very different testing regimes, so higher totals may simply reflect more testing.
Deaths are somewhat more reliable, but remain problematic because countries have different rules for what deaths to include in their official numbers. The most notable difference between countries’ Covid mortality figures is whether or not they include deaths outside hospitals, particularly in care homes. Some countries like France and the UK have even changed which deaths they include during the course of the epidemic. Between May 25 and July 3, Spain’s data was not readily comparable to its earlier figures, and was temporarily removed from the chart until a revised time series was published.
For either measure, we use a seven-day rolling average to adjust for the impact of administrative delays to reporting new data over weekends.
The FT is tracking excess mortality — the difference between deaths from all causes during the pandemic and the historic seasonal average — in the handful of countries and municipalities that publish suitably recent data, and has reported on the specific circumstances in Mexico, Russia, South Africa, Turkey and the UK."
Logarithmic or linear scales
The vertical axis of our charts are shown using a logarithmic scale, where the same distance on the scale represents multiplying or dividing by the same amount, instead of adding or subtracting the same amount as is the case with a linear scale. Log scales are particularly suited to displaying trends in relative rates of change, like a virus spreading. By comparing the slopes of two lines, a log scale allows us to compare epidemics at a very early stage with those that are much more advanced, even though they have very different absolute numbers of cases or deaths.
On a log scale, an epidemic looks like a steep diagonal line that flattens towards a horizontal line as its rate of growth slows. On the more familiar linear scale, the same data looks like a hockey stick shooting upwards, which gives a better sense of the overall size of each country’s epidemic.
Adjusting for population
Unusually for cross-national data, adjusting for population isn’t strictly necessary when analysing the speed at which a virus spreads. Viruses don’t respect borders, and the rate at which they spread is not affected by the overall population of the affected country.
Population matters least in the early stages of an epidemic because cases are likely to be highly concentrated in particular regions like Hubei or Lombardy. Later, though, viewing the values per 100,000 people gives a sense of the pandemic’s relative strain on countries’ resources. Switching to the “per 100k” view won’t alter the shape of each country’s curve, but will reorder them relative to one another.
Adjusted for population, small countries with broad definitions for what cases or deaths to include in their data will look particularly badly affected, while epidemics concentrated in parts of a very populous country look surprisingly small. Try changing this setting while comparing Belgium to the US or China."
We hide countries with populations under 80,000 to avoid distorting the scale of population-adjusted charts. You can still search from them, though: Try looking at San Marino and Andorra; both European microstates have large proportions of their population affected.
Unless otherwise stated below, the data used in thses charts comes from the Johns Hopkins University Center for Systems Science and Engineering, and reflects the date that cases or deaths were recorded, rather than when they occurred.
Data for the Cook Islands, Guernsey, Jersey, Kiribati, Nauru, Niue, North Korea, Palau, Pitcairn, St Helena, Ascension and Tristan da Cunha, Tokelau, Tonga, Turkmenistan, Tuvalu and Wallis and Futuna come from the the World Health Organization.
On October 1 2020, Argentina added 3,050 deaths to its official cumulative death toll, reflecting previous deaths recorded in the province of Buenos Aires that had not been attributed to a date. The time series up until that date has been adjusted to redistribute these deaths in proportion to the previously known distribution of the data.
On September 6 2020, Bolivia added 1,610 deaths to its official cumulative death toll without explaination. The time series up until that date has been adjusted to redistribute these deaths in proportion to the previously known distribution of the data.
Data for Chile before July 18 2020 has been adjusted to redistribute revised death totals published on June 6 and July 16, and the addition of previously unreported cases added on June 17, all in proportion to the original data.
Data for China from before April 17 2020 has been adjusted to redistribute a data revision published on that day in proportion to its original data.
On September 6 2020, Ecuador adjusted its methodology to cease distinguishing betwween confirmed and suspected Covid-19. This resulted in 3,752 additional deaths previously classified as suspected being added to its official cumulative death toll. The time series up until that date has been adjusted to redistributed these deaths in proportion to the previously known distribution of the data.
Data for Eswatini from before April 12 2021 has been adjusted to redistribute 1,017 previously unreported cases in proportion to its original data.
Data for France after April 4, 2020 comes from Public Health France. Data for deaths between January 14 and January 22 are estimates based on cumulative totals before and after this period. The data has been adjusted to redistribute nursing home deaths that were added to the official death toll on April 2 2020 as well as revised confirmed case counts on May 5 2020, May 28 2020 and May 20 2021, in proportion to its original data.
Data for India before June 16 2020 have been adjusted to redistribute older deaths added on that date by Maharashtra and Delhi, in proportion to its original data. On June 10 2021 Bihar added 3,951 deaths after an audit. These have been distributed between Mar 1 2021 and May 25 2021 in proportion to the original data.
Data for Ireland comes from the Health Protection Surveillance Centre and the Health Service Executive via Ireland's COVID-19 Data Hub.
Data for Italy before August 17 2020, has been adjusted to redistribute 154 deaths from March, April and May that the Parma Local Health Authority had not previously reported. These have been distributed in proportion to the previously-known data for the Emilia-Romagna region in those three months.
On January 30, Lebanon reported 290 deaths in addition to the 61 reported on that day. The health ministry said these were cases from 2020 which had previously been under investigation. The time series for 2020 has been adjusted to redistribute these deaths in proportion to the previously known distribution of the data.
On October 5 2020, Mexico’s health ministry said a record increase in cumulative cases and deaths was due to the inclusion of data dating back to June. The time series between June 1 and October 8 has been adjusted to redistribute an estimated 23,845 cases and 2,450 deaths in proportion to the previously known distribution of the data.
On April 6, the Philippines reported 341 deaths from previous months. These have been redistributed across the period prior to March 31.
Data for Slovenia comes from the Slovenian Ministry of Health and National Institute for Public Health.
Data for Sweden after April 5 2020, is calculated from the daily difference of cumulative figures published Tuesday through Fridays by the Swedish Public Health Agency. Unlike most other countries, Sweden uses "date of incidence" figures for its official death toll, so these "date of reporting" figures will not match official data for the most recent days. Correction: Between August 11 and September 15 2021 new cases and new deaths data for Sweden was displayed incorrectly due to a programming error. This was corrected on September 16, 2021.
The last known update of coronavirus data from Tanzania was released on May 7 2020.
Cases data for Turkey prior to November 26 2020, reflected a "new patients" metric that excluded asymptomatic cases. On December 10, Turkey updated its cumulative total of cases to reflect this change. The time series up to November 25 has been adjusted to redistribute these 792,801 additional cases in proportion to the previously known distribution of the data.
UK deaths and new cases data, and all data from that nations of the UK, comes from the UK Government coronavirus dashboard. The 20,537 cases reported in Wales between December 10 2020 and December 17 2020 have been redistributed equally across this period. This corrects for a backlog of more than 11,000 cases that was reported on December 17 2020.
Data for the US, its individual states, Puerto Rico, Guam, American Samoa, the US Virgin Islands and the Northern Mariana Islands is calculated from county-level data compiled by the Johns Hopkins CSSE. Substantial revisions to individual states’ data are distribtuted proportionally to the previously-known data in the approximate period that they are reported to have occured:
Delaware data between May 15 2020 and June 30 2021 has been adjusted to include 130 deaths added to the state total on July 30 2021 after a review of death certificates.
Iowa data prior to February 19 2021 includes 26,822 additional positvive tests added to the cumulative total due to a change in the definition of cases from unique persons testing positive to total postitive tests.
Kentucky data between November 2020 and January 2021 has been adjusted to include 604 deaths occurred in that period but were first recorded on March 18 and 19 following an audit. A further 88 deaths first reported on March 25 have been distributed over the period since October 2020 and 260 deaths first reported on June 1 have been distributed between March 20 2020 and October 26 2020.
Maryland data between May 26 2020 and May 26 2021 has been adjusted to include 517 previously misclassified deaths reported on May 27.
New Jersey data prior to April 24 has been reduced proporitionally to account for 10,442 duplicate postive tests that the state removed from its cumulative total on that date.
New Mexico data prior to May 25 2021 includes 113 deaths that the state added on May 24 after processing a backlog of death certificates.
New York state data between March 19 and March 23 includes 19,563 cases and 210 deaths added to its cumulative totals on March 24 following technical issues. The numbers of cases and deaths added were calculated by comparing data from Johns Hopkins CSSE to Governor Andrew Cuomo's daily updates.
Oklahoma added a large number of previously unreported cases and deaths to its cumulative totals on April 7. The 1,148 cases whose month of occurance were disclosed by the state have been redistributed in the appropriate months between December 2020 and March 2021; 1,701 deaths have been redistributed before March 31.
West Virginia data in December and January 2021 includes 165 deaths first reported on March 12 following an audit. The state's data prior to April 27 has been reduced proportionlly to account for 162 previously-reported deaths that were removed following a review.
Unless otherwise stated, population figures used to adjust data come from the 2020 estimates by the World Bank. Population data for Anguilla and Western Sahara come from the United Nations Population Division. Data for Eritrea comes from the WHO. Local population sources are used for: Ascension, Bonaire, Sint Eustatius and Saba, Cyprus, the Falkland Islands, Guernsey, Jersey, Malta, Moldova, St Helena, Taiwan, Tristan da Cunha the UK, the US and the Vatican City.
Help us improve these charts: We are looking for further sources of national or municipal mortality data showing total deaths from all causes, preferably broken down by day or week and including figures for recent weeks. If you know of a source of this data for your area, please email firstname.lastname@example.org.
US reporting: Brooke Fox
Editing: Martin Stabe