Detailed Notes

The detailed notes in this page explains the supporting data and evidence behind the facts we present in the book Factfulness, page by page, with links to the underlying sources.

VERSION 1 — Published: April 6, 2018  — This document is freely available under CC BY 4.0 LICENSE

 Download as PDF file »

About this document

At the end of the book Factfulness (page 275 in the English edition) we are listing notes about the core facts in the book, but then we link to this document which is an extended version with much more details and links to hundreds of underlying sources.

We want all facts in the book Factfulness to be based on the most reliable data that exist. But we can’t be experts in every field and you might very well know a better source than the one we are using, or you might have evidence that conflicts with the sources we use. We want to constantly improve the fact-base of this book. Please give us feedback or ask a question in Gapminder’s forum⨠


Anna Rosling Rönnlund & Ola Rosling, Authors of Factfulness and Co-founders of Gapminder Foundation, Stockholm, Sweden, April 6, 2018

Current status of notes

This is work in progress. Please expect lots of updates to these notes during the next couple of weeks and months. Hundreds of thousands of people are reading the book and they may be requesting all kinds of improvements of the documentation.

A note about acronyms of sources

In this list, we use the same acronyms and identifiers for sources, e.g. UN-Pop[1], as you find in the Sources section at the end of the printed book.

Data for 2017.

Throughout the book, where economic indicators do not extend to 2017, Gapminder has extended the series, mainly using forecasts from the World Economic Outlook from IMF[1]. For extending demographic data, we have used the World Population Prospect 2017—UN-Pop[1]. See

Country boundaries.

Throughout the book, we refer to countries in the past as if they always had the boundaries they have today. For example, we talk about Bangladesh’s family sizes and life expectancy in 1942 as if it had been an independent country at that time, although in reality it was still under British rule as part of British India. See

Inside Cover

World Health Chart 2017.

The graph shows 182 states recognized by the United Nations, only excluding those with the smallest populations (such as the Vatican). Each circle, or bubble, represents a country. The area of the circle is proportional to the population of the country, using data from UN-POP[1]. The color of the bubble shows its geographical region. Gapminder has defined these four regions and color coded each of them: green for the Americas; blue for Africa; red for Asia and Australia; and yellow for Europe, in which we have included Greenland, Russia, and some parts of Central Asia. For more on this division, see The chart’s x-axis shows GDP per capita (PPP in constant 2011 international $) on a logarithmic scale, which Gapminder[3] has divided into the four income levels. The GDP data comes from World Bank[1] and has been extended to 2017 by Gapminder[2] using forecasts from the IMF[1], the World Economic Outlook. Population data comes from UN-Pop[1]. The chart’s y-axis shows life expectancy at birth, based on data for 2016 from IHME[1] and extended to 2017 by Gapminder[4]. This chart, together with more information about the sources, are freely available at

Author’s note

Page ix. Gapminder.

Gapminder is an economically independent, educational non-profit foundation with no political or religious affiliations. Gapminder is a fact tank, not a think tank. It was founded in 2005 by the three authors of Factfulness, is based in Stockholm, Sweden, and reports to the city council of Stockholm. For more info see:


Page 1. X-ray.

The X-ray was taken by Staffan Bremmer at Sophiahemmet in Stockholm. The sword swallower is a friend of Hans’s, called Maryanne Magdalen. Her website is here: The original image that Hans saw was of Hjalmar Wickman, who later inspired Hans and taught him the art of sword swallowing.

Page 2. Hans’s sword swallowing.

You can see Hans swallow his bayonet at the end of his second TED-talk:

Page 3. Fact questions.

The 13 fact questions are freely available in multiple languages at

Page 3. Fact Question 1: Low-Income Countries and Girls in School.

Correct answer is C. 60 percent of the girls in low-income countries finish primary school. According to World Bank[3], the number was 63.2 percent in 2015, but we rounded it to 60 percent to avoid overstating progress. The definition of “low income countries” is intentionally left out of the question, since we also aim to investigate how people interpret the term, as described in chapter 1. Primary completion rate, or gross intake ratio to the last grade of primary education, is the number of new entrants (enrollments minus repeaters) in the last grade of primary education, regardless of age, divided by the population at the entrance age for the last grade of primary education, roughly age 11. The World Bank[2] defines 31 countries as low income countries in 2017. The data is based on estimates from UNESCO[3] compiled primarily from household surveys conducted by USAID-DHS[1] and UNICEF-MICS. See

Page 3. Fact Question 2: Where the majority of people live.

Correct answer is B. The majority of people live in middle-income countries. The definitions of the three income groups are intentionally left out, since we also aim to investigate how people interpret the term, as described in chapter 1. The World Bank[2] divides countries into income groups based on gross national income per capita in current US $. According to the World Bank[4], low-income countries represent 9 percent of the world population, the middle-income countries, 76 percent of the world population, and the high-income countries, 16 percent of the world population. The numbers are as follows: Low income: less than $1,005, 31 countries, total population: 0.7 billion. Middle income: $1,006 to $12,235, 109 countries, total population: 5.6 billion. High income: $12,236 or more, 78 countries, total population: 1.2 billion people. See

Page 3. Fact Question 3: Extreme Poverty.

Correct answer is C. The share of people living on less than $1.9/day fell from 34 percent in 1993 to 10.7 percent in 2013, according to World Bank[5]. Despite the impression of precision given by the precise threshold of $1.9/day and the use of decimals, the uncertainties in these numbers are very large. Extreme poverty is very difficult to measure: the poorest people are mostly subsistence farmers with insufficient harvests or destitute slum dwellers, with unpredictable and constantly changing living conditions and few documented monetary transactions. The staff at PovcalNet tell us that the uncertainty of their estimates is probably in the range of half a billion people. But even if the exact levels are uncertain, the trend direction is not uncertain, because the method for estimating has not changed, and thus the sources of error are probably constant over time. We can trust that the level has fallen to at least half, if not one-third. See

Page 3. Fact Question 4: Life Expectancy.

Correct answer is C. The average global life expectancy for those born in 2016 was 72.48 years, according to IHME[1]. The UN-Pop[3] estimate is slightly lower, at 71.9 years. We rounded to 70 to avoid overstating progress. The three answer alternatives were chosen by Gapminder after first having asked the question with an open answer field, letting respondents write any age they wanted. Most people wrote 50 or 60 years. See

Page 4. Fact Question 5: Number of Children in the World.

Correct answer is C. The UN experts (i.e. the demographers of the UN Population Division) publish population forecasts every second year in their publication the World Population Prospect. They work with multiple alternative scenarios. The one they think is most probable is called the ‘Medium fertility variant’, which falls between the highest and lowest predictions of fertility and mortality decline worldwide. During the last 10 years this scenario has been predicting that the number of children in 2100 will not be higher than today. In the latest revision 2017 (UN-Pop[2]) they estimate there are 1.975 billion children (aged 0 to 14 years) in 2017, and they forecast that the number will be 1.957 billion in 2100 (after having peaked at 2.094 in year 2057). Nobody can know for sure, so the question is only asking what the experts are forecasting as most likely. See

Page 4. Fact Question 6: Why Is the Population Increasing?

Correct answer is B. In their forecasts, the UN experts (i.e. the demographers of the UN Population Division) calculate that 1 percent of the population increase will come from 0.37 billion more children (age 0–14), 69 percent from 2.5 billion more adults (age 15–74), and 30 percent from 1.1 billion more very old people (age 75 and older). The reasons are described in chapter 3. Data is from UN-Pop[3]. See

Page 4. Fact Question 7: Natural Disasters.

Correct answer is C. Annual deaths from natural disasters have decreased by 75 percent over the past 100 years, according to the International Disaster Database; see EM-DAT. Since disasters vary from year to year, we compare ten-year averages. In chapter 4 where we further discuss the decline, we also use 25-year averages. In the last ten years (2007–2016), on average 80,386 people were killed by natural disasters per year. This is 25 percent of the number 100 years earlier (1907–1916), when it was 325,742 deaths per year. The huge decline in disaster deaths would be even more striking if two other major global changes are taken into account. First, the world’s population has increased by four, which calls for counting disaster deaths per capita. 1907–1916 there were 181 disaster deaths per million people. 2007–2016 the number was 11. The relative number has dropped to 6 percent of what it was 100 years ago. Second, 100 years ago the communication technologies for reporting disasters were very primitive, compared to the monitoring of today. Many catastrophes must have gone unrecorded or been underreported. The EM-DAT data include death toll estimates for 8969 disasters recorded worldwide since 1900. All known emergency events have been grouped into categories as follows: Animal accident, Complex disasters, Drought, Earthquake, Epidemic, Extreme temperature, Flood, Fog, Impact, Insect infestation, Landslide, Mass movement (dry), Storm, Volcanic activity, Wildfire. See

Page 4. Fact Question 8: Where People Live.

Correct answer is A. The world population in 2017 is 7.55 billion, according to UN-Pop[1]. That would usually be rounded to 8 billion, but we show 7 billion because we are rounding the population region by region. The populations of the four Gapminder[1] regions were estimated based on national data from UN-Pop[1]: the Americas, 1.0 billion; Europe, 0.84 billion; Africa, 1.3 billion; Asia, 4.4 billion. See

Page 4. Fact Question 9: Vaccination.

Correct answer is C. 88 percent of 1-year-old children in the world today are vaccinated against some disease, according to WHO[1] Global Health Observatory, estimate for immunization coverage of the vaccine against TB. We rounded it down to 80 percent to avoid overstating progress. The common vaccines that reached most 1-year-olds worldwide in 2016 are here sorted by coverage level: BCG (Tuberculosis): 88%, DTP3 (Diphtheria tetanus toxoid and pertussis): 86%, MCV1 (Measles, 1st dose): 85%, Pol3 (Polio): 85%, HepB3 (Hepatitis B): 84%, PAB (Neonatal tetanus): 84%, Hib3 (Haemophilus influenzae type b): 70%, MCV2 (Measles, 2nd dose): 64%, PCV3 (Pneumococcal conjugate): 42%, RotaC (Rotavirus): 15%. See

Page 5. Fact Question 10: Girls in School.

Correct answer is A. Worldwide, women aged 25 to 34 have an average of 9.09 years of schooling, and men have 10.21, according to IHME[2] estimates from 188 countries. Women aged 25 to 29 have an average of 8.79 years of schooling, and men 9.32 years, according to Barro and Lee (2013) estimates from 146 countries in 2010. As always there’s uncertainty in these kinds of estimates, but there’s no reason to assume that the difference between the genders is nearly as large as what people think. Gapminder first asked this question with an open answer field to see how responses were distributed, before deciding on three exact values for alternatives A, B and C, in order to make the skewed perception easier to compare to random results. See

Page 5. Fact Question 11: Endangered Species.

Correct answer is C. None of the three species are classified as more critically endangered today than they were in 1996, according to the IUCN Red List of Threatened Species. The tiger (Panthera tigris) was classified as Endangered (EN) in 1996, and it still is; see IUCN Red List[1]. But in 2017 WWF reports “After a century of decline, tiger numbers are on the rise. At least 3,890 tigers remain in the wild, but much more work is needed to protect this species that’s still vulnerable to extinction.” John R. Platt reports the same in the article Big News: Wild Tiger Populations Are Increasing for the First Time in a Century.” in Scientific American, 2016. According to IUCN Red List[2], the giant panda (Ailuropoda melanoleuca) was classified as Endangered (EN) in 1996. In 2015, new assessments of increasing wild populations resulted in a change of classification to the less critical status Vulnerable (VU), while their situation is still problematic see RED-LIST[2]. The black rhino (Diceros bicornis) was classified as Critically Endangered (CR) and still is; see IUCN Red List[3]. But the International Rhino Foundation states that populations in the wild are slowly increasing. In its Annual Report for 2016 the foundation estimated the population at 5,042-5,455. In March 2018 the status of black rhino was reported as “Population slowly increasing” on their website, see See

Page 5. Fact Question 12: Electricity.

Correct answer is C. A majority of the world population, 85.3 percent, had some access to the electricity grid in their countries, according to GTF, the Global Tracking Framework, a collaboration between the World Bank and the International Energy Agency. We rounded this down to 80 percent to avoid overstating progress. The term “access” is defined differently in all their underlying sources. In some extreme cases, households may experience an average of 60 power outages per week and still be listed as “having access to electricity.” The question, accordingly, talks about “some” access. See

Page 5. Fact Question 13: Climate Change.

Correct answer is A. “Climate experts” refers to the 274 authors of the IPCC[1] Fifth Assessment Report (AR5), published in 2014 by the Intergovernmental Panel on Climate Change (IPCC). To access a detailed list for each Working Group that contributed to the IPCC report, see The Climate Change 2014 Synthesis Report, IPCC[2] page 10, states: “Surface temperature is projected to rise over the 21st century under all assessed emission scenarios”—see the full report at See

Pages 7–8. Vaccination Data.

Vaccination data comes from WHO[1]. Even in Afghanistan, more than 60 percent of the one-year-olds today have received multiple vaccinations. None of these vaccines existed when Sweden was on Level 1 or 2, which is part of the reason lives were shorter in Sweden back then. See

Page 7 and Appendix. Poll results.

The results of the online polls by question and country are set out in the appendix. For the results of the polls we have conducted in our lectures, see

Page 8. Fact Questions: Online polls.

Gapminder worked with Ipsos MORI and Novus to test 12,000 people in 14 countries. Their polls were conducted with online panels weighted to be representative of the adult populations—Ipsos MORI[1] and Novus[1].

Page 8. Fact Questions: Results.

The average number of correct answers for the 12 questions (that is, excluding question 13 on climate change) was 2.2, which we rounded to 2. See more at

Page 8. Public awareness of climate change.

The first scientific hypothesis on man-made climate change due to CO2 emissions was published as early as 1896. The “few decades” mentioned here is the time elapsed since the formation of a broad consensus amongst scientist about fundamental facts in the matter, roughly in the 1980s, and the establishment of the Intergovernmental Panel on Climate Change in 1988.

Page 10. Hans testing students.

The early results from testing students are described in Hans’s first TED talk called “The best stats you’ve ever seen.” See Rosling (2006) at

Page 12. World Economic Forum lecture.

For a video recording of the lecture (the audience receives its results five minutes and 18 seconds in), see WEF.

Page 14. Illusions.

The idea of explaining cognitive biases using the Müller-Lyer illusion comes from Thinking, Fast and Slow by Daniel Kahneman (2011).

Page 13. The Ten Instincts and cognitive psychology.

Our thinking on the ten instincts was influenced by the work of a number of brilliant cognitive scientists. Some of the books that completely changed our thinking about the mind and about how we should teach facts about the world are: Dan Ariely, Predictably Irrational (2008), The Upside of Irrationality (2010), and The Honest Truth About Dishonesty (2012); Steven Pinker, How the Mind Works (1997), The Stuff of Thought (2007), The Blank Slate (2002) and The Better Angels of Our Nature (2011); Carol Tavris and Elliot Aronson, Mistakes Were Made (But Not by Me) (2007); Daniel Kahneman, Thinking, Fast and Slow (2011); Walter Mischel, The Marshmallow Test (2014); Philip E. Tetlock and Dan Gardner, Superforecasting (2015); Jonathan Gottschall, The Storytelling Animal (2012); Jonathan Haidt, The Happiness Hypothesis (2006) and The Righteous Mind (2012); and Thomas Gilovich, How We Know What Isn’t So (1991).

Many of these authors study cognitive biases. At the moment of writing the English Wikipedia lists 186 studied biases. Psychologists struggle to capture these quirks in the mental machine in lab experiments so that they can be replicated or falsified. That’s not how we developed our list of misconceptions and instincts. They only describe our hypothesis of how the common wrong thinking may work.

Chapter One: The Gap Instinct

Page 19. Child mortality.

The child mortality data used in the 1995 lecture came from UNICEF[1]. In this book we have updated the examples and use the 2017 mortality data from UN-IGME. The numbers used in the evening lecture back in 1995 were only slightly different: Saudi Arabia 1960: 292, 1993: 38; Malaysia 1960: 105 ;1993:17, Brazil 1960:181, 1993:63, Tanzania: 1960:249, 1993:167,

Page 21. Rate of improvement in Sweden and Saudi Arabia.

In Sweden in 1869, the child mortality rate was 249. It dropped below 35 in 1946, a process that took 77 years. Saudi Arabia moved from 242 in 1960 to 35 in 1993, roughly the same difference in 33 years. See Gapminder[6].

Page 25. Bubble chart: Family size and child survival 1965.

Each circle, or bubble, represents a country. The area of the circle is proportional to the population of the country, using data from UN-POP[1]. On the X-axis is the Total Fertility Rate, using data from UN-POP[3]. The scale is reversed putting large families to the left and small to the right. This is to show progress as a movement from left to right, which is more intuitive. The Y-axis shows Child Survival Rate in percent. These numbers are more commonly expressed as Child Mortality Rate in deaths before 5 years of age per 1000 live births. Instead of deaths per thousand we changed the rate to percent (deaths per 100) because it is more broadly understood, and we also show survival instead of mortality so that the positive direction is upward, which intuitively is more positive. Data comes from UN-IGME. The two boxes are not showing any official thresholds, they are there to visually emphasize the divided world that existed in 1965. 125 countries with 68% of the world’s population were in the “developing” box. Only 44 countries, with 30% of the world’s population, were in the “developed” box. An interactive version of the chart is freely available here:

Page 26. Bubble chart: Family size and child survival 2017.

Each circle, or bubble, represents a country. The area of the circle is proportional to the population of the country, using data from UN-POP[1]. On the X-axis is the Total Fertility Rate, using data from UN-POP[3]. The scale is reversed putting large families to the left and small to the right. This is to show progress as a movement from left to right, which is more intuitive. The Y-axis shows Child Survival Rate in percent. These numbers are more commonly expressed as Child Mortality Rate in deaths before 5 years of age per 1000 live births. We changed the rate to percent, because it is more broadly understood, and show survival instead of mortality so that the positive direction is upward, which intuitively is more positive. The UN-IGME data for Child mortality rate ends in 2016. Gapminder[6] extended the series by using the percentage change expected by UN WPP 2017 medium fertility forecasts from World Population Prospects 2017. The two boxes are not showing any official thresholds, they are there for comparison with the world of 1965. In 2017, only 13 countries with 6.4 percent of the world population are still in the “developing” box. Those are: Angola, Burkina Faso, Burundi, Chad, DR Congo, East Timor, Gambia, Mali, Mozambique, Niger, Nigeria, Somalia and Uganda. 37 countries with 8.4% of world population are between the boxes. 134 countries are in the “developed” box. An interactive version of the chart is freely available here:

Page 28. Danish TV interview.

The Danish TV show is called Deadline and the interviewing journalist was Adam Holm. The interview with English subtitles can be seen here:

Page 29. Primary School Completion Rate: Low-Income Countries.

The primary school completion rate for girls is below 35 percent in just three countries. But for all three, the uncertainty is high and the numbers are outdated: Afghanistan (1993), 15 percent; South Sudan (2011), 18 percent; Chad (2011), 30 percent. Three other countries (Somalia, Syria, and Libya) have no official number. The girls in these six countries suffer under severe gender inequality, but in total they make up only 2 percent of all girls of primary school age in the world, based on UN-Pop[4]. Note that in these countries, many boys are also missing school. See

Page 30. Low-income countries: Imagining the Worst.

Gapminder has asked the public in the United States and Sweden how they imagine life in “low-income countries” or “developing countries.” They systematically guessed numbers that would have been correct 30 or 40 years ago. On average, respondents believed life expectancy is roughly 45 years, while World Bank[7] data says 62 years. They believed roughly 20% of people in low income countries have access to an improved water source, while World Bank[8] data says 66%. They believed 40% of children are vaccinated, but it’s 78% according to World Bank[9] based on WHO[1]. They believed roughly 70% are undernourished but it’s only 26%, see World Bank[10], based on FAO[1]. All data from low income countries have large uncertainties, but still much smaller than the large public misconceptions. See

Page 30. How many live in low-income countries.

Gapminder polled the public, asking the question with an open answer field so respondents were not limited by our three predefined alternatives. In the US 61 percent of respondents entered a value above 50 percent, guessing that a majority of people live in low-income countries. The average guesses were 57 percent in the US and 61 percent in Sweden, see Novus[3]. Gapminder also asked the same question but with “low-income countries” replaced by “developing countries”. The results were the same, as if the terms were synonyms.

Page 32. Graph: The Four Income Levels.

The numbers are rounded to billion people to make it easier to remember. Incomes are in price adjusted PPP 2011 dollars, by ICP. Gapminder[8] estimates the number of people on each income level in 2017 as follows. Level 1 has 0.75 billion people living on less than $2 per day; Level 2 has 3.3 billion people living on incomes between $2 to $8 per day; Level 3 has 2.5 billion people living on $8 to $32 per day; Level 4 has 0.9 billion people living on more than $32 per day. These detailed estimates are based on the World Bank’s PovcalNet for 2013 and forecasts from IMF[1]. PovcalNet is the dataset that the World Bank uses to estimate the official rate of extreme poverty worldwide. Collection of data was made through household income surveys from across the world. National currencies are converted to comparable dollars adjusted for differences between countries in cost of living. The threshold of $2 per day is almost identical to the World Bank’s $1.9 per day. Gapminder rounded that up to $2 per day to make it easier to remember, and to avoid the false precision in poverty estimates that are very rough. See

Page 33. Income levels: doubling scale.

Throughout the book, when talking about personal income levels and countries’ average incomes, we use a doubling scale. Doubling (or logarithmic) scales are used in many situations when comparing numbers across a large range, or when small differences between small numbers are as important as big differences between big numbers. It’s a useful scale when it is not the size of the pay rise that matters, but the size of the rise in relation to what you had before, as explained further on page 98. See

Page 34. Children working in household.

In Somalia, Ethiopia and Rwanda, where a majority live in extreme poverty, most girls aged 5–14 spend at least 2 hours every day doing household chores like fetching water, gathering firewood and cooking. But where water is far away or firewood is scarce some children spend the whole day, everyday fulfilling these tasks. See UNICEF[2].

Page 35. The cost of illness on Level 2.

Life is uncertain on Level 1 and 2. Many (maybe half) of those leaving the extreme poverty of Level 1 fall back again within a year or two. In his great book One Illness Away: Why People Become Poor and How They Escape Poverty (2010), Anirudh Krishna describes the effect of disease on poor families who are making progress. Illness often forces them back into destitute, since they have no safety nets or health insurances whatsoever. Level 2 roughly corresponds to the World Bank income group called “Lower middle income”, where people pay on average 55 percent of their health expenditure with cash, see World Bank[11]. Voices of the Poor is a three volume publication from the World Bank[12] of interviews with poor people across 47 countries. Many describe the unfortunate events of illness that lead them back into extreme poverty again soon after escaping.

Page 36. Fridges and food.

Many people on Level 2 also have some kind of freezer, but with the unstable electricity grid it’s usually not until Level 3 that it’s worth mentioning. The increased variation of dishes can be seen by looking into the refrigerators of people on different incomes here:

Page 36. Traffic accidents on Level 3.

Road injuries is one of the leading causes of disability among working age people on Level 3, see IHME[3] and compar page 96-97 in the section about humps in chapter.

Page 37. Education on Level 4.

The average length of schooling is above 12 in all countries on Level 4, including in many of the gulf states that only recently reached Level 4, like Saudi Arabia and the UAE. The only exceptions on Level 4 are the three small gulf states Bahrain, Oman and Qatar. See IHME[2] and Barro-Lee.

Page 37. Travel and vacation on Level 4.

More than half the citizens in countries on Level 4 traveled abroad in 2014, and large amount of of those trips were for leisure, see The TripBarometer 2015, by TripAdvisor (based on an online survey conducted from 16 January to 2 February 2015 by Ipsos MORI). On Level 4, the average number of tourist departures abroad in 2015 was roughly 600 per 1000 citizens, which is 6 times higher than the average for Level 3, based on World Bank[13] and UN-POP[1].

Page 38. Historic poverty rate.

For the estimate of 85 percent on Level 1 in 1800, see note to the graph on page 52.

Page 38. Income distribution in Western Europe and the US in the 1950s.

Historic records of GDP per capita, adjusted for inflation and price differences, puts the majority in Western Europe and the US in the 1950s on Levels 2 and 3. That is where the majority of the world population is today. For more on the historic income distribution of Europe and the US, see Gapminder[8].  

Page 38. “Developing countries” and Hans’s lectures at the World Bank.

For one of these lectures, filmed June 8, 2015, see World Bank[14]. The outdated terminology of a divided worldview is specifically challenged beginning at 01:30:31. Five months later, the World Bank announced in a public blog post (World Bank[15]) that they had taken the challenge. With references to my latest presentation, and to an interview with Bill Gates in the New York Times, the World Bank finally announced they planned to phase out the use of the “developing world”. They illustrated their decision with a version of the two boxes bubble graph used in this book on pp 25-26.

Page 38. “Developing countries” in other organizations.

Large parts of the UN still use the term “developing countries”, but there’s no common definition. The UN Statistics Division (2017) uses it for something it calls “statistical convenience”. and finds it convenient to classify as many as 144 countries as developing (including Qatar and Singapore, two of the healthiest and richest countries on the planet). The World Economic Forum re-posted the World Bank announcement on phasing out the division in “developed” and”developing”, but continues using the term. From last years you will find posting on their site like “84% of refugees live in developing countries” (June 20, 2017), ”5 ways to make global trade work for developing countries” (September, 2016), and “Digital can lift the developing world out of poverty” (July 10, 2017). This illustrates one aspect of the problematic definitions. In the first article, where the number is taken from a report issued by UNHCR, “developing countries” refer to the outdated list of the UN Statistics Division. In the second and third, “developing countries” and “developing world” actually refers to the UN list of Least Developed Countries, which is updated every third year, taking into account per capita income, human assets and economic vulnerability.

Pages 41, 43–44. Graphs: Income levels in USA, Mexico and Brazil.

The graphs showing people distributed by income, comparing incomes in Mexico and the US in 2016, are based on the World Bank’s PovcalNet for 2013 and forecasts from IMF[1], slightly adjusted to align with the shape of the distributions from the latest available national income surveys. Brazil’s numbers come from World Bank[16] and PovcalNet, slightly adjusted to better align with CETAD. See

Pages 40–41. Math scores.

Part of the example is borrowed from Denise Cummins (2014).

Pages 41–42 Apartheid South Africa.

The average white household in South Africa today can spend roughly five times more money than the average black household. In the 1970s, during the Apartheid system, white people earned on average 12 times more than black people. 2016 in the US white households earned on average 1.6 times more than black households, $65,041 and $39,490 respectively. See US-CPS. (IRR South African Institute of Race Relations South Africa Survey Online 2009/2010.) Historic race specific employment and incomes are reported by BBC here:

Page 44. Poverty and Extreme poverty.

The term “extreme poverty” has a set technical meaning: it means you have a daily income of less than $1.9 per day, as explained in the note to Fact Question 3. The term “poverty” in many countries on Level 4 is a relative term, and the “poverty line” may refer to the threshold for eligibility for social welfare or the official statistical measure of poverty in that country. In Scandinavia, the official poverty lines are 20 times higher than the poverty lines in the poorest countries, like Malawi, even after adjusting for the large differences in purchasing power; see World Bank[17]. The latest US census estimates that 13 percent of the population lives below its poverty line, putting it at approximately $20 per day. In Sweden the official numbers of “poor” is defined relative to the median income in the country, by counting all individuals with incomes less than 60 percent of the median income. The social and economic challenges of being among the poorest in a rich country should not be neglected but it is not the same thing as being in extreme poverty. In extreme poverty, you can’t even afford a daily meal of staple grain porridge. You can’t get poorer without dying. See World Bank[5]. See

Page 45. Differences in poverty.

Voices of the Poor is a three volume publication from the World Bank[12] of interviews with poor people across 47 countries. The interviews shed light on the existing differences between levels of poverty. It is clear that those who live in poverty are themselves well-aware of these differences.

Chapter Two: The Negativity Instinct

Page 48. Living conditions in Sweden in the 1950s.

Hans grew up in Uppsala, Sweden, in the working class slum suburb of Eriksberg, next to the Ekeby brick factory. The sewage problem in the early 1950s was just as bad as it is in industrial slums on level 2 across the world today. Not until the 1970s did this part of Uppsala get improved sanitation. Source: (Uppsala kommun, KULTURMILJÖUTREDNING ERIKSBERG, page Eriksbergs Upsala-Ekeby.)

Page 49. Terrorism rising.

In 1999, terrorism worldwide reached its lowest annual death toll in two decades, with only 2,200 killed worldwide. It then started increasing, and multiplied tenfold over the next 12 years to 32765 in year 2014, after which it has declined slowly during 2 years. See GTD. For more on terrorism, see note to page 118.

Page 49. Overfishing and Red Listed Species.

Fishing ships are getting larger and larger, and go out further into the deep seas to find the remaining stocks of fish. Based on the analysis of Food and Agriculture Organization of the United Nations of assessed commercial fish stocks, the share of fish stocks within biologically sustainable levels decreased from 90 percent in 1974 to 68.6 percent in 2013 (see page 5 in FAO[2]). Paul Collier describes in “The Plundered Planet” (2010) page 160 how to calculate the real price of a natural resource, used by one generation of humans, based on the reproduction rate of the resource. This is a way of determining how much fish each generation can consume. UNEP[1] “Towards a Pollution-Free Planet” reports on more than 500 recorded dead zones in polluted coastal areas around the world. In 2015 the number of threatened species was 23,250 and in 2016 the number had increased to 24,307 according to IUCN Red List[4]. See

Page 49. Sea levels.

I presented the new forecasts of rising sea levels at the launch of the fifth assessment report of IPCC[1] in 2013. The video clip is called “Hans Rosling – 200 years of global change“.

Page 50. Graph: Better, worse, or about the same?

This poll was initially conducted by YouGov, and the results were so extraordinary that Gapminder decided to see if they could be replicated with a different polling company. In 2017 the same question were asked with Ipsos MORI and the results were very similar. The barchart mixes results from those two online polls. The YouGov 2015 poll asked 18 000 people in 17 countries, see YouGov[1]. The two countries where most people were optimistic about the world were China and Indonesia, where 41% and 23% said the world is getting better. But we have decided to remove these two outliers because the proportion of people with access to internet is not large enough to represent the whole population. It’s quite likely that people with internet access have a different perception of the world than the rest of the population. This is not to discard the plausible and interesting hypothesis that Asians may be more positive than westerners. See

Page 50. When to trust the data.

In this chapter we introduce the idea that you should never trust the data 100 percent. For Gapminder’s guidelines on reasonable doubt for dif­ferent kinds of data, see

Page 52. Graph: Extreme Poverty Rate.

The levels of extreme poverty historically can not be known exactly. Adjusting for changes in prices, currencies, food, employment and technology is very difficult. In this book we use numbers from Gapminder[9]. The numbers before 1980 are based on two sources. First: Bourguignon and Morrisson (2002) estimate that in 1820, the share of people below $2 per day in constant 1985 PPP dollars) was 94.4 percent, and the share of people below $1 per day was 83.9 percent. To express this in 2011 PPP dollar prices is not trivial. The two alternative rates from Bourguignon and Morrisson, Max Roser at OurWorldInData[1] use the higher estimate when showing a single line for the global trend of extreme poverty rate; Max Roser uses the higher estimate, while we have decided to go for a lower estimate. This is because the second source, van Zanden[1], indicates a lower rate. The paper World Income Inequality 1820-2000 uses historic GDP per capita from Maddison[1] to estimate what income levels people lived on. For the distribution of incomes within countries they use historic records of the differences in heights of people (such as military data archives). Insufficient food during childhood stops growth and leads to a shorter adult person. By estimating the childhood stunting they can guess the share of people missing food, hence living in extreme poverty. Based on those estimations they assess that 73 percent of people lived below $2 per day, and 39 percent below $1 per day (in constant 1990 PPP dollars). But they couldn’t construct height and GDP data for all countries and roughly 25% of humanity are missing from this estimate. The share missing were probably mostly the poorest, who didn’t even have organized military archives, hence we can add them to the extremely poor, and get 82 percent of humanity in extreme poverty in 1820. We then pull this number back 20 years and assume even more people were poor in 1800. We land on 85 percent on Level 1 at the start of the trend in 1800. After 1980 the data comes from PovcalNet and is described in the note to Fact Question 3. The official World Bank estimate of extreme poverty in year 2013 is 10.7 percent which Gapminder[9] has extended to 2017, by assuming income distributions being constant and IMF[1] GDP per capita forecasts are applicable on household incomes form PovcalNet.

Page 52. 19th century living conditions.

UK probably had among the world’s longest lives in 1800, on average 36 years according to Livi-Bacci (1989). Swedes and almost everyone else lived 32 years or shorter, see Gapminder[4]. British children started working early, on average at age ten, but this varied between regions. Young children were extremely valuable in British coal mines because they were small. Up to 1842, many children as young as five years old died while working their 10 hour shifts as trappers.

Page 53. Dip in Extreme Poverty: China, India and Latin America.

The numbers in the text on the reductions in extreme poverty in China, India, Latin America and elsewhere come from World Bank[5] extended to 2017 by Gapminder[9], assuming IMF forecasts are accurate. See

Page 53. Life Expectancy and Data Doubt.

Life expectancy data is from IHME[1]. In 2016, only the Central African Republic and Lesotho had a life expectancy as low as 50 years. However, uncertainties are huge, especially on Levels 1 and 2. Learn how much data doubt you should have at

Page 54. Historic child mortality.

As populations didn’t grow much before 1800 despite the large number of babies per woman, child mortality (death rate before age 5) must have been almost 50 percent on average throughout human history, and then almost half of those surviving their first five years, died before finishing their parenthood. See Livi-Bacci (1989) page 17, who suggests that 55 percent of children died when life expectancy was 20 years, and 40 percent died when life expectancy was 50 years.

Page 54–55. Deaths from starvation in Ethiopia.

This number is an average of two sources. FRD, page 87 says: “By the end of 1973, famine had claimed the lives of about 300,000 peasants of Tigray and Welo, and thousands more had sought relief in Ethiopian towns and villages.“ EM-DAT has a record of 100,000 deaths from the famine in Ethiopia in 1973.

Page 55. Graph: Average life expectancy from 1800 to today.

The famine in India between 1876 and 1878 began with a drought in 1875. This lead to food shortage and disease during several years, causing up to 5 million deaths according to the sources listed in the Wikipedia article about the Great Famine. As a result, the Indian life expectancy dropped to roughly 19 years, according to the economic historian Mattias Lindgren, see Gapminder[4]. In his book America’s Forgotten Pandemic (1989), Alfred W. Crosby estimates that the Spanish flu caused 50 million deaths. The number is confirmed by Johnson and Mueller (2002) and CDC[1]. The world population in 1918 was 1.84 billion, which means this pandemic wiped out 2.7 percent of the entire global population.

Page 56. World Food Programme.

The World Food Programme was established in 1961 as an experiment to provide food through the UN. The WFP started its life saving mission in September 1962 when an earthquake killed 12,000 people in Iran, leaving thousands without homes and food.

Page 56. Swedes living on Level 4.

82 percent of Swedes live on Level 4 according to PovcalNet, meaning they have an income above $32 per day (in price adjusted dollars, PPP 2011).

Page 58. Not uncommon for children to drown.

With “not uncommon”, we mean that the drowning percentage of all deaths is higher on Level 3 than on other Levels. This is further explained on page 73 and note.

Page 58. Citizens of Lesotho.

The citizens of Lesotho are usually referred to as the Basotho. Many Basotho also live outside Lesotho, but here we refer to those actually living in Lesotho.

Page 58. Life expectancy in Lesotho and uncertainty of data.

In 2016, only  two countries had a life expectancy as low as 50 years, according to IHME[1]. The Central African Republic at 50.2 years, and Lesotho at 50.3 years. The uncertainties are huge for all health estimates on Level 1. According to IHME’s models the confidence interval is for these numbers are ±2.5 years. But that is only expressing the uncertainty within the model. When IHME change their model and improve their estimation calculations, these country estimates may change outside these uncertainty intervals. The estimates  for 24 percent of all countries changed more than the confidence interval between the previous two revisions. For example: in the Global Burden of Disease Study 2013, Botswana’s life expectancy for 2013 was estimated between 62.6 and 68.7 years. But in the Global Burden of Disease 2015 revision, the lower estimate for year 2013 was adjusted downwards by 5.7 years, down to 56.9. This was due mainly to an improved modelling of the HIV pandemic. This is just one example of the actual size of uncertainty of health estimates on Level 1 and 2. There are no examples of such large  adjustments for countries’ estimates on Level 3 and 4, where data is more certain.

Page 58. Literacy in Sweden and India.

The most recent census data on literacy in India puts the overall literacy rate at 74 percent of the population aged 7 and above. While literacy differed between states—for example, 64 percent in Bihar and 94 percent in Kerala—the numbers show a 10 percent increase since the last census that was published 10 years earlier. Assuming that the literacy rate continues to increase, it is probable that the majority of India has a literacy rate of at least 80 percent in 2017. For Sweden, the literacy rate began to slowly increase in 1765 when the church decided to penalize those who didn’t participate in the catechetical exams, conducted in households to measure people’s literacy (see these paintings from the so called “House interrogations” (Swedish: “husförhör”). While literacy did increase, still a century later, many people in Sweden couldn’t read and write. But with the 1842 school reform it became obligatory for children to attend school, the numbers rose further (see By 1900, the literacy rate had climbed to roughly 87 percent.

Both in India today and in Sweden 100 years ago, “literacy” may only mean basic recognition of letters and the ability to parse text slowly. The figures do not imply an ability to understand advanced written messages. The literacy rate for India is from India Census 2011. Historic literacy numbers for Sweden are from van Zanden[2] and OurWorldInData[2], visually interpolated by Gapminder.

Page 59. Vaccination on Level 1.

In Afghanistan in 2016, more than 60 percent of the 1-year olds have received most of the vaccines as listed by WHO[1]: BCG (Tuberculosis): 74 percent, DTP3 (Diphtheria tetanus toxoid and pertussis): 65 percent, HepB3 (Hepatitis B): 65 percent, Hib3 (Haemophilus influenzae type b): 65 percent, MCV1 (Measles, 1st dose): 62 percent, PAB (Neonatal tetanus): 65 percent, PCV3 (Pneumococcal conjugate): 65 percent and Pol3 (Polio): 60 percent. Only MCV2 (the second shot against measles which brings immunity from 95 to 99.99 percent) is lower, at 39 percent, which is the case in many countries on Level 1. None of these vaccines existed when Sweden was on Level 1 or 2, which is part of the reason lives were shorter in Sweden back then.

Pages 60–63. Graphs: 32 Improvements.

The data behind each of the 32 line charts on pages 60–63, together with detailed documentation of how the many sources were used, can be found here:

Page 60. Graph: Legal slavery.

For detailed documentation and data behind this graph, visit:

The date used for abolishment is the earliest recorded year when the country either signed one of the UN treaties condemning slavery and/or forced labour of all citizens and guest workers, usually its one of ILO[1] from 1930 or ILO[2] from 1957, but some countries signing the UN Slavery Convention in Geneva already in 1926. At the time of signing these international agreements, many countries already had a law or constitution i place banning slavery. REading all laws and constitutions is a very time consuming task. Fortunately this task was already undertaken by Dr. Jean Allain and Dr. Marie Lynch at the School of Law, Queen’s University in Belfast in Northern Ireland. And they made the result freely available as an online database for Slavery in Domestic Legislation. We went over their webpages with with excerpts from domestic law and constitutions for all 193 UN-members and gathered the dates where legal slavery and forced labor was made illegal (see SDL among the sources of the book). In addition to official UN documents and constitutions we rely partly on Pinker (2011) and Wikipedia[1]. Official abolishment in most cases doesn’t mean that slavery disappears in practice.

In many countries law enforcement is still not taking sufficient action to make sure the law is followed to end the remaining slavery. But that’s a different question. Here we only count the fact that the law is in place, which is definitely a step in the right direction compared to slavery being legal. In principle, all countries have abolished slavery, but we have decided to exclude three countries even if their constitution and law make forced labor illegal. It would just be too hypocritical to classify them as officially condemn slavery, when there are strong evidence that the government itself is actively using forced labour. Those are: Turkmenistan and Uzbekistan where ILO’s investigators are not getting full insight (see ILO[3,4]) and journalists report forced labour in state owned cotton industry, see North Korea’s constitution makes slavery illegal, but in practice ILO[5] has little or no insight into this in reality. Reports from witnesses who have escaped the labor camps give no reason to believe that the current government is following its own constitution.

Page 60. Graph: Oil spills.

Not only has the amount of oil spilled been falling. The number of accidents dropped from an average of 24.5 per year in the 1970s to 1.7 per year between 2010 to 2016.

Page 60. Graph: Children dying.

The trend shows ten year averages from Gapminder[6], which ends with data from UN-IGME[1], published in 2017 for the period 1990 to 2016. Before 1990 estimates from hundreds of historical sources, primarily and Mitchell’s books on are combined into one trend line. In summary:

  • 1800 to 1950: Based in estimates documented in version 7 at the bottom of this page: These numbers were was compiled and documented by Mattias Lindgren from many sources, but mainly based on and the series of books called International Historical Statistics  by Brian R Mitchell, which often have historic estimates of Infant mortality rate which were converted to Child mortality through regression. See detailed documentation of v7 below.
  • 1950 to 2016: UNIGME, is a data collaboration project between UNICEF, WHO, UN Population Division and the World Bank. They  released new estimates of child mortality for countries and a global estimate on October 17, 2017, which is available at In this dataset almost all countries have estimates between 1970 and 2016, while roughly half the countries also reach back to 1950.
  • 1950 to 2017: UN Pop WPP, World Population Prospects 2017, provides annual data for Child mortality rate for all countries in the interpolated demographic indicators, called WPP2017_INT_F01_ANNUAL_DEMOGRAPHIC_INDICATORS.xlsx, accessed on September 2, 2017.

From these sources came our county estimates, and our global trend for Child mortality rate is using the UN IGME data for the period 1970 to 2016. All other years is a weighted mean of countries data. The proper way to calculate the global child mortality, would require estimating the total number of child births and child deaths each year. But we don’t have good estimates of the number of births, so instead we have used a proxy: Fertility rate multiplied by population. This method gets us very close to the properly calculated UNIGME numbers. For 1990, UNIGME has 93.4, and our weighted average is 96.6. We have linked our weighted average for the world, to the UNIGME series, by using the rate of change before 1990, and apply that backwards in time, so the whole series is moved down to meet UNIGME in 1990.

See for a detailed documentation.

Page 60. Graph: Death penalty.

The reasons to abolish the death penalty is not only the terrible risk of wrongly receiving a death sentence. It also breaches two essential human rights: the right to life and the right to live free from torture. Both rights are protected under the Universal Declaration of Human Rights, adopted by the United Nations in 1948, and should be respected by all 193 members states. Amnesty International keeps track of countries that have abolished the death penalty after 1990. For a country to qualify it requires that the use of the death penalty is prohibited as punishment for any crimes. Abolishment prior to 1990 in this trend comes from Wikipedia[2] and Pinker (2011).

Page 60. Graph: Leaded gasoline

Tetraethyl lead began to be added in gasoline for increased vehicle performance and fuel economy in the 1920s, see That was a stupid idea, since lead is a toxicant that affects multiple body systems and is particularly harmful to young children: See WHO’s Fact sheet, Updated August 2017, called “Lead poisoning and health”.

The first country to ban lead in gasoline was Japan in 1986. Since then 192 countries have followed. The only three countries where leaded gasoline is still not completely phased out are Iraq, Yemen and Algeria, according to UNEP[3].

Page 61. Graph: Child labor.

The current ILO definition of “child labor” excludes light part-time work, and includes only the so called “worst forms of child labour”; see ILO[6]. Gapminder[42] combines data from three different ILO reports, using three age intervals; ILO[7,8,9]. The actual degree of child labor is uncertain, but the trend of decline is evident in all sources measured consistently over time. Data from ILO[8] covers the years 2000–2012. It overlaps in time with ILO[7] but reports data for a wider age-interval spanning the years 5–17 years. The overlapping years in ILO[8] were used to align with ILO[7]. The ILO[9] data is from the ILO Programme on Estimates and Projections on the Elimination of Child Labour and covers the period 1950–1995. The age interval reported here is 10–14 years. For 1950, 27.6 percent is the estimate of children were involved in child labor. This is probably a low estimate, considering that a majority of children didn’t go to school in the 1950s; only one third did in China and India, according to Barro-Lee. We don’t know if these children out of school were used for labor under bad conditions, so we decided to use this official ILO estimate, even though it reports on an older age group. While the global decline of child labor is certain, the actual levels at all times are not. Since writing this book, new numbers supporting our general estimates have been published on OurWorldInData[3].

Page 61. Graph: Deaths from Disaster.

Annual deaths from natural disasters have decreased by 75 percent over the past 100 years, according to the International Disaster Database; see EM-DAT. Since disasters vary from year to year, here we compare ten-year averages. In chapter 4 where we further discuss the decline, we also use 25-year averages. In the last ten years (2007–2016), on average 80,386 people were killed by natural disasters per year. This is 25 percent of the number 100 years earlier (1907–1916), when it was 325,742 deaths per year. The huge decline in disaster deaths would be even more striking if two other major global changes are taken into account. First that the world’s population has increased by four, which calls for counting disaster deaths per capita. In 1907-1916 there were 181 disaster deaths per million people. In 2007-2016 the number was 11. The relative number has dropped to 6 percent of what it was 100 years ago. Second, 100 years ago the communication technologies for reporting disasters and their victims were very primitive, compared to the monitoring of today. Many catastrophes must have gone unrecorded ur been underreported. The EM-DAT data include death toll estimates for 8969 disasters recorded worldwide since 1900, including all known emergency events categorized as: Animal accident, Complex Disasters, Drought, Earthquake, Epidemic, Extreme temperature, Flood, Fog, Impact, Insect infestation, Landslide, Mass movement (dry), Storm, Volcanic activity, Wildfire.

Page 61. Graph: Nuclear arms.

The number of warheads peaked in the mid 1980s and has steadily decreased since. SIPRI estimates the total number in 2017 at 14,935. The Nuclear Notebook gives a detailed breakdown: “As of mid-2017, we estimate that there are nearly 15,000 nuclear weapons located at some 107 sites in 14 countries. Roughly, 9,400 of these weapons are in military arsenals; the remaining weapons are retired and awaiting dismantlement. Approximately 4,150 are operationally available, and some 1,800 are on high alert and ready for use on short notice.

By far, the largest concentrations of nuclear weapons reside in Russia and the United States, which possess 93 percent of the total global inventory.”

Page 61. Graph: Smallpox.

Smallpox had been one of the major killers of humans. In 18th century Europe, for example, it caused about 7% of all deaths according to Max Roser at OurWorldINData. Vaccines were invented already in 1796, but not until 1980 was the nasty viruses defeated thanks to massive global vaccination campaigns lead by the World Health Organization. The last known case was recorded in Somalia in 1977. Smallpox was the first (and so far the only) disease eradicated by global vaccination programs. The graph shows the years when the last case was recorded in each country, based on data that was kindly shared with us by Katie Hampson, at Wellcome Trust Boyd Orr Centre for Population and Ecosystem at University of Glasgow, Medical, Veterinary & Life Sciences. This data was published in the paper: “Towards the endgame and beyond: complexities and challenges for the elimination of infectious diseases“ in,

Published 24 June 2013, by Petra Klepac, C. Jessica E. Metcalf, Angela R. McLean, Katie Hampson, DOI: 10.1098/rstb.2012.0137

Page 61. Graph: Ozone depletion.

The data show the impressive decline in the use of gases that are causing a hole in the ozone layer of the Earth’s atmosphere. The ozone layer prevents harmful UVB wavelengths of ultraviolet light from reaching the Earth’s surface, where it causes skin cancer, sunburn, and cataracts, all of which were projected to increase dramatically as a result of thinning ozone. The UVB light is also harmful to plants and animals. When scientists presented evidence of the gases causing the ozone hole, the world reacted fast. All countries agreed to stop using these gases, and ratified the Montreal Protocol in 1987. Since then, humanity has almost completely stopped consuming Ozone Depleting Substances (ODS). In 2017 the hole finally showed signs of shrinking as reported by MIT (which may as well be caused by warmer temperature according to NASA. The trend line shows ODS Consumption in ODP Tonnes. The unit ODP tonnes is not an actual weight measurement, but represent the destructive effects of a substance compared to a reference substance. Data is compiled and available form UNEP[4], and the substances included are Chlorofluorocarbons (CFCs), Halons, Other Fully Halogenated CFCs, Carbon Tetrachloride, Methyl Chloroform, Hydrochlorofluorocarbons (HCFCs), Hydrobromofluorocarbons (HBFCs), Bromochloromethane and Methyl Bromide.

Page 62. Graph: New movies.

The Internet Movie Database (IMBd) is maintained by film enthusiasts across the world, and has a nearly complete coverage of all feature films, making 3.5 million movies searchable and filterable by year. Many early cinematic recordings were done in the last years of the 19th century, but they dare often experimental and short we don’t consider them feature films. We have chosen 1906 as the year of the first feature, referring to “The Story of the Kelly Gang”, a 70 minute long narrative film (see Wikipedia[3]).

Page 62. Graph: Protected nature.

This trend is based on UNEP[5], which keeps track of protected areas as defined by IUCN[1,2]. The trend after 1990 is based on the figure on page 30 in UNEP[6]. Between 1911 and 1990 the trend comes from Abouchakra et al (2016), which is based on UNEP[5] data from february 2012. Between 1900 and 1911 the data was aggregated form the historic records in UNEP[5] by Gapminder[31].

Page 62. Graph: Women’s right to vote.

In 1893 New Zealand took the step, and then all other countries followed, except one. Saudi Arabia started letting women vote in 2015 (and since 2017 women are also allowed to drive cars). As of 2017, The Vatican State is the only country left where voting requires male gender. One could argue that there is no popular vote at all in the Vatican, since the head of state is chosen by the cardinals, but nevertheless, it’s the only state where no woman’s voice is heard in election. The data from Gapminder[20] is mainly based on the Wikipedia[4] page about women suffrage.

Page 62. Graph: New music.

The first time music was “recorded” was for experimental purposes in 1857. Since then, music recording have increased both in terms of quality and quantity. The curve shows the number of songs on Spotify, based on the recording date as stated by the ISRC (International Standard Recording Code) which is a global identifier for sound recordings that let music artists track and charge based on how often their recordings are played in different channels. Of course, not all music recordings are available through Spotify, but the shape of the curve shows the mind blowing increase in cultural expression and consumption.

Page 62. Graph: Science

The final data-point in this curve, 2.6 million articles, is an estimate from Elsevier publishing, who published 400,000 peer reviewed articles in 2015, with the help of 700,000 peer-reviewers, from a scientific community of 7.8 million active researchers worldwide. The first peer-reviewed scientific journal was Philosophical Transactions published by the Royal Society in London. The first issue was published in March 1665 and during its first year in existence 119 scientific articles were printed (only counting articles that are more like modern day scientific articles, excluding book reviews and extracts from letters). Another scientific journal, Journal des Sçavans, was established in France in the same year, but without peer-reviewing.

Page 62. Graph: Harvest

Today, each acre of cropland produces on average 3.6 times more food compared to 50 years ago; see FAO[4].

Page 62. Graph: Democracy.

Putting countries into two groups as being a “democracy” or not is highly problematic. We use Max Roser’s data as compiled at OurWorldInData[4]. Roser has adopted the definitions of the Polity IV dataset but give the numbers in terms of inhabitants, not number of countries. The Polity IV dataset puts countries on a democracy scale, and defines non-democratic regimes as autocracies (e.g. China), closed anocracy (e.g. Morocco), open anocracy (e.g. Russia) or colonial regimes. In this graph, we only show democracies, hence disregarding all types of non-democratic regimes. For an alternate detailed list of development of democracy, see Mathew White’s “Chronological List of Democracies”:

Page 63. Graph: Child cancer survival.

This is not a global average for survival of all children with cancer. The data is for children treated in the US, which is representative for the cancer treatment given to children on Level 4, in Europe, Japan and elsewhere. That’s what we mean by “best treatment”. The trend shows that cancer research has been successful and the state of the art treatment is improving, but we should keep in mind that most people don’t have access to this treatment. The first part of the trend line comes from NCI[1]. The 2010 estimate is from NCI[2].

Page 63. Graph: Girls in school.

This trend is based on UNESCO[3], and shows the number of girls of official primary school age who are enrolled in primary or secondary school, expressed as a percentage of the population of official primary school age girls. The age interval for primary education varies between countries, but it’s often 6 to 11 years of age. Girls of primary school age who are still in pre-primary education are excluded and considered out of school. UNESCO[4] expresses the numbers in “out-of-school” children. We have inverted the numbers to show children not out-of-school: the in-school rate.

Page 63. Graph: Monitored species.

The IUCN Red List[4] now includes an impressive 87,967 wildlife species (animals, fungi and plants), each with an expert assessments of threat-status. Out of these, 25,062 (approximately 28%) today fall within one of the the three “threatened” categories: Critically Endangered (CR), Endangered (EN), or Vulnerable (VU). Despite the fact that the status of many of these species is not improving, we consider it a great improvement that they are at least monitored. Data after 2000 come from Red List[4]. The estimates for the following years are handpicked from the following publications and pages: 1986, 1990 and 1996 come from previous paper editions of the list: Page iv, Red List 1986 edition; Page ix, Red List 1990 edition; and page 4, Red List 1996 edition. The first systematic approach to register and monitor threatened species was the 1959 “Threatened Mammals Card Index”. It compiled data about 34 mammal species and was managed by The Species Survival Commission under Leofric Boyle, according to the about page at Red List.

Page 63. Graph: Electricity coverage.

Data comes from GTF, the Global Tracking Framework, a collaboration between the World Bank and the International Energy Agency. The term “access” is defined differently in all their underlying sources. In some extreme cases, households may experience an average of 60 power outages per week and still be listed as “having access to electricity.” The graph, accordingly, shows people with “some” access.

Page 63. Graph: Mobile phones.

Statistic showing the global increase of mobile penetration often uses data from ITU[1], counting the number of subscriptions, not subscribers. In the world in 2015 there were 7.2 billion SIM cards and 7.5 billion people, but the rate is misleading since many people have multiple SIM cards. GSMA publishes estimates of the number of unique subscribers, and their data series start with the 2010 numbers. Gapminder has extended the series, combining these two measures by calculating the rate of subscriptions per subscriber for the overlapping year 2010, and then assume that the same rate is applicable from the beginning of the ITU[1] subscription series in 1980 (when ITU reports 23,482 subscriptions worldwide). The ITU series beginning in 1980 is retrieved from the World Bank[18].

Page 63. Graph: Internet.

Data for 2005 to 2017 is taken from ITU[2] divided by UN-POP[1]. Data before 2005 comes from previous editions of the ITU report via World Bank[19]. These are combined in Gapminder[22]. The first data point, zero internet users in 1980, is based on the Internet System Consortium (ISC), which count internet hosts historically. The first record is from August 1981 with 213 hosting servers, why we assume that the year before, the number of users was practically zero.

Page 63. Graph: Immunization.

Data from WHO[1] gives the immunization coverage of all different common vaccines, as described in the note to Fact Question 9. Gapminder[23] has combined these to one single indicator: the share of children who has received at least one vaccine. This indicates at least some basic form of access to modern health service and scientific medicine.

Page 64. Guitars per capita.

This documentation will be added in the next version of this document. For more information about this chart, see (coming soon)

Page 66. Historic child murders.

In violent communities, children are not spared. Members of hunter-gatherer groups generally experienced lots of violence, as described in Gurven and Kaplan (2007), Diamond (2012), Pinker (2011), and OurWorldInData[5]. This doesn’t mean all tribes of hunter-gatherers are the same. In situations of extreme poverty all across the world, many cultures have accepted the practice of infanticide, the killing of one’s own children to reduce the number of mouths to feed in difficult times. This terrifying way of losing a child is just as painful as other ways, as consistently documented in traditional societies by anthropologists interviewing parents who had to kill a newborn—Pinker (2011, p. 417).

Page 69–70. Educating girls.

The data on girls’ and boys’ education comes from UNESCO[5]. Schultz (2002) describes clearly and in more detail how educating girls has proven to be one of the world’s best-ever ideas.

Page 73. Drownings.

The data on drownings today comes from IHME[4,5]. Up until 1900, more than 20 percent of the victims of drownings were children below the age of ten. The Swedish Life Saving Society started lobbying for obligatory swimming practice in all schools, which together with other preventive actions reduced the number; see Sundin et al. (2005).

Page 57. Graphs: Catching up with Sweden.

Use the animated version of the World Health Chart to see how almost all countries are now catching up with Sweden—or select another country to compare—at

Chapter Three: The Straight Line Instinct

Page 75. Ebola.

The data on Ebola is from WHO[3]. The material Gapminder produced to try to communicate the urgency of the situation is at You can see Hans explain the Ebola outbreak in the Factpods about Ebola here: and

Page 76. Lord Krishna’s Chessboard.

The Indian legend, depicting the effect of doubling, is called the Legend of the Ambalappuzha Paal Payasam, named after the temple where it supposedly happened. In calculating the volume of rice to cover India, the total number of grains Lord Krishna is owed were divided by India’s total land area. The grains’ estimated weight and volume—300 grains of rice equals 6 grams (thanks,—were converted into density per square meter, using this grain converter:

Page 78. Graph of Future Children and UN Population Experts.

The polling devices Hans used to test the audience at the teacher’s conference in Norway was from TurningPoint; see The 9 percent we are quoting on page 78, refers to the 7 out of 81 teachers who picked the correct answer. The dashed line at the bottom of the graph shows the official UN future forecast. The trend up to 2005 is the UN numbers for the global child population, ages 0–14. Two billion children is a rounded number. The precise UN numbers are 1.95 billion for 2017 and 1.97 billion for 2100. Hans showed this graph to the Norwegian teachers in August 2013. At that time, the forecast had been published two months earlier, but the numbers should not have been news. The line published in the edition two years earlier looked almost identical to the one Hans showed his teacher audience. In fact, the official future forecasts have stayed the same in the UN World Population Prospects the past four editions—throughout 2010, 2012, 2015, and 2017. The UN forecasts have been available to the public for almost 8 years and show that the global child population has stopped increasing; see You can find poll results from events, including this one, at

Page 79. Experts at the World Economic Forum.

For a video recording of Hans showing the audience results at the World Economic Forum in Davos 2015, see Poll results from events, including the results from the experts at WEF, are available at

Page 80. Graph: Historic Population Data.

The line showing the world population from 8000 bc to today uses data from hundreds of dif­ferent sources, compiled by the economic historian Mattias Lindgren. The sources listed under the chart are only the main sources. For more about our population data, see

Page 80. City Populations Today.

To compare the population in prehistory with the current population of three major cities, we used data from the UN Statistics Division published in the Demographic Yearbook–2015. The population was 6.5 million in Rio de Janeiro; 8.1 million in London; and 8.3 million in Bangkok; see Table 8, “Population of capital cities and cities of 100 000 or more inhabitants: latest available year, 1996-2015,” at

Page 82. UN Population Forecasts.

UN Population forecasts are based on UN-Pop[1,2,5]. Like forecasting the weather, it is almost impossible to perfectly predict the future population. The demography experts at the UN Population Division, however, have been very accurate in their forecasts for many decades, even before modern computer modeling was possible. For a video with Hans Rosling comparing these historic forecasts: It’s based on a paper form Nico Keilman who compiled the different forecasts from UN during the past decades and compared them: “Data quality and accuracy of United Nations population projections, 1950-95”, see

The National Research Council measured the accuracy of past projections and concluded that the average error in UN population projections from 1958 to 2000 was modest, only 2.8 percent—read the full chapter by Bongaarts, et al. (2000) here: UN forecasts for the future number of children have stayed the same in the past four editions of the publication. For more on the quality of UN forecasts, see On accuracy in UN population projections, see the article by Keilman (2010) available here: For the uncertainty intervals of the UN medium forecasts, see


Page 84. Graph: Babies per Woman.

We use the term “babies per woman” for the statistical indicator “total fertility rate.” We use UN-Pop[3] for post-1950 data. The long trend displayed on the graph shows Mattias Lindgren’s work that compiled multiple historical sources for the years before 1950. It is aligned with UN estimates from 1950 and onwards—see Gapminder[7]. Fertility rates decline during times of hardship. The dips and humps before 1965 shown on the graph were caused by the Second World War and famines. Fertility climbs to rates higher than usual when the crisis is over. In 1950, the rate was 5.05 children, according to UN-Pop[3]. The UN projection for 2017 is 2.48; the replacement rate with the current mortality worldwide is 2.3. The dashed line after 2017 shows the UN medium fertility projection, which is expected to reach 1.96 in 2099. See

Page 86. The Fill-up Effect.

We use the expression “the fill-up effect” to better explain this counterintuitive phenomena. The fill-up is also called “the demographic momentum”, a term demographers use for the delayed stopping of population increase. The process is almost identical to what we have simplified as the fill-up, where we only compare the size of cohorts. But future changes in populations can be attributed to three different factors: fertility, mortality, and momentum. To find out how a population will change, demographers make up a set of variations. By keeping fertility at replacement rate or letting mortality be constant, they can analyze the effect of different fertility and mortality scenarios. In that way, demographers can compare the outcome with their main scenario, and then attribute the difference to the factor they kept constant.

If you find it hard to understand the fill-up in the text and graphs in this book, we find it easier to explain with animations or with our own hands—see For more technical descriptions, see UN-Pop[6,7]. See also

Page 87. The Old Balance—Fertility and Mortality Before 1800.

The main sources behind our assumptions about fertility and mortality in pre-1800 families are Livi-Bacci (1989), Paine and Boldsen (2002), and Gurven and Kaplan (2007). Nobody knows the fertility rate before 1800, but 6 is a commonly used and likely average. See

Page 88. The New Balance.

The technical term “replacement-level fertility” is a measure of the average number of children that is needed for the next generation of a population to replace itself. The population curve is flat when the fill-up is completed and childbearing is on the level of replacement. About 2.1 children per woman is the often assumed replacement-level. But this is only the case in periods of good health and well-being in a population. The value requires that no more than 5 percent of children die. If 4 out of 6 people die before reaching childbearing age in a population, the average babies per woman would have to be 6. In “The surprising global variation in replacement fertility,” Thomas et al. (2004) describe the ranges in fertility. You can access the full article at

Page 90. Graph: Family Size by Income.

The graph presenting family size by income uses the most recent available data from World Bank[5] estimating that 10.7 percent of the population lived in extreme poverty in 2013. Gapminder[9] has extended the data to 2017, using IMF’s GDP per capita growth, and estimates that 0.75 billion people of the world population live on less than $2 a day; that is 10 percent of 7.55 billion people in the world. UN and the World Bank has set the threshold of extreme poverty at $1.9. Since these numbers are very rough estimates, Gapminder has rounded up the threshold to $2 a day. Our estimates for families on dif­ferent income levels are based on household data compiled by Countdown to 2030 and GDL[1,2], combining hundreds of households surveys from UNICEF-MICS, USAID-DHS[1], IPUMS, and others. Instead of using national averages, household data allow us to include families on Level 1, as well as the poorest families living in countries on Level 2 and 3. The income levels of households in these datasets are estimated from their material assets, for instance by the number of people per sleeping room, floor-material and means of transport. For more about the data and methods behind Gapminder’s four income levels, see

Page 91. Changing the typical family size.

For more on how societies transition from large to small families, see Rosling et al. (1992), Oppenheim Mason (1997), Bryant (2007), and Caldwell (2008). Babies per woman seem to start to increase again when people reach really high incomes on Level 4; see Myrskylä et al. (2009). This video shows how saving lives leads to fewer people:

Page 92. Two Public Health Miracles.

Life expectancy suffered an extreme dip in Bangladesh 1971 because of the Bangladesh war of independence. In 1972, the total fertility rate was 6.93 children per woman and life expectancy was 47 years according to UN, or 52 years according to IHME. In Bangladesh today, the total fertility rate is 2.07 children per woman and life expectancy is 72.8 years according to UN, or 72.7 according to IHME. Since 1972, the under-five mortality rate per 1,000 live births (U5mr) has reduced significantly. The mortality rate fell from 221.7 children in 1972 down to 34.2 in 2016. In other words, the child survival rate in 1972 was 778 children per 1,000, representing 77.8 percent. Today, the number is 966 out of 1,000 which means that 96.6 percent of children in Bangladesh survive. The child mortality rate was 313 children per 1,000 live births in Egypt 1960. Child mortality rate in Egypt and Bangladesh are based on UN-IGME. At we describe in full how we have combined child mortality data from multiple sources.

The construction of the High Aswan Dam began in 1960 to control flooding along the Nile. The dam was completed in 1970 and installed 1971; see In our animated World Health Chart you can see the progress of Egypt, Bangladesh, or most other countries—search by country and click Play:

Page 92. Child Survival.

See the talk at TEDxChange where Hans describes how reducing child mortality is both a moral and environmental imperative, here:

Page 93. 14 Straight Lines, S-bends, Slides, and Humps.

These 14 graphs of differently shaped lines are all derived from plotting two indicators against each other and then drawing a line in the middle—just like the World Health Chart you see in the beginning of the book. We removed the background bubbles in the small images, because it got too cluttered. Most of these charts use national averages, aggregated by the national income level; see Gapminder[3]. A few (the straight line on recreational spending, the S-bend on vaccinations and fridges, and the slide on fertility) use household data. Our estimates of typical families on different income levels are not based on country averages, because that would severely underestimate the number of the poorest and mask the wide range of differences for countries on levels 1–3. Very few countries follow these lines exactly, but the lines show the general pattern of all countries over several decades. In each example, there are huge differences between countries on every level. You can explore the actual plotted bubbles behind these lines here:

Page 97. The Growth of E.coli Bacteria

The generation time of E.coli bacteria, or the time required for the bacteria to double in number, is 15–20 minutes in a laboratory “but in the intestinal tract, the coliform’s generation time is estimated to be 12-24 hours,” as described in this bacteriology textbook:

Page 99. What part of the line are you seeing?

Many lines that are not straight can look straight if you zoom in enough—even a circle. This idea was inspired by Jordan Ellenberg, How Not to Be Wrong: The Power of Mathematical Thinking (2014). See

Chapter Four: The Fear Instinct

Page 105. Fear in Polls.

In the passage about the most common fears, we refer to those self-reported among adults. The polls, conducted in the United States and the United Kingdom respectively, reported similar results. In the Gallup[2] survey, top US fears were snakes; public speaking; and heights, followed by entrapment; spiders and insects; needles; ice; flying; and dogs. In the poll from YouGov[2], the leading fears in the United Kingdom were heights; snakes; and public speaking; followed by spiders; entrapment; mice; needles; and airplanes. Read more about the fear instinct at

Page 107. Disaster Data.

The International Disasters Database (see CRED) estimates that the earthquake in Nepal 2015 killed 9,034, injured 200,000 and affected 5.6 million people. The government of Nepal estimates a slightly higher death toll of 10,000 deaths; see PDNA. To avoid underestimating the suffering, we have used the higher number from PDNA. Numbers for the 2003 heat wave in Europe are from UNISDR, estimating the total death toll for Western Europe to be 46,730 fatalities. All other disaster data used here comes from EM-DAT. Nowadays, Bangladesh has a very advanced flood-monitoring website; see See

Page 110. ReliefWeb.

A specialized digital service of the UN Office for the Coordination of Humanitarian Affairs (OCHA), the ReliefWeb is a global humanitarian information platform. On their FAQ page—— they explain that their main function is to provide humanitarian workers with key information on global crises and disasters. The ReliefWeb’s yearly budget of 3.8$ million is funded by the OCHA and include contributions from Sweden, Japan, the United States, the United Kingdom, Denmark and others. In detailed reports, the ReliefWeb describes how the money was used, like this one,, stating “… the Red Crescent is also working with the United Nations’ World Food Programme to distribute 112.5 tonnes of biscuits to 30,000 families in Cox’s Bazar.”

Page 111. Child Deaths from Diarrhea.

Our calculations of child deaths from diarrhea caused by contaminated drinking water are based on numbers from IHME[11] and WHO[4]. See

Page 112. Plane accidents.

The data on fatalities in recent years is from the International Air Transport Association, IATA, and the data on passenger miles is from the UN agency that managed to reduce the number of accidents, see ICAO [1,2,3]. Gapminder[16] Airplane fatalities—v1. based on IATA, ICAO[3], BTS[1,2] & ATAA. See and

Page 113. War and Conflict: Deaths in wars.

The figure of 65 million World War II deaths includes all deaths and comes from White[1,2]. Estimates of fatalities in Syria are from UCDP[2].

Page 114. Graph: War and Conflict: Battle Deaths.

The data sources for battle deaths—Correlates of War Project, Gleditsch, UCDP[1] and PRIO—include reported deaths of civilians and soldiers during battle, but not indirect deaths like those from starvation. We strongly recommend watching this interactive data-driven documentary, which puts all known wars in perspective—at To interactively compare fatalities in wars since 1990, go to Measuring battle deaths is not trivial, as a war zone is no place for careful data collection. These studies publish numbers estimated by combining official sources and media reports from conflicts. But this method of estimating conflict fatalities has been disputed by several other researchers. Obermeyer and Murray showed in 2008 ( that the number of battle deaths in recent wars seems much higher if estimated with a different method, based on sample surveys of the local population in war-torn areas, who report how many family members they lost in conflict, with the so called sibling method. These authors claim explicitly that “there’s no evidence to support a recent decline in war deaths” since the Vietnam War. But the representativeness of the sample is not a trivial problem in these extreme events, and the number easily gets exaggerated when a local death toll is multiplied to a broader population. As there is little chance that new primary data will show up about past conflicts, the chances of new reliable estimates from other sources are small. The methodological discussion seems to have ended with a response in 2012 by Lacina & Gleditsch (see, making the case that their data-sources are indeed bias in an unknown way and that the bias may not be the same over the decades. Maybe the tendency is to under-report in some wars in some decade and then over-report in others. But still, even if they take into account the experts highest levels of doubts and use the widest reasonable uncertainty estimates and they try their hardest to generate an increasing impression of fatalities, by drawing a trendline from the lowest estimates of past conflicts to the highest estimates of the recent conflicts, even such line would be steadily falling. See

Page 114. Contamination: Fear of Nuclear.

The data on Fukushima is from the National Police Agency of Japan and Ichiseki (2013). According to police records, the Tōhoku earthquake and tsunami caused 15,894 confirmed deaths, and 2,546 people are still missing (as of December 2017). Tanigawa et al. (2012) concluded that 61 very old people in critical health conditions died during the hasty evacuation. About 1,600 further deaths were indirectly caused by other kinds of problems for mainly elderly evacuees, reports Ichiseki. According to Pew[1], in 2012, 76 percent of people in Japan believed that food from Fukushima was dangerous. The discussion of health investigations after Chernobyl is based on WHO[5]. Data about nuclear warheads is from the website Nuclear Notebook. See

Page 115. Contamination: DDT.

Paul Hermann Müller won the Nobel Prize in Physiology and Medicine in 1948 for “his discovery of the high efficiency of DDT as a contact poison against several arthropods.” Hungary was the first country to ban DDT, in 1968, followed by Sweden in 1969. The United States banned it three years later; see CDC[2]. An international treaty against various pesticides, including DDT, has since entered into force in 158 countries; see Since the 1970s, CDC[4] and EPA have issued directives on how to avoid the dangers of DDT to humans. Today, the World Health Organization promotes the use of DDT to save lives in poor settings by killing malaria mosquitoes, within strict safety guidelines; see WHO[6,7].

Page 116. Contamination: Chemophobia.

Gordon Gribble (2013) tracks the origin of chemophobia back to the publication of Silent Spring (1962), by Rachel Carson, and chemical accidents in the decades that followed. He argues that the exaggerated and irrational fear of chemicals today leads to wrong usage of common resources. See

Page 116. Contamination: Refusing vaccination.

In the US, 4 percent of parents think that vaccines are not important, according to Gallup[3]. In 2016, Larson et al. found that, across 67 countries, an average of 13 percent of people were skeptical about vaccination in general. There were huge variations between countries: from more than 35 percent in France and Bosnia and Herzegovina to 0 percent in Saudi Arabia and Bangladesh. In 1990, measles was the cause of 7 percent of all child deaths. Today, thanks to vaccination, it is only 1 percent. Deaths from measles mainly happen on Level 1 and Level 2, where children only recently started to get vaccinated; see IHME[7] and WHO[1]. See

Page 118. Terrorism.

The data about fatalities from terrorism comes from the Global Terrorism Database; see GTD.

The data on terror deaths per income level comes from Gapminder[3]. See Gallup[4] for the poll about fear of terrorism. See

Page 121. Alcohol deaths.

Our calculations on deaths involving alcohol draw on IHME[9], NHTSA (2017), FBI, and BJS. See

Page 122. Risks of dying.

The percentages we quote take the death tolls on Level 4 for the past ten years divided by the number of all deaths on Level 4 over that period, and are based on the following data sources: EM-DAT for natural disasters, IATA for plane crashes, IHME[10] for murders, UCDP[1] for wars, and GTD for terrorism. A more relevant risk calculation should not just divide by the number of all deaths, but rather should take into account exposure to the situations in which these kinds of deaths can occur. See

Page 122. “How Many Deaths Make a Natural Disaster Newsworthy?”

To compare dif­ferent kinds of disaster deaths, see “Not All Deaths Are Equal: How Many Deaths Make a Natural Disaster Newsworthy?” online at OurWorldInData[8]. Gapminder is currently compiling data about the skewed media coverage of dif­ferent kinds of deaths and dif­ferent kinds of environmental problems. When ready, it will be published here:

Chapter Five: The Size Instinct

(More notes coming soon for this chapter.)

Page 124. Nacala child deaths calculation.

The births and population data used for these calculations is based on the Mozambique census of 1970, the Nacala hospital’s own records, and UN-IGME of 2017.

Page 127. Basic Health Care: Saving lives.

The list of the low-cost, high-impact interventions that save the most lives comes from UNICEF[2], which also set out the essential basic health care to which all citizens should have access before public health budgets start being spent on more advanced care.

Page 128. Wrong proportions.

The examples of proportions that people tend to overestimate come from Ipsos MORI[2,3], which reveal misconceptions across 33 countries. Innumeracy (1988) by John Allen Paulos is full of fascinating examples of disproportionality, asking, for example, how much the level of the Red Sea would rise if you added all the human blood in the world. See

Page 129. Educated mothers and child survival.

The discussion on how educated mothers lead to higher child survival is based on a study of data from 175 countries between 1970 and 2009, by Lozano, Murray et al. (2010). See

Page 130. 4.2 million.

The data on infant deaths in recent years comes from UN-IGME. The data on births and infant deaths in 1950 comes from UN-Pop[3].

Page 132. Bears and axes.

This striking comparison was brought to the public’s awareness by a man named Hans Hansson. He wrote to his local newspaper about the absurd neglect of domestic violence against women and went on to start a network for men to help them break their violent behavior. Read an interview with him in English here:

Page 133. The Spanish Flu.

In his book America’s Forgotten Pandemic Crosby (1989) estimated that the Spanish flu caused 50 million deaths. The number is confirmed by Johnson and Mueller (2002) and CDC[1]. The world population in 1918 was 1.84 billion, which means this pandemic wiped out 2.7 percent of the entire global population.

Page 133. Tuberculosis (TB) and Swine Flu.

The data on swine flu comes from WHO[17] and the data for TB from WHO[10,11]. See

Page 134. Energy Sources.

The data comparing energy sources is from Energy Transitions: Global and National Perspectives by Smil (2016). Smil describes the slow transition away from fossil fuels and also debunks myths about food production, innovation, population, and mega-risks. See

Page 138. Graphs: West and Rest: Future Consumers.

For an interactive visualization of the graphs on page 138, see Two great books on this are The Post-American World by Fareed Zakaria (2008) and The World Is Flat by Thomas L. Friedman (2005).

Page 139. CO2 per capita.

The data on CO2 per capita for China, the United States, Germany, and India comes from CDIAC. See

Chapter Six: The Generalization Instinct

Page 159. Graph: Difference within Africa.

For an interactive version of the graph on page 159, see

Page 160. Contraceptives in Sweden.

The individual and national gains of making birth control easily available is described in “Children of the Pill: The Effect of Subsidizing Oral Contraceptives on Children’s Health and Wellbeing” (2012). The study, conducted by Andreas Madestam and Emilia Simeonova, looks at the long-term effects of Sweden’s subsidized birth-control policy in several municipalities between 1989 and 1998. Madestam and Simeonova report that women who decided to have children and were eligible for the subsidy achieved better health, economy and education for the next generation than women who decided to have children but were not eligible for improved access to birth-control.

The data on the use of contraceptives comes from UNFPA[1] and UN-Pop[9]. See

Page 160. Everything is made from chemicals.

People with chemophobia divide the world into “natural” (safe) and “chemical” (industrial and harmful). The world’s largest database of defined chemical compounds sees it differently. CAS contains 132 million organic and synthetic chemicals and their properties. It shows that toxicity is not related to who produces the compound. Cobratoxin (CAS registry number 12584-83-7), for example, which is produced by nature, paralyzes your nervous system until you can’t breathe. See

Pagi 161. The Salhi family on Dollar Street.

See more about the Salhi family at If you think we have too few homes from Tunisia or elsewhere on, feel free to contribute. Read more about how you can do it at:

Page 163. The recovery position.

Before the 1950s war surgery guidelines didn’t have a specific body position recommended  for the care of trauma patients. When the so-called NATO coma position had been successfully used by the US in the Korean War, however, it began to be promoted for saving unconscious soldiers—see Högberg and Bergström (1997). The recovery position was not standardized until decades later, in the early 1990s, when it emerged in general first aid handbooks. For more on the history of the recovery position, see Wikipedia[10].

Page 163. Hong Kong report on Sudden infant death syndrome (SIDS).

The conclusion that it was public health policy on the prone position that caused the increase in SIDS in Sweden is described by Högberg and Bergström (1997) and Gilbert et al. (2005). The report from Hong Kong is from Davies (1985).

Chapter Seven: The Destiny Instinct

Page 169. The sense of superiority.

For more on the sense of superiority over other groups, see Haidt, The Righteous Mind: Why Good People Are Divided by Politics and Religion (2012). See

Page 170. Societies and cultures move.

To see the World Health Chart in motion over 200 years, visit and click Play.

Page 170. Africa can catch up.

The data for life expectancy for countries and regions comes from Gapminder[4]. Paul Collier writes in The Bottom Billion (2007) about the future prospects for the world’s poorest people. Our rough estimate of people in extreme poverty close to conflicts is based on ODI (2015), preliminary results by Andreas Forø Tollefsen and Gudrun Østby of the number of people who live close to conflict worldwide (743 millions in 2016), and maps from WorldPop, IHME[6], FAO[4] and UCDP[2]. See the speed of improvement over the past decades here:

Page 171. Progress in China, Bangladesh, and Vietnam.

The Population Bomb by Paul and Anne Ehrlich (1968) contributed to a widespread idea that Asia and Africa would never be able to feed their growing populations. A total of 3.5 million people died in the so-called Bengal famine in 1942–1943: 2 million in Bangladesh and 1.5 million in India, based on the territorial boundaries today. The data on deaths from famines is from the International Disaster Database, EM-DAT. The Peace Research Institute Oslo (PRIO) produces maps of conflicts and poverty: For global textile production, see

Page 172. IMF forecasts.

Our comments on the IMF’s forecasting track record are based on the World Economic Outlook IMF[2]. See

Page 173. Fertility in Iran.

Today Iran has a lower fertility rate than the United States or Sweden. Professor Hossein Malek-Afzali at Tehran University of Medical Science was my host in Iran. He showed me the infertility clinic and taught me about Iran’s family planning and sexual education programs. To compare Iran—the world champion in family planning—against other countries over time, see

Page 176. Graph: Religions and babies.

In most countries, a majority of the population belongs to one of the large religions, and this guides which chart the country shows up in. However, in many countries there is no clear majority. In Nigeria, for example, 49 percent of the population was Christian and 48 percent Muslim in 2010 according to our data on religion, Pew[2,3]. We have split 81 such countries into three separate bubbles in the relevant charts, using Pew[2] and USAID-DHS[2] to estimate each religious group’s fertility rate, and roughly estimating each religious group’s per capita income based on GDL[1,2], OECD[3] and other sources. See:

Page 177. Asian values.

In “Explaining Fertility Transitions” (1997), Karen Oppenheim Mason discusses changing family norms. Gender roles change quite fast in all cultures as people get richer and their way of living is modernized. In cultures with an emphasis on extended families, values may change a bit more slowly. See

Page 178. Asian University for Women in Bangladesh.


Page 179. Nature reserves.

The data on protected nature is based on data from The World Database on Protected Areas (UNEP[5]), with the Protected Planet report (UNEP[6]) and IUCN[1,2]. The trend for 1911–1990 comes from Looking Ahead: The 50 Trends That Matter; see Abouchakra et al. (2016). See Gapminder[5] for details.

Page 180. Outdated Chimpanzee Questions.

In the 1990s, students at Karolinska Institutet did not know that many European countries had worse health outcomes than many Asian countries. These are the results that Hans show in his first TED talk—Rosling (2006). Thirteen years later, when we wanted to check whether people’s knowledge had improved, we could no longer use the original questions since the European countries had managed to caught up. This is shown in the animated graph here—at

Page 181. Attitudes toward same-sex marriage.

The data on attitudes toward same-sex marriage in the United States is from Gallup[5].

Chapter Eight: The Single Perspective Instinct

Page 187. Expert forecasts.

People with extraordinary expertise in one field score just as badly on our fact questions as everyone else. This wouldn’t surprise Philip E. Tetlock and Dan Gardner, the authors of Superforecasting (2015). In this book they describe a systematic way to test people’s ability to predict the future, and they find that one thing that can really impair good judgment is narrow expertise. Even forecasts by so-called experts in the media are often overly dramatic. When checking against actual outcomes, Tetlock and Gardner show that the experts not only score worse than the public but also worse than random. The personality traits that often come with good judgment are humility, curiosity, and a willingness to learn from mistakes. The Good Judgement Project, a forecasting services firm co-created by Tetlock, runs a public tournament called the Good Judgement Open. You can practice your forecasting here:

Page 188. Lindau Nobel laureate meeting.

This is a great annual gathering of brilliant young researchers who, thanks to this wonderful organization, get the chance to learn from Nobel laureates once a year. We are not criticizing that! We are just using their really low score on the vaccination question to make the case that expert knowledge doesn’t guarantee general knowledge. Read more about the presentation on the Lindau website:

Page 188. Poll results from groups of professionals.

For poll results for the groups of professionals mentioned here, and others, see

Page 189. Plundered natural resources.

For discussions about the commons and how to avoid exploitation, see The Plundered Planet: Why We Must—and How We Can—Manage Nature for Global Prosperity, by Paul Collier (2010), and IUCN Red List[4].

Page 193. Education needs electricity.

For more on this, see UNDESA.

Page 198. US health spending.

The spending data comes from WHO[12]. The comparison between US spending and spending in other capitalist countries on Level 4 comes from OECD[1], a study named “Why Is Health Spending in the United States So High?” It concludes that costs in the US health-care system are higher across the board, but in particular costs of outpatient care and administration; and that this does not lead to better outcomes, because the system is not incentivizing doctors to spend time with the patients most in need of care. See

Page 201. Democracy.

Paul Collier’s books are just as disturbing as they are fact-based. See his Wars, Guns and Votes: Democracy in Dangerous Places (2011) for more on how democracy can destabilize countries on Level 1 rather than make them safer. More disturbing problems with democracy are discussed in Fareed Zakaria’s The Future of Freedom: Illiberal Democracy at Home and Abroad. We must remind ourselves of Winston Churchill’s wise words: “No one pretends that democracy is perfect or all-wise. Indeed it has been said that democracy is the worst form of Government except for all those other forms that have been tried from time to time.” See

Page 201. Fast economic growth and democracy.

This discussion is based on economic growth data from IMF[1] and the Democracy Index 2016, from The Economist[2]. This index gives countries “democracy” ratings between 1 and 10, with the lowest score, 1.8, going to North Korea and the highest score, 9.93, to Norway. Here are the ten countries with the fastest economic growth over the past five years and their democracy scores (fastest first): Turkmenistan, 1.83; Ethiopia, 3.6; China, 3.14; Mongolia, 6.62; Ireland, 9.15; Uzbekistan, 1.95; Myanmar, 4.2; Laos, 2.37; Panama, 7.13; Georgia, 5.93. Only one of the ten fastest-growing economies scores well on democracy.

Chapter Nine: The Blame Instinct

Page 204. Neglected Diseases.

The World Health Organization has a list of neglected tropical diseases (See WHO[15]) are primarily affecting people living on income Level 1 and they are markedly neglected in research and technological development (R&D). This list of illnesses are not profitable to the pharmaceutical industry—only until recently, Ebola was on this list.

Page 207. Systems Thinking.

Peter Senge developed the idea of systems thinking within corporate organizations as a way of stopping people from blaming one another and helping them to understand the mechanisms that are causing problems. But his ideas apply to all kinds of human organizations where blaming individuals blocks understanding. See Senge, The Fifth Discipline: The Art & Practice of the Learning Organization (1990). See

Page 208. UNICEF’s Low Costs.

UNICEF’s streamlined logistics and supply chain are amazing. If you want to place a bid, you can see the supplies and services UNICEF is looking for right now at You can read more about its procurement process at UNICEF[5].

Page 212. Why Refugees Don’t Fly.

Sweden did not confiscate the boats of those smuggling refugees from Denmark during the Second World War—see the BBC documentary “How the Danish Jews Escaped the Holocaust.” According to Goldberger (1987), 7,220 Danish Jews were saved by these boats. Today, EU Council[1] Directive 2002/90/EC defines “smuggler” as anyone facilitating illegal immigration, and an EU Council[2] framework decision allows “confiscation of the means of transport used to commit the offence.” While the Geneva Conventions say that many of these refugees have the right to asylum, see UNHCR. See and

Page 214. CO2 Emissions by Income.

Researchers are trying to figure out how to adjust emissions quotas for changing population sizes; see Shengmin et al. (2011) and Raupach et al. (2014). See For more on CO2 emissions at dif­ferent incomes, see

Page 216. Syphilis.

If you think you are not living in the best of times, search for images of syphilis and you will soon feel blessed. We got the many names of this disgusting disease from Quétel (1990) via the University of Glasgow Library.

Page 216. 1 Billion People and Mao.

1 billion is a rounded-down approximation of the number of people whose lives were affected by Chairman Mao, who ruled China from 1949 until his death in 1976.

Page 216. Falling Fertility Rates and Powerful Leaders.

In 1949 China’s population was 0.55 billion. Between 1949 and 1976 the Chinese population increased by another 0.7 billion, according to UN-Pop[1]. From 1970 to 1976—preceding the 1979 one-child policy—the total fertility rate in China had already declined by half, see this graph: This interactive chart shows how all countries’ birth rates have fallen since 1800:

Page 218. Abortion.

The WHO guidelines on Access to Safe Abortion say: “Restriction in access to safe abortion services results in both unsafe abortions and unwanted births. Almost all deaths and morbidity from unsafe abortion occur in countries where abortion is severely restricted in law and/or in practice.” See WHO[2].

Page 218. Institutions.

Institutions are best understood through the work performed by the people maintaining them. In their book Poor Economics, Banerjee and Duflo (2011) describe the very basic institutions needed to make the escape out of poverty easier. See

Page 219. The governmental employees who saved the world from Ebola.

Dr. Mosoka Fallah is one of the Ebola contact tracers who Hans worked with in Monrovia. Listen to Dr. Fallah’s own words about the government’s employees and their commitment to their society when it needed them most, and watch him describe how to maintain trust within the community while hunting the infection, in his TEDx Monrovia talk—at

Page 219. The washing machine.

See the magic washing machine in action in this TED talk:

Chapter Ten: The Urgency Instinct

Page 223. Konzo.

To understand the lives of the villagers and their children suffering from konzo, watch the film by Thorkild Tylleskär (1995), recorded in the Bandundu Province, in present-day Democratic Republic of Congo:

Page 227. Now or Never.

Learn to defend yourself against common sales tricks in Robert Cialdini’s Influence (2001).

Page 227. The Urgency Instinct.

The instinct to think—and act—in oppositional pairs are deeply rooted in our evolutionary survival. For the gazelle, watching out for a possible predator, the difference between “yes” and “no” could be a matter of life and death. Humans, as well, are notoriously bad at keeping a reasonable range of options; like the gazelle, we tend to instinctively pick between opposites and choose the one that would best benefit our survival. A evolutionary remnant that once saved our lives, black-and-white-thinking nowadays blocks us from weighting the good and bad, yes and no, and select something more useful, such as a “maybe”. See Superforecasting by Tetlock and Gardner (2015) for more about our tendency toward urgent decision-making.

Page 232. The Melting Ice Cap.

The website Greenland Today shows the melting at the North Pole every day; see

Page 232. Fresh Numbers for GDP and CO2.

The OECD regularly publishes data for its 35 rich member countries. As of December 2017, the most recent number for GDP growth is from six weeks ago. The most recent number for CO2 emissions is from three years ago; see OECD[2]. For Sweden, CO2 emissions data that is not older than three months can be found at the website for Sweden’s System of Environmental and Economic Accounts; see SCB.

Page 233. Climate Refugees.

Many studies claim to show that the number of refugees will increase dramatically because of climate change. The UK Government Office for Science showed in their study Migration and Global Environmental Change (Foresight, 2011) fundamental weaknesses in the common assumptions underlying these claims. First it found that most of the frequently quoted studies refer back to just two original sources, one estimating that climate change will create 10 million refugees and the other anticipating 150 million; see Box 1.2: “Existing estimates of ‘numbers of environmental migrants’ tend to be based on one or two sources.” And second, it found that these original sources underestimate people living on Levels 1 and 2 and their ability to cope with change. They describe migration as their only option in the face of climate change. For a fact-based picture of global migration and the refugee situation, see UNHCR Population Statistics—at Important books on global migration are Alexander Betts’s and Paul Collier’s Refuge (2017) and Collier’s Exodus (2013).

Page 234. Ebola.

The WHO[13] lists all situation reports produced to track the Ebola pandemic since 2014. They still show suspect cases, and the CDC[3] continues to use the high estimates, which include suspected and unconfirmed cases.

Page 237. The Five Global Risks.

For a fact-based view of a longer list of major risks, see Global Catastrophes and Trends: The Next Fifty Years by Smil (2008). For those who find numbers calming, this is where you will find the big picture of the proportional risks and uncertainties of all kinds of possible fatal discontinuities. See

Page 237. The risk of global pandemic.

A small version of the Spanish flu is more likely to occur than a large one; see Smil (2008). While we should work against the obscene overuse of antibiotics in the meat industry—see WHO[14]—we must also be careful not to make the same mistake as we did with DDT and become overprotective. Antibiotics could save even more lives if they were less expensive. See

Page 238. The risk of financial collapse.

During the past ten years, the external environment is volatile, with capital markets increasingly characterized by more extreme events, observe Dobbs et al. in No Ordinary Disruption (2016), illustrated by the peaks in the trend line on page 88. In his article “How Should We Prevent the Next Financial Crisis?” (2015) Ricardo Hausmann suggests that since the financial system constantly changes, it is difficult to learn from past mistakes—our next financial crisis will simply not look like any other that has happened before. Many countries have regulated their financial systems to prevent future crises by restricting loans and risk-taking. While this strategy is good for some countries, preventative measures do not benefit all. A number of countries in Latin America for example, Hausmann writes, have instead made their economies too secure. Read Hausmann’s article in full at See

Page 239. The risk of World War III.

In his book (2008), Smil was already discussing ten years ago how six unfolding trends of the new world order were slowly leading to intensified conflicts between parts of the world: Europe’s place, Japan’s decline, Islam’s choice, Russia’s way, China’s rise, and the United States’ retreat. See

Page 239. The risk of climate change.

This passage draws on The Plundered Planet by Paul Collier (2010), OurWorldInData[7] and the thinking of economist Elinor Ostrom. In her work Governing the Commons (1990) Ostrom explores the way in which humans across the world has managed to regulate their common resources to avoid overexploitation. Read more about Ostrom and her design principles of common pool resource administration—at


Page 240. The risk of extreme poverty.

The passage draws on World Bank[26], ODI, PRIO, Paul Collier’s The Bottom Billion (2007), and the BBC documentary “Don’t Panic—End Poverty,” Gapminder[11]. While extreme poverty has fallen, the number of extremely poor people living in conflict has been stable or even increased, based on preliminary data from PRIO. If current wars continue, soon the vast majority of extremely poor children will live behind military lines. This poses a cultural challenge to the international aid community. The International Dialogue on Peacebuilding and Statebuilding had its fifth global meeting in Stockholm, 2016. The meeting, following the Stockholm Declaration against war and conflict, discussed how the aid community is now preparing for the risk of extreme poverty in conflict prone countries—see See

Chapter Eleven: Factfulness in Practice

Page 247. Teachers.

Visit to find our free teaching materials and join the community of teachers promoting a fact-based worldview in their classrooms.

Page 251. Diversified economies.

MIT has produced a free-of-charge tool ( to help countries work out how best to diversify, given its existing industries and skills; see or read Hausmann et al. (2013).

Page 252. Speling miskates.

This typo is intentional, inspired by the fact that oriental rugs should always contain at least one deliberate mistake. At least one knot must always be wrong in every rug. It is to remind us that we are humans and we should not pretend we are perfect. Deliberately, we have no source behind this fact.

Page 252. Constructive news.

Here are two very dif­ferent approaches for fixing the news problem: and

Page 253. Local ignorance and data.

Don’t miss Alan Smith’s TEDx talk “Why you should love statistics” where he shows great examples of local misconceptions in the UK. Gapminder is starting to develop localized visualizations, like these about Stockholm: see Each bubble represents a small area of the city. The differences in income are disturbingly large, but just like the global picture, the areas in Stockholm are NOT divided into two groups: poor vs. rich, as it’s usually sounds like in the Swedish news. Push play and see how most of Stockholm is getting richer and more educated, despite the common feeling of decline.

A Final Note

Free global development data.

Open access to data and research made this book possible. In 1999, the World Bank produced, on a CD-ROM, the most comprehensive set of global statistics ever: “World Development Indicators.” We uploaded the content to our website in our animated bubble graphs to make it easier for people to use. The World Bank got a bit angry, but our argument was that taxpayers had already paid for this official data to be collected; we were just making sure they could reach what they already owned. And we asked, “Don’t you believe in free access to information in order for global market forces to work as they should?” In 2010, the World Bank decided to release all of its data for free (and thanked us for insisting). We presented at the ceremony for their new Open Data platform in May 2010, and since then the World Bank has become the main access point for reliable global statistics; see

This was all possible thanks to Tim Berners-Lee and other early visionaries of the free internet. Sometime after he had invented the World Wide Web, Tim Berners-Lee contacted us, asking to borrow a slide show that showed how a web of linked data sources could flourish (using an image of pretty flowers). We share all of our content for free, so of course we said yes. Tim used this “flower-powerpoint” in his 2009 TED talk—see—to help people see the beauty of “The Next Web,” and he uses Gapminder as an example of what happens when data from multiple sources come together; see Berners-Lee (2009). His vision is so bold, we have thus far seen only the early shoots!

Unfortunately, this book uses almost no data from the International Energy Agency (, which, together with OECD, still puts price tags on lots of taxpayers’ data. That probably will—and has to—change soon, as energy statistics are way too important to remain so inaccessible.

Date Posted: 2018-03-08