This entry focuses on the history of famine and famine mortality over time. Our data include information only up to 2016. It does not include any data on current food emergencies. More information on current crises can be found at FEWS.net.
A famine is an acute episode of extreme hunger that results in excess mortality due to starvation or hunger-induced diseases.1
It is this crisis characteristic that distinguishes it from persistent malnutrition, which we discuss in another entry on this website. As we discuss in the Data Quality and Definition section below, the term ‘famine’ can mean different things to different people and has evolved over time. It is only in recent years that more precise, measurable definitions – in terms of mortality rates, food consumption and physical signs of malnutrition – have been developed.
But despite these ambiguities, it is nonetheless very clear that in recent decades the presence of major life-taking famines has diminished significantly and abruptly as compared to earlier eras. This is not in anyway to underplay the very real risk facing the roughly 80 million people currently living in a state of crisis-level2 food insecurity and therefore requiring urgent action. Nevertheless, the parts of the world that continue to be at risk of famine represent a much more limited geographic area than in previous eras, and those famines that have occurred recently have typically been far less deadly – as we will go on to show in this entry.
For this entry we have assembled a new global dataset on famines from the 1860s until 2016. We estimate that in total 128 Million people died in famines over this period.3
Famines have always occurred as the result of a complex mix of ‘technical’ and ‘political’ factors,4 but the developments of the modern industrial era have generally reduced the salience of natural constraints in causing famine. This includes many developments discussed in other pages of Our World in Data, such as the increasing availability of food per person, made possible through increasing agricultural yields; improvements in healthcare and sanitation; increased trade; reduced food prices and food price volatility; as well as reductions in the number of people living in extreme poverty. Over time, famines have become increasingly “man-made”-phenomena, becoming more clearly attributable to political causes, including non-democratic government and conflict. Paradoxically, over the course of the 20th century famine was virtually eradicated from most of the world, whilst over the same period there occurred some of the worst famines in recorded history. This is because many of the major famines of the 20th century were the outcome of wars or totalitarian regimes. As such, the waning of the very high levels of warfare over the last decades (as seen in the reduced number of battle deaths in recent times) and the spread of democratic institutions has also played a large part in the substantial reduction in famine mortality witnessed in recent decades.
Emergency food aid provided by relief agencies continues to play a crucial role in preventing loss of life, and the international relief community has recently developed much better monitoring systems, such as the Famine Early Warning System, that has allowed for greater preparation and more timely interventions. Where poor harvests are the main cause of famine, as in Niger in 2005, relief provision tends to prevent marked increases in mortality. It is the presence of conflict, or abuses of political power that can block food supplies reaching populations which represents the most pertinent trigger of ‘death-dealing’ famines today.5
Thus, overall, we can see in the rapid decline of famine mortality one of the great accomplishments of our era, representing technological progress, economic development and the spread of stable democracies. Viewed in this light, however, it also serves to highlight the appalling continued presence of famines which are, in the modern world, entirely man-made.
The entry is based on a global dataset of famines since the mid-19th century produced by us. This ‘Our World in Data-Dataset of Famines’ can be found at the very end of this document and is preceded by a discussion of how this dataset was constructed and on which sources it is based.
All our charts on Famines
Compared to earlier historical periods, very few people have died in famines in recent decades. Here we show two bar charts based on our dataset of famines. The blue bars show the number of famine deaths in each decade since 1860. The number of famine deaths varies hugely from decade to decade depending on the occurrence of individual catastrophic famines. Nevertheless the last four decades have seen low numbers of famine deaths by historical standards.
The red bars show famine mortality relative to the growing world population over this time. Viewed in this way, the trend is all the more notable. The chart shows the rate of famine deaths globally, expressed as the number of people dying each year per 100,000 people of the world population. The chart presents this rate averaged across each decade since 1860. You can see that the famine mortality rate fell to very low levels over the second half of the 20th century onwards.
The sharp reduction in famine mortality represents “one of the great unacknowledged triumphs of our lifetime”, as famine researcher, Alex de Waal describes it.6
As de Wall explains, a continued decline is by no means assured: the future of famine will depend largely on the nature and prevalence of war. However, as we discuss here, the long-run developments that have contributed to the sharp decline in famine mortality do suggest that the type of catastrophic famine seen in the nineteenth and twentieth centuries is very unlikely to return.
The geographic spread of famines has also reduced over this period, as we can see in these two charts, which give two ways of visualizing famine deaths by continent.
While in earlier periods Asia suffered heavily from major famines, this came to a halt over the mid-20th century. Since then, famines have almost exclusively been restricted to Africa, with the famine in North Korea being a stark exception.
This chart shows the estimated number of people dying in individual famines since the 1860s, based on our dataset of famines. The length of each line shows the duration of the famine and the color shows the continent in which the famine occurred.
What this chart doesn’t show however is the significant uncertainty that surrounds many of these estimates. As we discuss in our section on Data Quality, historical famine estimates are frequently based on very scant demographic information, and even where such evidence is available there is still disagreement in its interpretation.
As such, many of the famines included in our dataset are associated with a wide range of plausible mortality estimates. In our data, these are represented by upper- and lower-bound estimates, with the mid-point being shown in the visualization above. Two apt examples are the famines in the Democratic Republic of Congo, which took place amidst the Second Congo War beginning in 1998, and that of North Korea from 1995 to 1999. These famines stand out in recent decades for their particularly high mortality. But in both cases, the range of mortality estimates available in the literature is large, with high and low estimates varying by several millions of deaths.12
Where such differences are present, our midpoint estimates are clearly very sensitive to our choice of upper and lower bounds. We have not simply taken the highest and lowest figures published in the public domain, given that more accurate estimates often emerge with time. As noted by the World Peace Foundation, “generally speaking, better demographic calculations lead to lower estimations of excess deaths than those provided by journalists and other contemporary observers.” Rather, we have sought to select upper and lower estimates based on the balance of opinion in commonly-cited sources, all of which are detailed (for each individual event) in our dataset of famines. Inevitably though, this meant us taking a position in some controversial cases, discussed in detail here.
But whilst the number of deaths caused by individual famines is often subject to a good deal of uncertainty, the overall trend over time is very clear: compared to earlier historical periods, far fewer people have died in famines in recent decades. Whether we consider high or low estimates, or something in between, does not affect this conclusion.
Number of deaths and duration of individual famines, 1860s-201611
In today’s developed countries peacetime famines had largely ceased by the mid-19th century.13
In England this was achieved at least a century earlier. This graph shows estimates of the crude population increase – the number of births minus the number of deaths divided by the population – taken from Campbell (2009).14
These figures are based on a national sample of parish register entries, which are available with good coverage from 1538 when the registration of baptisms, marriages and burials become enforced. By comparing the price of grain (which was well documented in England from the twelfth century onwards) with estimates for real wages and grain yields, the author was able to make reasoned speculations about which of the documented sudden drops in population were likely to have resulted from famine, as opposed to disease epidemics unrelated to food shortages. The idea is that even if harvests were bad, if there was no simultaneous rise in grain prices it seems more likely that disease would have been the main driver of population losses, as opposed to famine (the lower population reducing the demand for food, thereby offsetting the reduced supply to keep prices roughly level).
Those population crises potentially consisting of famine conditions are indicated with an ‘F’ in this visualization. The last was in 1741-2 which was “brought on by an extreme short-term weather anomaly of at least three-years’ duration” that affected much of northwest Europe, causing an even more severe famine in Ireland.
Estimated crude rates of natural increase in England, 1550–1849, with possible famines highlighted – Campbell (2009)
Saito (2010) has created a chronology of famines in Japan since the 6th century. Before the 14th century data is judged to be incomplete (although the records for the 8th and 9th are surprisingly complete – there were more than 35 famines in each of the two centuries).
The number and intensity of famines is shown as ‘points’ in Saito’s visualisation shown here: 1 point being given to widespread famines, 0.5 points being given to more localised events.
Overall Saito’s chronology comprises information on 281 famines. None of these 281 famines happened in the twentieth century and the graph here shows that the end of starvation in Japan arrived gradually. Before 1550 there were more than 10 famines per 50 year-interval and since then famines have became less and less common in Japan.
The number of famine points by half-century, 1300-1900 – Saito (2010)15
It is very difficult to know how common famines were in the distant past given the absence of historical record.
As noted by the World Peace Foundation, “generally speaking, better demographic calculations lead to lower estimations of excess deaths than those provided by journalists and other contemporary observers. We might therefore reasonably expect an upward bias in the figures for earlier famines on the record.” On the other hand, there is an obvious risk that existing historical records underreport long-past famines and the number of their victims due to the lack of documentation being made at the time or their being lost subsequently. Loveday, an early researcher of Indian famines, noted in 1914 that, “The frequency of the mention of famine in the later history […] increases in exact proportion with the precision and accuracy in detail of her historians.”16
At least in proportionate terms, it seems safe to conclude that the nineteenth century suffered far more intensely from famine than did the twentieth century, with Ó Gráda (2007) considering one hundred million deaths a ‘conservative estimate’ for the nineteenth century as a whole: higher than the combined figure for the twentieth century, and in the context of a much lower population.17
On the other hand, it seems unlikely that famines dominated to the degree assumed by some early famine scholars such as Robert Malthus, not least because ‘normal’ mortality rates would have been very high anyhow. “Given that life expectancy was low even in noncrisis years, frequent famines would have made it impossible to sustain population”, concludes Ó Gráda (2007).
Based on consideration of a patchwork of burial records and other historical accounts, Menken and Watkins (1985) conclude that famines in which death rates doubled for two years or more were ‘rare’, and that famines of even greater intensity were ‘highly unusual’, if they occurred at all.18
From what evidence there is, it seems unlikely that famine served as a primary check to population growth in the past, with non-crisis malnutrition and disease generating high enough death rates to act as “more potent positive checks on population growth in the long run than the Third Horseman.”19
We might naturally tend to associate famine with drought or other natural phenomena, and indeed most documented famines have occurred in the context of harvest failures, often due to droughts or flooding. However, in recent times, aggregate food availability per person has increased dramatically, and given the comparable ease of transportation and communication, localized shortfalls can – in theory at least – be met by importing food from surplus areas far quicker and at a much lower cost nowadays. As such, lack of overall food availability per se plays a less prominent role in causing famine today than it did historically.
Better integrated food markets have on the whole helped to ease acute localized food price volatility due to bad harvests. Food shortages that lead to higher prices create an incentive for traders to increase the supply of food, thereby preventing shortages developing into outright food crises.
Thus the absence of markets, or presence of badly functioning markets, can be a key part of why people are not able to obtain enough food. Where means of transport is lacking, trade between surplus and deficit regions can be hampered, as well as making the distribution of food aid much harder during crises. Comparable climatic conditions that sparked two famines in northern China, in the 1870s and 1920s respectively, brought about 9-13 million deaths in the first case and half a million in the latter. One important difference that can explain this moderated impact is the availability of greatly improved transport infrastructure that was constructed in the interim which allowed for prompt relief efforts to take place. The more limited development of transport infrastructure in parts of Africa has played a contributory role in a number of recent famines on the continent.20
Where markets function badly, supply may be restricted ‘artificially’. For example, Amartya Sen argues that ‘speculative withdrawal and panic purchase of rice stocks’ was one of the primary causes of the Bengali famine of 1943, which turned a ‘moderate short-fall in production… into an exceptional short-fall in market release’.21
Where traders have some monopoly power over local markets, hoarding can be a way of increasing profits by making prices rise. Even without monopoly power, where traders collectively expect prices to increase, for whatever reason, it can make sense for them not to sell storable food to final consumers immediately, but rather wait for the higher prices, thereby restricting the current overall supply to consumers. Again, this is part of the normal functioning of a market which encourages food to be transferred from periods of relative plenty to those of relative scarcity. But where such trading leads to excessive speculation on price increases, ‘price bubbles’ can emerge such that prices no longer bear any relation to the actual relative scarcity any more. In this sense badly functioning markets can produce ‘artificial’ scarcities, where food is prevented from reaching final consumers not because of actual falls in production, but only due to the anticipation of higher future prices.
Such self-fulfilling expectations of price increases can occur simply where people have mutually reinforcing, but nonetheless mistaken beliefs about future supply. According to Ravaillion (1987), such a dynamic was indeed at play during the Bangladesh famine, in which food prices soared despite there being no significant drop in food production or in overall food availability per person.22 He suggests that the severe flooding that occurred during the famine created the expectation of a shortfall and related price increases, but that the resulting panic buying and price speculation themselves brought about the scarcity, rather than any realised drop in production. It is argued by others that food price speculation at the time was directed towards a perceived weakness in the government’s ability to continue with a policy of buying food at below-market prices in order to keep prices from rising too much.23
For instance, where a weather event (such as the severe flooding that occurred during the Bangladesh famine of 1974) makes people think there will be shortages, resulting panic buying and price speculation can itself then produce scarcity artificially.
The chart shown, taken from Ó Gráda (2006), shows the very dramatic peak in food prices during the Bangladesh famine that happened despite there being no decline in overall food availability. It also shows a sharp increase in the differences in food prices between different regions in Bangladesh (as measured by the standard deviation). This is evidence that, during the famine, markets became more spatially segregated – i.e. that food was not able to move to those regions where it was in highest demand, and thereby lower local price differences. The absence of properly functioning markets thus contributed to the localized scarcities.
International aid continues to play a large role in addressing food security, both in emergency situations and to help relieve more persistent periods of food insufficiency (the World Food Program collects data on the quantity and value of international food aid and is available here). The development of better monitoring systems, such as the Famine Early Warning System, has given the international relief community more advanced notice of developing food crises, although such early warnings by no means guarantee a sufficient aid response, nor that secure access to affected areas will be granted.
Thus, whilst drought or flood-caused crop failure might naturally seem to be high up on a list of causes of famine, this was far truer of famines in the past. Nowadays, crop failure is better understood as an important contributing factor rather than a sufficient cause of famine: food crises due simply to localized drops in production do not tend to develop into full-blown famines with high excess mortality unless exacerbated through more overtly human influences.
Average Rice Prices and Regional Variation in Bangladesh 1972-7524
As Amartya Sen argued, the fact that there may be enough food available in aggregate within a given area does not necessarily mean that everyone will be able to afford it.25
Food crises are often precipitated by spikes in the price of food relative to wages, or the collapse in the price of assets owned. The latter commonly accompanies famines due to many people all at once trying to sell their assets (for instance their livestock) in order to be able to buy more food. Such shocks can mean that those already living close to the level of subsistence may find their ‘exchange entitlement’ – that which they can obtain on the market in exchange for their labour or other assets – fails to provide them with enough food, even if the aggregate local supply is sufficient.
Wealthy countries have very few people living in such extreme absolute poverty, both because of higher incomes before tax and benefits, but also due higher government expenditures and transfers. It is therefore unsurprising that those countries in which famines occur do tend to be very poor. Here we show the inflation-adjusted income per capita of each country at the time they experienced a famine, with some reference points on the vertical axis. As you can see, most countries in which a famine took place had, at the time, average incomes less than half of that of the UK at the outset of its industrial revolution. You can see that average incomes in India – a country that historically suffered very heavily from famine – has grown rapidly in recent decades, and at the same time has been famine free. African countries, conversely, have on the whole has remained very poor and make up the majority of recent famines.
As well as proxying for the presence of extreme poverty, this relationship also reflects the fact that poorer countries also tend have less adequate facilities like transport infrastructure, sanitation and systems of healthcare that play a key role in preventing or moderating the impacts of food shortages.
Whilst poverty certainly increases the vulnerability of a country, we should be careful not to think of it as the single, or even the most important, cause of famine, given the typically political nature of most outbreaks of famine.
Famines and real GDP per capita, 1860-201626
Amartya Sen famously noted in his 1999 book Development as Freedom that “there has never been a famine in a functioning multiparty democracy”. He suggested that democratic authorities are incentivized by elections to be more responsive to food crises and that the presence of a free press can quickly draw attention to the event and hold government to account.
Whilst exceptions to this rule can be found – depending on the definition of ‘democracy’ and ‘famine’ being employed – the visualization here corroborates the idea that famines tend not to happen in democracies, by grouping them according to the political regime under which they took place. Here we use our list of famines since 1850 which can be found at the bottom of this page, and we define the political regime type according to the Polity IV score (discussed more in our entry on Democracy), collecting the various scores into three clusters: Democracy (>5), Autocracy or Anocracy (-10 to 5), and Colony (-20).
According to the definitions we have adopted, three famines since 1850 took place in democracies. However, in two of them the ‘democratic’ classification is rather ill-fitting. The St. Lawrence Island famine of 1878-80 is listed as occurring in the USA. However, it took place on a very remote Alaskan island populated by the indigenous Yupik people, that had relatively little interaction with mainland USA. In the case of Sudan, according to its Polity IV score, there was a brief spell of democracy, following elections held in 1986. But it is somewhat misleading to consider the famine occurring in southern Sudan in 1988 as happening under conditions of a functioning democracy. It took place during the Second Sudanese Civil War, which was organized primarily along a North/South division and marked by many human rights violations. According to Wikipedia, in the majority of seats in the southern regions voting was in fact suspended for the 1986 election.
In the third case, that of Maharashtra in India in 1972-3, whilst there is no ambiguity as to the political regime type, there has been some disagreement as to whether excess mortality in fact occurred. This is discussed further in the Data Quality and Definition section below. In any case, whilst in absolute terms it is certain that the drought caused enormous suffering, whatever excess mortality that did occur in Maharashtra was very much lower than the major famines occurring under totalitarian regimes in roughly the same period.
It is important to note, however, that the question of how often famines have occurred within democracies crucially depends upon the definition of famine being used. In particular, what, if any, excess mortality lower-bound is being used yields different answers. As Thomas Plümper and Eric Neumayer (2007) point out, a number of smaller-scale events in which drought-related mortality did occur have happened in functioning democracies.28 As the authors argue, even within democracies it can still be politically advantageous for governments to allow small minorities to starve if in doing so they are able to win more votes by distributing benefits to others. As discussed in the Data Quality and Definition section below, in compiling our table we have omitted events where the excess mortality is estimated to be lower than one thousand deaths, to reflect that the term “famine” has in its common usage typically been reserved for larger-scale events with crisis characteristics. This leaves only the three democracy famine events discussed above. Regardless of the threshold though, the main point remains the same: famines tend not to occur in democracies, and none of the catastrophic life-taking famines documented in history have occurred in the context of functioning democratic institutions.
Famines since 1850 by political regime27
Many of the major famine events in our table occurred as the consequence of international or civil war. For some of these, famine was used as an intentional part of political or military strategy. The ‘Hunger Plan’ pursued by Nazi Germany as part of its attempted invasion of the Soviet Union is an example of the latter. Despite the plan only being partially executed, over 4.5 million famine deaths are attributable to the offensive, significantly more than have occurred globally since the turn of the 21st Century.29
In terms of more recent events, from the second half of the 20th century onwards, famines in Africa have become increasingly associated with civil war, and include a number of crises in places that were not previously prone to famines at all, such as Mozambique and Biafra in Nigeria. In addition to the direct casualties, conflict can also generate disruption to production and trade and can encourage the spread of disease epidemics, particularly through forced migration.30
Crucially, it can also block the arrival of humanitarian relief to those in need.
It is mainly in the context of conflict that major death-dealing famines can be expected today. The majority of ‘areas of concern’ currently noted by the Famine Early Warning System are listed with armed conflict mentioned in the ‘reasons for concern’.31
It is important to note that, as opposed to dying from literal starvation, the vast majority of people that die during famines actually succumb to infectious disease or other illnesses, with some diseases being more directly linked to diet than others. Famines brought on by drought often go hand-in-hand with a scarcity of clean drinking water that increases the threat of cholera and other diseases. Increased migration and the disruption of personal hygiene and sanitation routines and healthcare systems also increases the risk of outbreaks of infections diseases, all in the context of a population already weakened through malnourishment.
This is particularly true in places where such diseases are already endemic. Thus, in sub-Saharan Africa where vaccination rates for measles have been relatively low, the disease has been a big killer during modern famines in the region alongside other infectious and parasitic diseases common in non-crisis times.32
The table here is taken from Ó Gráda and Mokyr (2002) and shows the percentages of excess deaths during a selection of famines attributable to different proximate causes, including the most typical famine diseases. In each case, it can be seen that communicable diseases were the ultimate cause of death in the majority of cases. (Note that, for India and Moscow, the excess mortality attributable to starvation is not available separately).
This is in contrast to some famines that occurred in industrialised countries during WWII, in a context of overall healthier populations and systems of sanitation that were maintained to some degree despite the crisis. In these instances disease played far less of a role, with deaths from starvation correspondingly higher. Ó Gráda (2009) gives the example of the siege of Leningrad in which “few of Leningrad’s 0.8 million or so victims perished of contagious diseases,” noting that the number of people dying from the main infectious diseases were actually lower in 1941 – amidst an overall vast increase in excess mortality – than they had been in 1940 before the blockade began.34
Causes of excess deaths in selected famines in Ireland, Russia and India – Ó Gráda and Mokyr (2002)33
In February 2017, parts of South Sudan were officially declared by the UN as being in famine – the first such declaration since 2011. By May the famine had apparently receded, thanks to an effective aid response that averted large-scale loss of life. And yet, the crisis was far from over. Indeed the overall food security situation in the country had, in fact, ‘further deteriorated’ over the same period, according to official reports35 – even as the ‘famine’ status was being withdrawn.
Here we look in more detail into the famine declaration in South Sudan, in order to understand better how famine is defined today and how this fits in with our understanding of famines in the past.
The IPC lays out thresholds across three dimensions of outcomes, all of which must be evidenced in order for a famine to be declared in a certain area:
- Food consumption and livelihood change: At least one in five households faces an extreme lack of food, as evidenced by insufficient food consumption (in terms of quantity and/or quality) and by the coping strategies employed by or available to households (e.g. the extent to which assets like livestock or seed may have been sold off)
- Nutritional status: More than 30% of the population is suffering from ‘wasting’ (being below 2 standard deviations below the median weight for a given height in the reference population)
- Mortality (due to inadequate food consumption):36At least two people out of every 10,000 and at least four children under five out of every 10,000 are dying per day.37
A few things are worth noting about this definition. Firstly, these thresholds represent only the most severe rank of the IPC food insecurity classification. The system ranges from Phase 1 to Phase 5, with 5 corresponding to a famine situation. Lower phases of food insecurity are categorized by lower thresholds in each of the three dimensions above.
Secondly, it is important to see that such thresholds are a measure of intensity rather than magnitude.38 That is to say, rather than trying to capture the absolute number of people in a certain situation of food insecurity, it looks at proportions within given geographic areas. Thus different assessments of food security trends will often be made depending on the geographic level of analysis. An amelioration at a very local level is perfectly compatible with an overall deterioration of the food security status of a country as whole. And this is exactly what happened in South Sudan over the course of 2017.
Here we show two maps of South Sudan showing the IPC classification of each county of South Sudan, in January and May 2017. It was the intensity of the food security situation in Unity State in January (shown in dark red colors), which brought about the famine declaration later in February, with IPC Phase 5 thresholds being confirmed in some parts.
By May the situation in Unity State had somewhat abated due to humanitarian relief efforts, but the food security situation of most other parts of the country had deteriorated significantly. Thus while the ‘famine’ was over – in the very particular sense of there being no area where intensity thresholds met Phase 5 criteria – the food emergency had in fact become worse for most people.
Just as different parts of a country can have different food security statuses, different households can, and typically do, experience different levels of food insecurity within any given geographic area. Rather than looking at geographical subdivisions, one way of getting a sense of how different people are faring in a food emergency is to look at the numbers of individual households experiencing different levels of food insecurity.
The IPC sets out such a ‘Household Group Classification’ alongside the ‘Area Classification’ outlined above. It mirrors the area classification in providing a Phase classification from 1 to 5, with 5 consisting of a ‘Catastrophe’ situation for the household. Since nutritional status and mortality data are typically collected for whole populations in a given area, only the food consumption and livelihood change dimension is used to categorize food security at the household level – though signs of malnutrition or excess mortality within the household are used to confirm the presence of extreme food gaps at the higher insecurity rankings.41
So whilst the household-level classification considers fewer outcomes (only food deficits, as opposed to nutritional or mortality outcomes), it does allow for an assessment of the magnitude of a food emergency in terms of the absolute number of people being affected at different levels of severity. Looking at the household data for South Sudan over 2017 offers another angle on the evolution of the crisis. The two tables shown give the number of people estimated to be at a given level of insecurity across the different States in January (first table) and May (second table).
With such a disaggregation we can see that the humanitarian provision, targeted to the most in need in Unity State, did indeed bring down the number of people experiencing the very worst food insecurity. It was on this basis that that country was no longer officially ‘in famine’. However, if we look at the number of individuals in Phase 3 (Crisis) or worse food insecurity, we see not only a deterioration in the country as a whole (45.2 % of the population in May compared to 32.3% in January), but even in Unity State itself (with 58.7% and 54.7% respectively).
Thus whilst a binary ‘famine/no-famine’ categorization is very useful in terms of being able to draw international attention and relief efforts to the most dire situations, there are other dimensions that we should be aware of in trying to get a sense of the gravity of a food crisis, particularly in terms of its magnitude.
The Household Group IPC classification can be used to get a sense of the scale of the food emergencies currently underway. The Famine Early Warning Systems Network (FEWS), for instance, publishes estimates for the number of people in need of emergency food assistance, defined as those experiencing, or imminently likely to experience Phase 3 (‘Crisis’) food insecurity or worse. This corresponds to households experiencing “food consumption gaps with high or above usual acute malnutrition” or those “marginally able to meet minimum food needs only with accelerated depletion of livelihood assets that will lead to food consumption gaps.”
The numbers estimated to be in need of emergency assistance in 2017, as defined by FEWS, did represent a peak in recent times45 – and humanitarian needs remained high in 2018.
It is these high estimated levels of emergency assistance need that led UN Emergency Relief Coordinator, Stephen O’Brien, to announce in 2017 that the world was facing the ‘largest humanitarian crisis since the creation of the UN’. This, however, does not imply an expectation that famine mortality would rise to the levels seen in the mid-20th Century. The broad developments that have reduced populations’ vulnerability to such severe famine mortality, discussed here, make this unlikely.
Peak population in need of emergency food assistance (IPC Phase 3+), 2015-2018 – FEWS46
The IPC system is fundamentally geared towards preventing famines, rather than assessing their severity after the event. As stated in the IPC Manual,47
“The purpose of the IPC… is not to classify various degrees of famine, nor is it to categorize the “worst famine”. Rather, in order to inform real-time decision-making, the IPC thresholds for famine… are set to signify the beginning of famine stages.”
It is important to bear this in mind when trying to compare such assessments with famine trends over time. Our table of famine mortality since 1860, provides estimates of the ‘excess mortality’ associated to individual famines.48
In reference to the discussion above, this can be thought of as a measure of magnitude only along one dimension: mortality.
Official famine declarations based on the IPC Area classification, like that made for South Sudan in 2017, do not straightforwardly map on to such an analysis. For instance, given the larger population being affected, it is quite possible that more people have died due to food consumption deficits since early 2017 in Yemen than in South Sudan, despite the intensity of the former crisis not having brought about a famine declaration in any part of the country so far.49
It’s no good blaming climate change or food shortages or political corruption. Sorry to be neo-Malthusian about it, but continuing population growth in this region makes periodic famine unavoidable… Many of the children saved by the money raised over the next few weeks will inevitably be back again in similar feeding centres with their own children in a few years time.
– Blog entry from British environmentalist, Sir Jonathan Porrit, 11/07/201150
It is not uncommon to see arguments along the lines of this quote from Sir Jonathan Porritt, claiming that famines are ultimately caused by overpopulation. Porritt – former director of ‘Friends of the Earth’ and also former chairman of the UK Government’s Sustainable Development Commission – was talking about the 2011 famine in Somalia that went on to kill roughly 250,000 people.51 He seems certain that the rapid population growth witnessed in East Africa had made famine there ‘unavoidable’.
There is something compelling about this logic: a finite land area, with a limited ‘carrying capacity’, cannot continue to feed a growing population indefinitely. From such a perspective, the provision of humanitarian aid to famine-afflicted countries, however well intended, represents only a temporary fix. In this view it fails to address the fundamental issue: there simply being too many mouths to feed.
As mentioned in the quote, this suggestion is commonly associated with the name of Thomas Robert Malthus, the English political economist writing at the turn of the nineteenth century. Malthus is famous for the assertion that in the absence of ‘preventative checks’ to reduce birth rates, the natural tendency for populations to increase – being ‘so superior to the power of the earth to produce subsistence for man’ – ultimately results in ‘positive checks’ that increase the death rate. If all else fails to curb population, ‘gigantic inevitable famine stalks in the rear, and with one mighty blow levels the population with the food of the world’.52
But does the evidence support this idea? Here we look into the relationship between population growth and famine, as well as that between population growth and hunger more generally.
This chart compares the number of famine deaths per decade – based on our famine dataset – with the world population over the same period.
Looking at the world as whole, it is very difficult to square Malthus’ hypothesis with the simple but stark fact that, despite the world’s population increasing from less than one billion in 1800 to more than seven billion today, the number of people dying due to famine in recent decades is only a tiny fraction of that in previous eras.
We might naturally think that the explanation for this trend lies in increasing agricultural production. Indeed, food supply per person has consistently increased in recent decades, as we can see in the interactive line chart shown. The large increase in global population being met with an even greater increase in food supply (largely due to increases in yields per hectare).
However, looking at the issue in this way is too simple. As we discuss in our entry on Famines, insufficient aggregate food supply per person is just one factor that can bring about famine mortality. Contemporary famine scholarship tends to suggest that insufficient aggregate food supply is less important than one might think, and instead emphasises the role of public policy and violence: in most famines of the 20th and 21st centuries, conflict, political oppression, corruption, or gross economic mismanagement on the part of dictatorships or colonial regimes played a key role.53
The same also applies for the most acutely food-insecure countries today.54
It is also true of the 2011 famine in Somalia referred to above, in which food aid was greatly restricted, and in some cases diverted, by militant Islamist group al Shabaab and other armed opposition groups in the country.55
Famine scholar Stephen Devereux of the Institute of Development Studies, University of Sussex, summarizes the trajectory of famines over the 20th century as follows: “The achievement of a global capacity to guarantee food security was accompanied by a simultaneous expansion of the capacity of governments to inflict lethal policies, including genocidal policies often involving the extraction of food from the poor and denial of food to the starving.”56
Thus, all in all, the recent history of famine mortality does not fit the Malthusian narrative particularly well. Firstly, contrary to what Malthus predicted for rapidly increasing populations, food supply per person has – in all regions – increased as populations have grown. Secondly, famines have not become more, but less frequent. Thirdly, in the modern era the occurrence of major famine mortality, and its prevention, is something for which politics and policy seem the more salient triggers.
Famine victims by decade and world population, 1860s-2010s
Famines tend to be thought of as acute periods of crisis, and are in that sense to be distinguished from more chronic manifestations of hunger that may in some places represent ‘normal’ circumstances, despite being responsible for large numbers of deaths.57
Given the typically political nature of outbreaks of such famine crises, it may make more sense to look for an effect of population growth on the longer-term trends of hunger and malnutrition.
But again, at the global level, we know that population growth has been accompanied by a downward trend in hunger. As we discuss in our entry on Hunger and Undernourishment, in recent decades the proportion of undernourished people in the world has fallen, and, although more muted, this fall is also seen in the absolute number. The number of people dying globally due to insufficient calorie or protein intake has also fallen, from almost half a million in the 1990s to roughly 300,000 in the most recent data, as shown in the visualisation.
We can also look at the experiences of individual countries, rather than just at the global level. Do those countries with particularly high population growth rates find it harder to adequately feed its population?
In order to get some idea about this, we can compare countries’ Global Hunger Index (GHI) score with their population growth rates. GHI is a composite measure, out of 100, that combines four indicators: undernourishment, child wasting, child stunting, and child mortality.58
The first scoring was conducted in 1992, and was then repeated every eight years with the most recent being in carried out in 2017. The score is based on data collected in the years leading up to the scoring year, and as such reflect the hunger levels in this period rather than solely capturing conditions in the year itself. All the countries for which there was GHI data available between 1992 and 2017 are shown in the three charts.59 Crucially, this excludes a number of very food-insecure countries including the Democratic Republic of Congo, South Sudan and Somalia, which have also seen high levels of population growth.60 This should be borne in mind when interpreting the following results.
Of the countries for which we do have GHI data, it is clear that those with higher levels of hunger have also tended to have had higher population growth over the last 25 years (first chart).61
It is important to see though that among the countries for which we have GHI scores in both 1992 and 2017, the level of hunger went down in all but one – Iraq (second chart). Over the same period population went up in almost every case. Moreover, those countries that experienced higher levels of population growth in fact saw a bigger drop in their GHI score over this period.62
The countries that saw high population growth over this period started with higher levels of hunger in 1992. So what we are seeing here is that countries are converging towards lower levels of hunger: it fell quickest in countries with the highest levels of hunger (third chart).
So whilst countries that experience hunger do tend to have high levels of population growth, the idea that population growth necessarily leads to increased hunger is clearly mistaken: many countries with high population growth have recently managed to decrease levels of hunger substantially.
Environmental degradation, including climate change, does pose a threat to food security, and the growth of human populations has undoubtedly exacerbated many environmental pressures. However, this represents only one aspect of the complex explanation of why so many people suffer and die from undernourishment today, despite their being adequate food available for consumption globally.63
‘Malthusian’ explanations of famine and hunger thus fall short for the following reasons, the evidence for which we reviewed above:
- Per capita food supply has increased as populations have grown, largely due to increasing yields.
- Famine deaths have decreased, not increased, with population growth.
- Food scarcity has played a smaller role in famines than suggested by the Malthusian narrative. It ignores other factors like conflict, poverty, access to markets, healthcare systems, and political institutions.
- Population growth is high where hunger is high, but that does not mean that population growth makes hunger inevitable. On the contrary, we see that hunger has fallen fastest in countries with high population growth.
If we want to put an end to hunger, we need to understand the diverse causes that bring it about. Oversimplifications that mistakenly see hunger and famine as an inevitable consequence of population growth do not contribute to this end.
English political economist, Thomas Robert Malthus, writing at the turn of the 19th century, is famous for describing famine as “the last, the most dreadful resource of nature” which acts to “level the population with the food of the world” should other forces fail to lower birth rates or increase death rates.64
We can think of this ‘Malthusian’ proposition as containing two separate hypotheses: firstly, that unabated population growth ultimately leads to famine; secondly, that famine acts as a ‘check’ to population in this eventuality.
As we discuss here, recent trends in famine mortality, and hunger more generally, largely contradict the first hypothesis. Here we investigate the second, by considering the contribution of famines to long-run population trends. What role has famine played in shaping birth and mortality rates throughout history?
We begin by considering two examples of famines which, from a demographic point of view, differ enormously: the Chinese famine of 1959-61 and that in Ireland in the late 1840s.
The ‘Great Leap Forward’-famine in China from 1959-61 was the single largest famine in history in terms of absolute numbers of deaths. Excess mortality estimates vary hugely, but based on our midpoint estimates, it cost more than double the number of lives than any other famine. Our table of famine mortality estimates is available here. This chart shows the estimated changes to birth and death rates during and after this famine period, in addition to the effect of these changes on overall population size. Alongside a significant jump in death rates, there was also a large fall in births — a trend very typical of famines.65
However, this was immediately followed by a spike in birth rates in the years immediately following the famine, offsetting to a large degree its demographic effect. This was followed by a steady decline in birth rates throughout the 1960s and 70s, concurrent with domestic birth control policies, but also in line with many other rapidly developing countries.
The trend in mortality rates is similar: the peak generated by the crisis was followed by a continued decline that forms part of the common experience of countries as they develop. The key thing to note is that these secular shifts in births and deaths far outweigh the short-lived impact of the famine in determining the long-run trajectory of population growth in China. Despite causing an excess mortality of 2-5% of the total population, and a similar number of ‘lost births’, we can see from the lower panel in the chart below the famine had next to no discernible impact on population in the long run.
Long-term demographic impact of China’s Great Leap Forward Famine66
This picture contrasts somewhat with the developments following the Great Irish Famine of the late 1840s, as shown in the chart below. After a period of rapid growth that brought the population to over 8 million, a famine struck that was, in relation to the country’s population, far more severe than the ‘Great Leap Forward’-famine. It is thought to have reduced the population by about one quarter in its immediate effects: one million perishing and a further million emigrating. But one striking feature of Irish demographics is how the population then continued to shrink following the famine. By 1911 there were in Ireland about half as many people as in 1841. The population only began to grow again in the late 20th century. Is this then an example of a country that learnt from its ‘Malthusian lesson’?
This continued depopulation was partly due to low birth rates (which were considerably lower than in England and Scotland in the second half of the 19th century) and partly due to very high levels of outward migration, particularly to the US and Great Britain. Between 1851 and 1900, there were almost as many outward migrants as there were deaths in Ireland (4.18 million and 4.56 million, respectively).68 According to Cormac Ó Gráda, “during the decades between the Great Famine and World War One the probability of a young Irishman or Irishwoman not emigrating was less than one in two.“
As Ó Gráda argues, the only way a famine can have any real lasting demographic impact is if it “teaches” the population to alter marriage and family planning practices to reduce fertility rates.69
There is some evidence of changing behaviour in Ireland following the famine, including more people choosing to marry later or not all. However it is difficult to know if this is directly attributable to the famine, or if it instead reflects people’s responses to other changes taking place at the time, such as increasing life expectancy or increasing incomes. Furthermore, whilst total birth rates were low in the post-famine period, the number of children being born to married couples remained high, and the rate of natural increase was highest in those parts of the country worst hit by the famine, complicating any simple explanation along these lines. Similarly, whilst the famine itself clearly provided the impetus for mass emigration, high levels of outward migration began some decades before the famine and continued long afterwards in the context of a much-ameliorated standard of living. Thus, it seems likely that it was the promise of improved economic opportunities, rather than fear of famine which drove emigration between 1851 and 1900.70
Overall then, even in this seemingly paradigmatic ‘Malthusian’ example, whilst Ireland undoubtedly did suffer some lasting demographic impacts from the famine, subsequent economic and social developments unrelated to the famine explain the majority of the depopulation the country experienced in the decades following it.
Population of the Island of Ireland, 1500-200067
But what can we say about the impact of famines on long-term population trends more generally? To help answer this question, we look instead at the role of ‘population crises’ of all kinds – that is to say severe spikes in mortality in general, not just those due to famine. What impact have such crises played in shaping population trends, relative to other global developments?
Here are two charts showing the historic evolution of death rates in England and Wales, and in Norway. Data points mark the annual ‘Crude Death Rate’ (total deaths per 1,000 people) in each country, and a line plotting the 20-year moving average is shown in each case. You can see that the decline in average mortality rates in both countries was preceded by a reduction in the spread around the average i.e., the number and extent of ‘crises’ of high mortality.71
However, when such spikes were common, they in fact played a relatively small part in keeping average mortality rates as high as they were. Economic Historian, Robert Fogel, in considering the data for England concludes that “crisis mortality72 [including famine] accounted for less than 5 percent of total mortality in England prior to 1800 and the elimination of crisis mortality accounted for just 15 percent of the decline in total mortality between the eighteenth and nineteenth centuries.”73
You can picture this by imagining what would happen if you took the highest points in the charts above – representing the crises in mortality – and moved them downwards towards the average for that time. The line showing the 20-year moving average would indeed fall, but only by a small amount compared to the overall decline.
Historic crude death rates in England and Wales
Historic crude death rates in Norway
Falling death rates, and increasing life expectancy, are trends that took place first in early industrialising countries, but have been a common experience in all parts of the world as poverty has declined, and healthcare and nutrition has improved. But since the 1960s this has been outpaced by a fall in birth rates, such that overall the global population growth rate has been steadily falling, and is likely to approach zero towards the end of this century. The rapid growth in population witnessed since the early 20th century was due to the fall in death rates happening ahead of the fall in birth rates, generating a period of natural increase in between.
This is known as the ‘demographic transition‘: a shift from stable populations with high birth and death rates to stable populations with low birth and death rates, with a period of rapid increase in between due to the fall in mortality preceding the fall in fertility. This chart shows this transition as it occurred in five very different countries. Here also we can see that the secular decline in death rates follow a reduction in its volatility. But again, the height of the peaks in earlier decades are generally small relative to the overall decline. Even larger crises, such as the “Great Leap Forward”, or the spike in mortality in Mauritius in the late 1920s,74 translate into very small changes in overall population trends, if at all. It is the persistent long-term decline in ‘normal’ birth and death rates that represent the more significant development for population trends, not the absence of temporary mortality spikes.
Demographic transition in 5 countries, 1820-201075
As with any living organism, humans cannot sustain a given population without sufficient energy resources. Given this, at first glance, it does seem intuitive to assume population growth and famines to be closely linked via food availability. The evidence discussed here (and also here) contradicts any simplified view of this relationship that fails to acknowledge the diverse causes of famines and population dynamics.
The analogy to other living organisms can obscure what is different about the human species. We organize ourselves into complex social and political structures capable of incredible joint accomplishments – such as the eradication of diseases. But we are also capable of inflicting, or consciously allowing, unimaginable suffering – including the majority of famine deaths to date. The capacity of the planet to feed us is not a fixed constant, imposed by nature, but rather it depends also on us: on our agricultural practices, the development and transmission of knowledge and technology, and also crucially upon our choice of diet – an inherently cultural act.
The population growth rate is now declining, not, thankfully, due to more frequent crises of mortality but because people, through their own volition, are choosing to have fewer children. This change is very much associated to rising incomes and other social developments in health and education, and has tended to happen more quickly in countries that have developed more recently. If we need any generalization here, it is plenty – in terms of improved access to adequate food, clean water, sanitation, healthcare, education and so on – rather than scarcity, that is slowing down our species’ multiplication.
This isn’t to say that increased populations and affluence haven’t brought about environmental damage, nor that environmental degradation poses no risks to our future well-being. But the idea we are helpless to stop famines in the face of high population growth in some parts of the world, or that famines represents any kind of ‘solution’ to the environmental problems humans are causing, are two hypotheses that do little to help either humanity or the planet.
This entry is based on our ‘Our World in Data-Dataset of Famines’ which covers the period since the mid-19th century and which can be found at the end of this document. In the following we discuss how famines are defined and in particular our reasoning for how we constructed the dataset.
Today the Integrated Food Security Phase Classification (IPC) provides a definition of 5 levels of food insecurity of increasing severity, with level 5 constituting famine. According to the IPC, in order for a food insecurity situation to be declared a famine it must meet three specific criteria:
- At least one in five households faces an extreme lack of food
- More than 30 per cent of the population is suffering from acute malnutrition (wasting)
- At least two people out of every 10,000 are dying each day
Whilst providing a more objective, and hence de-politicized, benchmark for declaring a famine – vital for eliciting a timely humanitarian response – a key aspect of the IPC classification is to provide a graduated system that fits the reality of food crises better than a binary ‘famine or no-famine’ approach. Notice that it focuses on the intensity of the crisis. As discussed by Howe and Devereux (2004), this is distinct from the magnitude of the event, typically understood in terms of the total (excess) mortality that occurred.76 In compiling our table of famine deaths over time, we have naturally used estimates of the latter. It is important to note that there is no institutionally-agreed classification of famines in terms of magnitude. Indeed, for some people, a crisis that resulted in no excess mortality might still be properly considered a famine under some circumstances – there are many terrible outcomes that a severe food crisis can produce other than mortality, such as losses of livelihoods or long-term health impacts. Nevertheless, in producing our table we decided to implement a lower threshold of mortality for a crisis to be included (see Famines with low mortality, below).
It is worth seeing that these two dimensions – intensity and magnitude – whilst clearly related are nevertheless independent of each other. A very high-intensity famine, resulting in high mortality rates, may only effect a very small group of people and thus represent a relatively low-magnitude event. Or a low-intensity crisis may extend across a wide area and over a long period, resulting in a high-magnitude famine. A threshold in terms of intensity (i.e. IPC level 5) does not, therefore, straightforwardly map onto any given magnitude threshold.
Our visualisations show data relating to the period up to and including 2016. No estimates of excess mortality for the major food emergencies currently affecting Yemen, South Sudan, Somalia, Nigeria and Ethiopia have to our knowledge been released. However, in each case, there was a crisis-level food insecurity in 2016 present also. It is therefore possible that as any such estimates emerge, some excess mortality will be seen as having occurred in 2016. This is particularly pertinent to the case of South Sudan, an area of which was officially declared as being in famine throughout early 2017 according to the IPC system. We will update our table accordingly as more clear information becomes available.
In constructing our table of famine mortality over time, we have relied on a variety of secondary sources (listed below), themselves generated from historical accounts that did not make use of such precise definitions, nor would they have been able to do so given the absence of demographic records. Even if we may imagine a relative degree of conformity through time to the notion that famine consists of ‘a widespread lack of food leading directly to excess mortality from starvation or hunger-induced illnesses’,77 it is important to note that most of the mortality estimates listed in our table are typically very rough and are often the subject of a good deal of controversy (some examples are discussed in more detail below).
It should be borne in mind that there may be many motives for different observers, record keepers or historians to (intentionally or otherwise) over- or underestimate mortality levels according to their sympathies with the affected population or suspected perpetrators.78
As noted by the World Peace Foundation, “generally speaking, better demographic calculations lead to lower estimations of excess deaths than those provided by journalists and other contemporary observers. We might therefore reasonably expect an upward bias in the figures for earlier famines on the record [i.e. from 1870s].”
The earlier back one goes though, the more one might suspect that the written historical record is incomplete. We start our table from the 1860s. Whilst records for this period are no doubt more complete than earlier times, it is likely that some smaller events that would count as famines by today’s standards may not have been recognized or recorded as such and are therefore missing from our table.
The aim of the table below is to show estimates of ‘excess mortality’ – that is to say, the extra number of deaths that occurred during the famine as compared to the number there would have been had the famine not occurred. As noted above, it should be borne in mind that those dying of infectious diseases during famines are normally also included in this. As such, mortality estimates typically try to subtract the ‘normal’ death rate – that expected in the famine’s absence – from the actual total death rate during the famine. Estimating the latter is far from straightforward given the paucity of reliable demographic statistics typical of even recent famines. This also requires making assumptions about what the ‘normal’ death rate is, leaving even more room for disagreement (see discussion of the Democratic Republic of Congo famine below for an example).
Given this focus on excess mortality, some severe food insecurity situations involving high levels of mortality can nonetheless result in next to no excess mortality where the ‘normal’ death rate for the area in question is already very high. Niger in 2005, which many observers at the time considered to be a famine, is an example of this.79
It is worth noting as well that what is ‘normal’ changes through time. Modern definitions of famine include criteria for nutrition and mortality that would correspond to conditions typical or near-typical of non-crisis conditions in earlier periods for much of the world.80
It should be noted that, based on the work of Alex de Waal at the World Peace Foundation, included in our famine list is a number of events that are described as ‘episodes of mass intentional starvation’. Some of these events are not included in other lists of major famine events of the 20th century (notably some of them are missing from Stephen Devereux’s much-cited 2000 paper, Famine in the twentieth century).81 This suggests that some authors might consider these deaths to be attributable more to conflict and not reliably attributable to famine. Nonetheless, we decided to include these events in our table. Our reasoning here is that the excess mortality associated to many of the famines listed in Devereux (2000) would not have occurred in the absence of conflict, and many of them are not without similar controversy (see below for some more discussion). Thus any distinction between ‘famine’ and ‘episodes of mass intentional starvation’ seems to be a matter of degree, and as such there appeared no clear reason not to include the latter in our table.
In our table we have excluded crises where reported excess mortality was lower than 1,000. Our reasons for doing so were twofold. Firstly, in the context of very large margins of error for many of the famines in our table (with upper and lower estimates of excess mortality sometimes several millions apart), we felt that including events in which very few people are recorded as dying might give a misleading impression of the accuracy of the rest of the estimates in the table. Secondly, for many people, excess mortality (due to starvation or hunger-induced diseases) would normally be seen as an integral part of what it means for a crisis to constitute a famine.82
It should be noted that there might be good reasons to make use of a definition of famine that allows for zero or very low excess mortality.83 Nevertheless we felt some lower threshold would be appropriate, though the exact cut-off was a somewhat arbitrary choice.
One of our main data sources is the International Disaster Database, which lists mortality estimates for a range of disasters. In particular, it provides data on a number of smaller-scale events often not given in the main lists of major famines we have used. We considered those events listed as ‘Droughts’ as being famines, though we excluded any such disasters with a mortality estimate lower than 1,000 as per our threshold.
In addition, we also chose to omit two recent drought events listed in the database for China of 1,400 and 2,000 excess deaths in 1988 and 1991 respectively, having failed to find any corroborating cross-references to famine having occurred in these years.
Relatedly, some events often described as famines are not included in the table below where the reported excess mortality is considered to be in some sense ‘negligible’. As with shifting understandings of what the ‘normal’, non-crisis death rate consists of, no doubt this is a threshold that has changed considerably over time as demographic analysis of famines has become more precise and excess mortality a relatively rare event. Examples of potentially controversial omissions we have made along these lines include the Highland Potato Famine in Scotland (1846-56), the Bihar famine in India 1966-7 (discussed in more detail below) and Niger in 2005.
Various secondary sources that we have used to compile our table (listed in the Data Sources section below) themselves use some excess mortality cut-off, but one typically higher than our threshold of 1,000. This means that there may exist records of famines of a magnitude larger than 1,000 excess deaths that are not included in our table (if they did not appear in the International Disaster Database).84 However given the large-magnitude events in our table, this can only have had a very small effect on the overall trend outlined in our charts.
The International Disaster Database lists a drought in India in 1965 as killing 1.5 million people. The only food crisis around this time that we could find cross-references for was that in Bihar, more commonly cited as occurring in 1966-67. Official statistics, however, suggest very low excess mortality. Indeed, the famine was sometimes invoked as evidencing that independent India had turned a corner in its development, such that it could now cope with a serious drought without sustaining major loss of life.
Dyson and Maharatna (1992), however, regarded the official mortality data to be highly deficient. They concluded that ‘while the available data show little sign of excess mortality in Bihar, we probably cannot exclude this possibility’.85 Drèze (1990) similarly came to the conclusion that ‘there is precious little evidence to support the self-congratulatory statements that have commonly been made about the Bihar famine, e.g. “no exceptional mortality was recorded” or “no one died of starvation”.’86
For our table we decided to exclude this famine given such uncertainty. Notably, we chose to excluded the EM-DAT figure for 1965: such a high mortality seems questionable given the absence of other corroborating references.
Similar issues surrounded the determination of an excess mortality figure for the Maharashtra crisis in 1972-3. For Drèze (1990) it is clear that, whilst “the crisis was of extreme severity, famine was uncontroversially averted“. This was largely due to an enormous public employment programme which at its peak employed as many as 5 million people in Maharashtra state alone.
Whilst Drèze considers the available demographic statistics to imply that “mortality rose only marginally, if at all,”, and notes that that there were no confirmed instances of ‘starvation deaths’, Dyson and Maharatna (1992) insist that the mortality rates do imply a significant excess mortality of 130,000. They arrive at this conclusion based on adjusting the figures to account for systematic under-registration of deaths, the pre-crisis trend in mortality rates, inter-census population growth and the possibility of excess mortality also occurring in 1972.
In keeping with many other of our listed famine mortality estimates, we decided to provide that figure cited by Devereux (2000), itself quoting the 130,000 figure from Dyson’s work.87
It is worth seeing though that our choice to attribute a mortality figure to the Maharashtra drought, but not that of Bihar, stands in contrast to the conclusion of Drèze (1990) – based on consideration of nutrition surveys, asset disposals and land sales (signs of acute distress), and the extent of migration – that the Bihar famine struck considerably harder. Nevertheless, in the absence of a specific mortality estimate for the Bihar famine it has been excluded from our list of famines. In any case, the level of uncertainty surrounding both of these famines should be borne in mind.
By far the largest single event in our table is that of China at the turn of the 1960s associated with the economic and social campaign led by Mao Zedong known as the Great Leap Forward. During and immediately after the Chinese famine, however, it remained ‘shrouded in mystery’, with the Chinese authorities and some Western observers insisting that, despite successive poor harvests, famine had been averted. In the post-Mao era of the early 1980s, some official demographic data was newly released allowing for the first systematic investigations of the death toll.
Initial results from this suggested an excess mortality of around 30 million, and this figure gained some currency. Subsequent estimates have tended to be lower.
One of the key issues is how these official data compare with UN estimates that exist for infant mortality and life expectancy for the period 1950-5, which imply significant under-registration in official data. Exactly what assumptions are made about such under-registration have consequences for the ultimate mortality estimate produced. There is necessarily a degree of arbitrariness to such assumptions, with different hypotheses often standing in contradiction to alternative sources of evidence such as historical documentation and conflicting with the demographic patterns typically observed in famines.
Whilst there is much uncertainty about the exact number of deaths attributable to the Great Leap Forward famine, it seems certain that it represents the single biggest famine event in history in absolute terms. Relative to the size of the population however, the death rate was “modest” compared to that of Ireland in the 1840s or Finland in 1867-8, and was comparable to that of the 1876-9 famine in China.
For our table we use the midpoint between the lowest and highest estimates given in our main sources, 15 million being the lower bound given by Ó Gráda (2009) and 33 million being the upper bound given by Devereux (2000).
A good summary of these issues is given by Ó Gráda (2008).
The most commonly-cited excess mortality estimate for the conflict in the Democratic Republic of Congo (DRC) is the 5.4 million given in a 2007 report by the International Rescue Committee (IRC – report available here). The estimates were based on ‘retrospective mortality surveys’ in which interviewers asked a sample of respondents to report the number of deaths that had occurred within their household over a given period. These were then used to make inferences about the number of deaths across the country and, in conjunction with an assumed baseline mortality rate capturing the number of people that would have died anyway in the absence of the conflict, were used to generate the overall excess mortality figure.
Some controversy was generated in 2009 with the publication of the 2009/10 Human Security Report which presented a number of criticisms of the IRC methodology and argued that it had significantly overestimated the death toll.88 The key debate concerned the baseline mortality rate used, which the Human Security Report considered to be too low, thereby inflating in its view the number of deaths that could be associated to the conflict. In addition, the Report argued that the samples of respondents used in the earliest IRC surveys were unrepresentative and also too small to provide reliable estimates. In particular, it suggested that the areas visited were atypical in that many of them were selected because of there being existing or planned humanitarian operations already in the vicinity, so they were therefore likely to have higher mortality rates than the average location.
Contrariwise, the IRC authors point to the fact that access to some of the most insecure zones was impossible during the surveys, suggesting a sample bias in the opposite direction.
The overall argument of the Human Security Report is that the available data is not sufficient to form the basis for a credible excess mortality estimate, and any attempt to make one is very sensitive to the choice between a range of plausible alternatives and subject to a very wide margin of error. It does produce an estimate, but only for the period between 2001-7 for which the surveys conducted were more representative and numerous. The report’s ‘best estimate’ for excess mortality over this period is 863,000, compared to the 2,830,600 of the IRC for the same period. However, it points out that this is very sensitive to assumptions about whether the counterfactual baseline mortality rate should be considered to have a trend. For short-lived events a point estimate for the baseline mortality rate is sufficient. To estimate the excess mortality of a long-lived event, the report argues, one should allow for the possibility that the baseline mortality rate would have changed over this period in the absence of the event being studied. In the case of DRC it might be reasonable to assume that a negative trend in mortality rates observed prior to the outbreak of war would have continued, in which case the Report’s ‘best estimate’ for the 2001-7 period would increase to 1.5 million.
As such, the 863,000 figure that we include as a lower bound in our table should be treated with extreme caution in that it completely excludes the period prior to 2001 and also ignores the downward pre-trend in mortality rates (as does the IRC estimate).
On the other hand, all these estimates of excess mortality include ‘violent deaths’ i.e. those directly attributable to conflict and not to the ensuing famine conditions. The IPC report cited does not provide an exact number of violent deaths, but it does claim that “less than 10 percent of all deaths were due to violence, with most attributed to easily preventable and treatable conditions such as malaria, diarrhea, pneumonia and malnutrition”. As such we do not attempt to subtract violent deaths from the total.
The number of people that died in the North Korean famine remains highly uncertain, largely due to the closed nature of the country which has precluded access to official data and other channels of inquiry, such as surveys. Estimates range from the North Korean Government’s ‘quasi-official’ estimate of 220,000 to the 3.5 million arrived at by South Korean NGO, Good Friends Centre for Peace, Human Rights and Refugees by extrapolation from interviews conducted with refugees fleeing the country. Whilst one might naturally be suspicious of the Government’s own estimate, the approximate figure has been leant some credence by a recent study by Spoorenberg and Schwekendiek (2012). Via a reconstruction of demographic trends between 1993 and 2008 census data, the authors deduce an estimated mortality between 240,000 and 420,000. A rough consensus seems to have emerged that the 3.5 million is not reliable: the sample of interviewees – people from areas so badly affected that they sought to emigrate – was almost certainly unrepresentative of the country as a whole.89
Over time, estimates made via a variety of methods have tended suggest increasingly lower excess mortality. For instance, Goodkind and West (2001) put forward 600,000-1 million, with a subsequent study by Goodkind, West and Johnson (2011) suggesting a mortality towards the lower end of that range. Ho Il Moon in an article for VOX argues for a figure of 336,000, again based on reconstruction of intercensal demographics. We take as our lower bound the 240,000 from Spoorenberg and Schwekendiek (2012) and Goodkind, West and Johnson (2011)’s higher figure of 600,000 as our upper bound.90
Pierre van der Eng collates local and international newspaper reports of a series of localized famines that may have affected specific parts of Indonesia intermittently during this period, against a backdrop of more generalized and persistent malnutrition in much of the country (his paper is partly available here). As news reports, these figures are clearly not necessarily all that reliable and naturally focus on total numbers of deaths rather than excess mortality. Nevertheless, taken together they probably do point towards some excess famine mortality. Moreover, this was a period of significant repression of press freedoms in which the Government appears to have sought to actively restrict reporting on food crises, such that the reports collated may only represent a subset of famine events that occurred.
In our table we include a zero lower bound and use van der Eng’s total figure of 135,400 deaths as the upper bound, taking the midpoint of these two for inclusion in the graphs presented in this entry.
Most of the visualisations in this entry are based on the ‘Our World in Data-Dataset of Famines’ assembled by us.
Our dataset is based on four main sources:
- Data: Information on the occurrence of natural and technological disasters.
- Geographical coverage: Global
- Time span: 1900-present
- Available at: Online at www.emdat.be
- We have used information on events listed as ‘Droughts’ with mortality estimates above 1,000 (see ‘Famines with low mortality’ above). Previously, there was a distinction in the database between droughts and famines, but these have been subsequently merged.
- This resource is published by the World Health Organization Collaborating Centre for Research on the Epidemiology of Disasters (CRED). EM-DAT states that the database “contains essential core data on the occurrence and effects of over 18,000 mass disasters in the world from 1900 to present. The database is compiled from various sources, including UN agencies, non-governmental organisations, insurance companies, research institutes and press agencies.”
- Neumayer and Plümper (2007) give an overview of natural disasters covered by EM-DAT.91
- The EM-DAT data for the time post 1970 is also available through Gapminder here.
- Data: Famines with mortality above 100,000 deaths and episodes of mass intentional starvation with killed more than 10,000
- Geographical coverage: Global
- Time span: 1870-present
- Available at: the World Peace Foundation website here.
- Data: Overview of the history of famines and of famine scholarship, with a selection of famine mortality estimates in Table 1.1.
- Geographical coverage: Global
- Time span: Table 1.1 includes selected famines since 1693
- The book’s website is available here.
Additional sources used in assembling the table below are as follows:
Kumar and Raychaudhuri [Eds.] (1983) – The Cambridge economic history of India, Volume 2, Cambridge University Press, 1983
Ó Gráda (2007) – Making Famine History. Journal of Economic Literature. Vol. 45, No. 1 (Mar., 2007), pp. 5-38. Available online here.
Ó Gráda (2008) – “The ripple that drowns? Twentieth-century famines in china and India as economic history.” The Economic History Review 61:5-37.
Ó Gráda (2010) – Famines past, Famine’s future. UCD CENTRE FOR ECONOMIC RESEARCH WORKING PAPER SERIES, WP10/20
Coghlan, Benjamin, Pascal Ngoy, Flavien Mulumba. Colleen Hardy, Valerie Nkamgang Beno, Tony Stewart, Jennifer Lewis and Richard Brennan, 2007. Mortality in the Democratic Republic of Congo: An ongoing crisis. New York, International Rescue Committee.
Davis (2001) – Late Victorian Holocausts; Verso Books, ISBN 978-1-85984-739-8
Maharatna (1992) The Demography of Indian Famines: A Historical Perspective; doctoral thesis, LSE. Available online here.
van der Eng (2012) – All Lies? Famines in Sukarno’s Indonesia, 1950s-1960s; Crawford School of Public Policy. Available online here.
Human Security Report Project. Human Security Report 2009/2010: The Causes of Peace and the Shrinking Costs of War. New York: Oxford University Press, 2011.
Iliffe (1987) – The African Poor: A History. Cambridge: Cambridge University Press
Imperial Gazetteer of India vol. III (1907), The Indian Empire, Economic (Chapter X: Famine, pp. 475–502, Published under the authority of His Majesty’s Secretary of State for India in Council, Oxford at the Clarendon Press. Pp. xxx, 1 map, 552. Through Wikipedia here.
Shoko Okazaki (1986) – The Great Persian Famine of 1870-71. Bulletin of the School of Oriental and African Studies, University of London, Vol. 49, No. 1, In Honour of Ann K. S. Lambton (1986), pp. 183-192
Crowell and Oozevaseuk (2006) – The St. Lawrence Island Famine and Epidemic, 1878–80: A Yupik Narrative in Cultural and Historical Context. Online here.
‘Wikipedia List’ refers to the List of Famines here.
|Year||Country||Excess Mortality midpoint||Excess Mortality lower||Excess Mortality upper||Source|
|1846–52||Ireland||1,000,000||1,000,000||1,000,000||Ó Gráda (2007)|
|1860-1||India||2,000,000||2,000,000||2,000,000||Kumar and Raychaudhuri [Eds.] (1983)|
|1863-67||Cape Verde||30,000||30,000||30,000||Ó Gráda (2009), p. 22|
|1866-7||India||961,043||961,043||961,043||Kumar and Raychaudhuri [Eds.] (1983)|
|1868||Finland||100,000||100,000||100,000||Ó Gráda (2009) Table 1.1|
|1868-70||India||1,500,000||1,500,000||1,500,000||Imperial Gazetteer of India vol. III (1907),|
|1870–1871||Persia (now Iran)||1,000,000||500,000||1,500,000||Okazaki (1986)|
|1876–79||Brazil||750,000||500,000||1,000,000||WPF; Davis (2001)|
|1876–79||India||7,176,346||6,135,000||8,217,692||Maharatna (1992). See also Ó Gráda (2009) Table 1.1|
|1877–79||China||11,000,000||9,000,000||13,000,000||WPF; Ó Gráda (2009) Table 1.1|
|1878-1880||USA (St. Lawreence Island Alaska)||1,000||1,000||1,000||Crowell and Oozevaseuk (2006)|
|1888-9||India||150,000||150,000||150,000||Kumar and Raychaudhuri [Eds.] (1983)|
|1896-7||India||3,887,287||2,624,574||5,150,000||Kumar and Raychaudhuri [Eds.] (1983); Maharatna (1992)|
|1899-1901||India||2,699,790||1,000,000||4,399,579||WPF; Maharatna (1992)|
|1900-03||Cape Verde||15,500||11,000||20,000||EM-DAT; Ó Gráda (2009), p. 22|
|1903-06||Nigeria (Hausaland)||5,000||5,000||5,000||Devereux (2000)|
|1906-07||Tanzania (south)||118,750||37,500||200,000||Devereux (2000); WPF|
|1913-14||West Africa (Sahel)||125,000||125,000||125,000||Devereux (2000); WPF|
|1915-18||Greater Syria (including Lebanon)||350,000||350,000||350,000||WPF|
|1917-19||Persia (now Iran)||455,200||455,200||455,200||WPF|
|1917-19||East Africa||300,000||300,000||300,000||WPF; see also Devereux (2000).|
|1920-22||Cape Verde||24,500||24,000||25,000||EM-DAT; Ó Gráda (2009), p. 22|
|1920-21||China (Gansu, Shaanxi)||500,000||500,000||500,000||Devereux (2000)|
|1921-22||USSR||9,000,000||9,000,000||9,000,000||Devereux (2000); Ó Gráda (2009) Table 1.1|
|1927||China (northwest)||4,500,000||3,000,000||6,000,000||Devereux (2000); Ó Gráda (2009) Table 1.1|
|1929||China (Hunan)||2,000,000||2,000,000||2,000,000||Devereux (2000); Ó Gráda (2008)|
|1932-34||USSR (Ukraine)||5,650,000||3,300,000||8,000,000||Devereux (2000); WPF|
|1932-34||USSR (Russia, Kazakhstan)||1,500,000||1,500,000||1,500,000||WPF|
|1934, 1936-7||China (Sichuan)||5,000,000||5,000,000||5,000,000||WPF; Ó Gráda (2008)|
|1940-43||Cape Verde||20,000||20,000||20,000||EM-DAT; Ó Gráda (2009), p. 22|
|1941-50||Europe/USSR (collection of WWII-related events)||6,333,000||6,333,000||6,333,000||WPF|
|1941-45||East/Southeast Asia (collection of WWII-related events)||5,444,000||5,444,000||5,444,000||WPF|
|1943||China (Henan)||3,250,000||1,500,000||5,000,000||Devereux (2000); WPF. See also Ó Gráda (2008)|
|1943||India (Bengal)||2,550,000||2,100,000||3,000,000||Devereux (2000); WPF|
|1946-48||Cape Verde||30,000||30,000||30,000||EM-DAT; Ó Gráda (2009), p. 22|
|1946-47||USSR (Ukraine and Belorussia)||1,300,000||600,000||2,000,000||Devereux (2000); WPF|
|1957-58||Ethiopia (Tigray)||248,500||100,000||397,000||Devereux (2000); WPF|
|1959-61||China||24,000,000||15,000,000||33,000,000||Devereux (2000); Ó Gráda (2009) Table 1.1; WPF|
|1966||Ethiopia (Wallo)||52,500||45,000||60,000||Devereux (2000)|
|1962-68||Indonesia||67,700||-||135,400||van der Eng (2012)|
|1968-70||Nigeria (Biafra)||750,000||500,000||1,000,000||Devereux (2000); WPF; see Ó Gráda (2009) p.98|
|1969-74||West Africa (Sahel)||50,500||-||101,000||Devereux (2000); WPF|
|1972-73||India (Maharashtra)||130,000||130,000||130,000||Devereux (2000)|
|1972-75||Ethiopia (Wallo & Tigray)||350,000||200,000||500,000||Devereux (2000); WPF|
|1974||Bangladesh||1,000,000||500,000||1,500,000||Devereux (2000); Ó Gráda (2009) p. 98|
|1979||Cambodia||1,605,000||1,210,000||2,000,000||Devereux (2000); WPF; see Ó Gráda (2009) p.98|
|1980||Chad||3,000||3,000||3,000||EM-DAT; see also Iliffe (1987) p.253|
|1983-85||Ethiopia||795,000||590,000||1,000,000||Devereux (2000); see WPF and Ó Gráda (2009) p.98|
|1984-85||Sudan (Darfur, Kordofan)||245,000||240,000||250,000||Devereux (2000); WPF|
|1988||Sudan (south)||175,000||100,000||250,000||Devereux (2000); WPF|
|1991-93||Somalia||360,000||220,000||500,000||Devereux (2000), WPF|
|1995-99||North Korea||420,000||240,000||600,000||WPF; see Devereux (2000) and Ó Gráda (2009) p.94|
|1998||Sudan (Baht el Ghazal)||85,000||70,000||100,000||Devereux (2000); WPF|
|1998-2007||Democratic Republic of Congo||3,131,500||863,000||5,400,000||Coglan et al (2007); 2009/10 Human Security Report|