The statistical model identifies links between battles in Africa and can be applied to other armed conflicts, according to a study at the Complexity Science Hub 

Around the world, political violence increased by 27 percent last year, affecting 1.7 billion people. The numbers come from the Armed Conflict Location & Event Data Project (ACLED), which collects real-time data on conflict events worldwide.


Some armed conflicts occur between states, such as Russia’s invasion of Ukraine. There are, however, many more that take place within the borders of a single state. In Nigeria, violence, particularly from Boko Haram, has escalated in the past few years. In Somalia, populations remain at risk amidst conflict and attacks perpetrated by armed groups, particularly Al-Shabaab. 


To address the challenge of understanding how violent events spread, a team at the Complexity Science Hub (CSH) created a mathematical method that transforms raw data on armed conflicts into meaningful clusters by detecting causal links. 


“Our main question was: what is a conflict? How can we define it?,” says CSH scientist Niraj Kushwaha, one of the authors of the study published in the latest issue of PNAS Nexus. “It was important for us to find a quantitative and bias-free way to see if there were any correlations between different violent events, just by looking at the data.”


“We often tell multiple narratives about a single conflict, which depend on whether we zoom in on it as an example of local tension or zoom out from it and consider it as part of a geopolitical plot; these are not necessarily incompatible,” explains coauthor Eddie Lee, a postdoctoral fellow at CSH. “Our technique allows us to titrate between them and fill out a multiscale portrait of conflict.”


In order to investigate the many scales of political violence, the researchers turned to physics and biophysics for inspiration. The approach they developed is inspired by studies of stress propagation in collapsing materials and of neural cascades in the brain.


Kushwaha and Lee used data on violent battles in Africa between 1997 and 2019 from ACLED. In their analysis, they divided the geographic area into a grid of cells and time into sequential slices. The authors predicted when and where new battles would emerge by analyzing the presence or absence of battles in each cell over time. 


“If there’s a link between two cells, it means a conflict at one location can predict a conflict at another location,” explains Kushwaha. “By using this causal network, we can cluster different conflict events.”

Snow and sandpile avalanches

Observing the dynamics of the clusters, the scientists found that armed clashes spread like avalanches. “In a way evocative of snow or sandpile avalanches, a conflict originates in one place and cascades from there. There is a similar cascading effect in armed conflicts,” explains Kushwaha.


The team also identified a “mesoscale” for political violence —a time scale of a few days to months and a spatial scale of tens to hundreds of kilometers. Violence seems to propagate on these scales, according to Kushwaha and Lee.


Additionally, they found that their conflict statistics matched those from field studies such as in Eastern Nigeria, Somalia, and Sierra Leone. “We connected Fulani militia violence with Boko Haram battles in Nigeria, suggesting that these conflicts are related to one another,” details Kushwaha. The Fulani are an ethnic group living mainly in the Sahel and West Africa. 


Policymakers and international agencies could benefit from the approach, according to the authors. The model could help uncover unseen causal links in violent conflicts. Additionally, it could one day help forecast the development of a war at an early stage. “By using this approach, policy decisions could be made more effectively, such as where resources should be allocated,” notes Kushwaha.


The study “Discovering the mesoscale for chains of conflict” by Niraj Kushwaha and Eddie Lee appeared in PNAS Nexus.

Early malnutrition increases the risk of type 2 diabetes and other diseases later in life. In a recent study, researchers at the Complexity Science Hub and Medical University of Vienna showed this to be true.

A number of studies have already shown that malnutrition during pregnancy may increase the risk of developing type 2 diabetes later on in life. According to a 2013 study by Peter Klimek and his team, people born during a famine have more than twice the risk of diabetes compared to those born one year earlier or later. 




Now, for the first time, Klimek and his team have succeeded in measuring not only the total number of diabetes cases (prevalence), but also the incidence, or the number of new cases, in a recent study. 


 “Among men born during the two most severe famine periods, 1939 and 1946/1947, the rate of new cases of diabetes is up to 78 percent higher in 2013 to 2017 than in comparable years, and up to 59 percent higher among women,” explains Klimek, from the Complexity Science Hub and the Medical University of Vienna. The effect is strongest in those born in 1939. 


The incidence rate rose from 3.9 percent to 6.9 percent among men and from 3.4 percent to 5.4 percent among women. Additionally, both groups have an increased incidence of concomitant conditions such as heart failure, arterial hypertension, chronic obstructive pulmonary disease (COPD), and kidney disease.




Scientists believe this is a result of genetic programming that occurs during pregnancy, which increases the risk of these diseases. As a result of deficiency, the unborn child’s metabolism adjusts to a nutritionally poor environment. If this does not prove true later in life, a maladaptation occurs that leads to increased metabolic and cardiovascular diseases in these birth groups.




“One strength of our study is the new, large dataset on which it is based,” says Klimek. This covers 99.9 percent of the Austrian population between 2012 and 2017, and all insured patients aged over 50 and under 100 were examined. Of these approximately 3.5 million people, 746,184 were treated for diabetes. The comprehensive dataset allowed researchers to measure age-specific and regional incidence rates directly for the entire population, without additional assumptions that would be required for modeling. 


“Our results clearly demonstrate that public health efforts to address diabetes should not focus solely on lifestyle factors. The importance of reproductive health, as well as adequate nutrition during pregnancy and in the early postnatal period, must also be considered,” Klimek said.





The study “Diabetes incidence in Austria: The role of famines on diabetes and related NCDs” was recently published in the journal Heliyon.

A new study out of the Complexity Science Hub concludes that social disintegration and violent conflict played a crucial role in shaping the population dynamics of early farming societies in Neolithic Europe

Complexity scientist Peter Turchin and his team at CSH, working as part of an international and interdisciplinary collaboration, may have added a meaningful piece to a long-standing puzzle in archeology. Scholars have long tried to understand why Neolithic farmer populations go through boom-bust cycles, including “collapses” when whole regions are abandoned. According to one common explanation, climate fluctuations are the main driver, but empirical tests do not fully support this claim. In a new paper, published in the latest issue of Nature Scientific Reports, Turchin and his team seem to have come up with a new piece of information.


“Our study shows that periodic outbreaks of warfare — and not climate fluctuations – can account for the observed boom-bust patterns in the data,” argues Turchin, who’s a project leader at the Complexity Science Hub (CSH).


The team tested the two rival theories attempting to explain these dynamics – climate change and social conflict – in computer simulations and compared the results with historical data.


“This is the first time an agent-based model has been applied to this scale for this period in history, pre-state and pre-empire. The model covers most of the European continent and works with small units, such as independent villages. Previous simulations for this period have been conducted by dividing the area into a few large regions, but we wanted to examine interactions at the village level,” explains CSH scientist Dániel Kondor, who’s the first author of the study.

Change of heart

Turchin has been applying mathematical models of social integration and disintegration to analyze the rise and fall of complex societies, such as agrarian empires in history or modern nation-states. He admits he wasn’t convinced that such ideas would also apply to prehistory, such as the European Neolithic, where most of the time people lived in small-scale farming communities with no deep social inequalities and limited political organization beyond local settlements.


“I confess that until recently I thought that such societies were quite resilient and not susceptible to social disintegration and collapse,” says Turchin. “There is no state or nobles to rebel against and, in any case, what’s there to ‘collapse’?,” adds the complexity scientist.


Turchin, however, now holds a different view. Increasing evidence suggested that “simple” Neolithic farmers’ societies also collapsed. “In fact, such cases are much more profound than the social and political breakdown of more recent societies, because archaeology indicates that substantial regions were depopulated.”

Computer simulations 

In the study, the researchers focused on the period from the first evidence of agriculture in Europe to the beginning of the Bronze Age – between 7000 BCE and 3000 BCE. The simulation begins with each small unit of the map either empty or occupied by a village of independent farmers. The simulation combines two components: population change in each unit based on climate variability during the time period; and interactions, which include populations in each unit splitting, migrating, or coming into conflict with each other.

The patterns created by the computer simulations were then compared to real-world data. A radiocarbon dating database was used by the team. “Archaeological data on settlements and Carbon-14 dating indicate boom and bust cycles. As settlement data are limited to a few regions and periods, we rely on Carbon-14 data in our model predictions,” explains Kondor.


Based on the study’s findings, climate variation is not capable of explaining boom-bust dynamics during the time period. In contrast, simulations taking into account the social conflict produced patterns similar to those observed in radiocarbon dating.


“Of course, we can’t prove that this is the only mechanism behind the population declines during that time period. There could be other [mechanisms], but we demonstrate that internal conflict producing population cycles is consistent with real-world data,” argues Kondor.

(c) CSH
In the top graph, simulated populations under climate variation and no inter-group conflict (in red) are compared to population distributions from observed radiocarbon data (in grey). There is no match. The bottom graph shows simulated populations under inter-group conflict and climate variation (in red), compared to population distribution from observed radiocarbon data (in grey), with a good match between the two.

Intense times

The study presumes a complex social landscape in this time period. Such a notion is consistent with the results of intense archaeological research in Europe over the past century. “This period was indeed much more dynamic than non-specialists might think,” adds Kondor.


“Since we don’t see consistent large-scale political organization during this time, it would be easy to imagine that things were static, such that people settled in a village and lived there for three or four thousand years without much happening in between. That doesn’t seem to be the case. Sadly, this also means that this period was more violent than previously thought.”


“Many case studies have shown that early farming societies underwent socio-political cyclical dynamics from integration to disintegration. These social cycles run more or less parallel to population cycles with outbreaks of considerable violent conflict occurring during the disintegration periods,” explains archaeologist Detlef Gronenborn, from the Leibniz Centre for Archaeology in Mainz, Germany, one of the coauthors of the study.


“With this supra-regional study, we were able to show that the previous can be applied to a much broader region and to a much longer time period. Disintegration and warfare seem to be a general behavioral pattern,” evaluates Gronenborn.


“Additionally, the study indicates that humans and their interactions, whether friendly or violent, form a complex system, regardless of their political or economic organization. It doesn’t matter if you don’t want to organize into a state, you are still affected by your neighbors and their neighbors as well,” adds Kondor.


The study “Explaining population booms and busts in Mid-Holocene Europe,” by Dániel Kondor, James S. Bennett, Detlef Gronenborn, Nicolas Antunes, Daniel Hoyer and Peter Turchin, was published in Nature Scientific Reports 13, 9310 (2023).

Social networks are designed to appeal to our emotions. Education, however, is not in vain, according to CSH scientist Hannah Metzler.

Hannah Metzler, a neuroscientist and psychologist at the Complexity Science Hub, recently contributed to a Stockholm Resilience Center report on climate misinformation. In a chapter, she discusses the role of emotions in spreading misinformation online about climate change.


People are not per se more susceptible to false news when they’re emotional, according to Metzler. “Strong emotions do not automatically mean people will believe a message and continue to share it. That people judge new information based on trust in sources, and their knowledge about the world, means that good arguments, education, and science communication are not futile.”


As Metzler points out, there are ways to reduce misinformation’s spread. Redesigning platforms would allow nuanced majorities and overlap in views of different groups to be more apparent. A redesign of social media algorithms might also be an option.


In her research, Metzler applies methods like text analysis and machine learning to capture digital traces of emotions, to investigate their validity, as well as their effects on misinformation spreading on social media.


Read her chapter in the report:

Being obese significantly increases the chances of also developing mental disorders.

This applies to all age groups, with women at higher risk than men for most diseases, a recent study of the Complexity Science Hub and the Medical University of Vienna shows.

“We analyzed a population-wide national registry of inpatient hospitalizations in Austria from 1997 to 2014 to assess the relative risks of comorbidities for obesity and statistically significant gender differences,” explains Elma Dervic of the Complexity Science Hub.


Consequently, it became evident that an obesity diagnosis significantly enhances the likelihood of a wide range of mental disorders across all age groups – including depression, nicotine addiction, psychosis, anxiety, eating and personality disorders. “From a clinical point of view, these results emphasise the need to raise awareness of psychiatric diagnoses in obese patients and, if necessary, to consult specialists at an early stage of diagnosis”, says Michael Leutner of the Medical University of Vienna.


“In order to find out which illness typically appeared prior and subsequently to the obesity diagnosis, we had to develop a new method. This allowed us to determine whether there are trends and typical patterns in the occurrence of diseases,” explains Dervic.

In case of all co-diagnoses, with the exception of the psychosis spectrum, obesity was in all likelihood the first diagnosis made prior to the manifestation of a psychiatric diagnosis. “Until now, physicians often considered psychopharmacological medications to cause the association between mental disorders and obesity as well as diabetes. This may be true for schizophrenia, where we see the opposite time order, but our data does not support this for depression or other psychiatric diagnoses,” explains Alexander Kautzky of Department of Psychiatry and Psychotherapy of the Medical University Vienna.

However, whether obesity directly affects mental health or whether early stages of psychiatric disorders are inadequately recognised is not yet known.


Surprisingly, the researchers found significant gender differences for most disorders – with women showing an increased risk for all disorders except schizophrenia and nicotine addiction.


While 16.66% of obese men also suffer from nicotine abuse disorder, this is only the case in up to 8.58% of obese women. The opposite is true for depression. The rate of diagnosed depressive episodes was almost three times higher in obese women (13.3% obese; 4.8% non-obese). Obese men were twice as likely to be affected (6.61% obese; 3.21% non-obese).

New study: obesity increases risk of mental disorders - with women at higher risk than men
Gender gaps towards male overrepresentation are seen in schizophrenia and nicotine use disorders, while all other comorbidities were showed increased risk in females. Gender gaps further widened in the presence of a diagnosis of obesity @ Complexity Science Hub


At present, obesity is a highly prevalent disease worldwide and affects more than 670 million people. The fact that the disease promotes metabolic disorders and serious cardio-metabolic complications (diabetes mellitus, arterial hypertension, and dyslipidaemia) has already been extensively researched.


Since this study now also shows that obesity often precedes severe mental disorders, the findings underscore its importance as a pleiotropic risk factor for health problems of all kinds. This is primarily true for young age groups, where the risk is most pronounced. For this reason, thorough screening for mental health problems in obese patients is urgently needed to facilitate prevention or ensure that appropriate treatment can be given, so the researchers conclude.


The study “Obesity as pleiotropic risk state for metabolic and mental health throughout life” has been published in Translational Psychiatry.

Would you like to see more women on corporate boards and in leadership positions in academia? A new CSH study recommends aligning quotas with an inclusive culture

What is the impact of affirmative actions, such as quota systems, on minorities’ representation in top ranks of the academic and corporate worlds? Scientists used mathematical models for the first time to quantify how successful quota systems can be for improving women’s visibility in science.


Their findings were published in the latest edition of the journal Communications Physics.


“The main question we wanted to answer was: the number of women has increased in academia and in the corporate sector over the past 100 years, but why do they not reach the top positions in their network?,” points out CSH scientist Fariba Karimi, who’s a co-author of the paper.

(c) Shutterstock
A multidimensional approach is needed to improve minorities' visibility, according to the study

According to the study, quotas alone are not sufficient to make minorities more visible in a network. “In essence, the results show that having even extreme quotas does not necessarily ensure that minorities will be represented in top ranks of the network as we would expect from their size,” says Karimi.


“In contrast, a very moderate quota would be extremely useful when it is combined with an inclusive environment in which people, especially those in high power positions, are open to bringing minorities into their personal networks,” adds Karimi. “By doing so, they are basically helping minorities grow their social capital through those connections.”

Hypothetical scenarios

In the study, the researchers created a network growth model to analyse how successful interventions can be for improving minorities’s visibility in social networks. Two kinds of interventions were tested: group size interventions, such as quotas; and behavioral interventions, such as changing the way groups interact.


“We ran this two hypothetical scenarios, sometimes isolated, sometimes combined, as we wanted to evaluate which combination of interventions would be more effective in pushing minorities to the top of the ranking,” explains Karimi.


The model took into account two key social processes. First, the formation of structural inequalities that emerge within social networks due to certain preexisting societal biases, such as in-group favoritism or homophily – the notion that humans tend to preferentially interact and connect with individuals who are like them in some way. Second, the impact of different interventions on changing those initial structural inequalities.


The results show, for instance, that even a very strong group size intervention – to have a 90 percent quota – will not improve minorities’ representation in the top ranks to a level proportional to their total size if the initial configuration is strongly homophilic. As a result of historical and cumulative structural inequality, minorities are locked in their initial network position.

(c) CSH
Group size and behavioural interventions were investigated for their impact on minority nodes' position (red) in degree rankings over time.


“The study shows that the discussion [of improving minorities’ visibility] should not be one-dimensional”, highlights Leonie Neuhäuser, from RWTH Aachen University and co-autor of the study.


From a network perspective, increasing the size of a group does not necessarily increase the visibility of minorities, according to Neuhäuser. “Obviously, this is a necessary step, but we should also consider the social network structure and behavioural aspects when designing interventions.”


The results indicate some behavioural interventions that may affect minority representation in top ranks. A minority group could benefit from increasing networking if they are large enough to gain a cumulative advantage in a growing social network. Alternatively, if quotas are not large enough, the majority group should be encouraged to mix with the minority group, since the latter will not gain visibility without connecting to the former.

Hard to change

Despite the difficulty of changing behavior, Karimi emphasizes that increasing diversity depends on it. Leaders and top-level professionals can be educated about this issue and be inclusive when bringing people with diverse backgrounds into important positions on social networks, according to Karimi.


“We also need some regulation,” adds Karimi. “As humans, we have a tendency to prefer interacting with people who are similar to ourselves, since it is less cognitively demanding – the homophily principle. Evolutionarly speaking, we are rewired to avoid interacting with outgroup members and that is why incentivising and educating people about the benefits of diversity can help overcome those barriers.”


Due to the external regulations, after a certain number of minority people are incorporated into the system, things begin to change. Having more people of diverse backgrounds around us reduces our fear of strangeness. As a result, behavior would follow.”

The study Improving the visibility of minorities through network growth interventions by Leonie Neuhäuser, Fariba Karimi, CSH researcher Jan Bachmann, CSH external faculty member Markus Strohmaier and Michael T. Schaub appeared in Communications Physics 6, 108 (2023).



:: A team from the Complexity Science Hub and Central European University developed poverty maps of Sierra Leone and Uganda which visually identify poor areas with greater accuracy


:: Policymakers can use the new maps to allocate resources more effectively and make better policy decisions


:: An online interactive tool was created by the researchers where users can compare wealth inferred in both countries


:: Check our short video and learn how to navigate the tool:

A team of researchers from the Complexity Science Hub (CSH) and Central European University (CEU) created more-detailed poverty maps using computational tools that bring together survey information, and data and images provided by public sources such as Google and Meta (Facebook).


The poverty maps for Sierra Leone and Uganda, two Sub-Saharan African countries with extreme poverty, will be presented in early May at the Web Conference (ACM) in Texas, US.


For years, policymakers, planners, and researchers have relied on surveys and census data to track and respond to poverty. “Gathering this information, however, can be time-consuming and expensive,” explains Lisette Espín-Noboa, a postdoc at CSH and CEU and first author of the study. “And a census may not include hard-to-reach areas.”


“If you ask why we need high-resolution poverty maps, the answer is simple. These incredibly detailed poverty maps can identify places in need that otherwise wouldn’t be captured in maps using only census data,” adds CEU associate professor Márton Karsai, co-author of the study.


“In regions with a lack of direct information, it is challenging to estimate the socio-economic status of people. In our study we focused on two countries, Uganda and Sierra Leone, where high-resolution, large-scale estimates are not available,” points out professor János Kertész, head of the Department of Network and Data Science at CEU and co-author of the study.

(c) CSH


In the first step, the team gathered demographic data from two surveys. “In particular, we focused on the housing characteristics questionnaire. This helps estimate household wealth by considering the quality and quantity of available facilities or assets at home,” say the researchers.


For both countries, International Wealth Index (IWI) scores were calculated. IWI indicates how well a home possesses a basic set of assets. The lower the value, the lower the quality of housing. And the higher the value, the wealthier the household.


Second, Espín-Noboa and colleagues compiled a second dataset of satellite imagery and geographic information system data provided by crowdsourcing and social media sources, like Google and Meta (Facebook).


Over 900 features were extracted that offer insight into the economic status of the population, the infrastructure development of the area, and other wealth indicators. “This data shows, for instance, how many antennas are located in a certain area, or the number of people visiting the same area. It can also indicate how many Facebook users own an iPhone,” explains Espín-Noboa.


The team created three machine learning models that were trained to determine not only a place’s average wealth, but also its standard deviation. Ultimately, the goal is to provide a more accurate picture of the wealth distribution within each populated area. “We wanted to know how wealth varied within an area, or if there was inequality,” points out Espín-Noboa.


The models were trained to learn correlations between the demographic data and IWI scores and the features extracted from data and images provided by public sources. “They learn, for instance, that a specific wealth value correlates to a specific set of features,” says Espín-Noboa. “Next, we tested the models by asking them to predict the wealth of different areas.”


The team developed three models: one based on satellite imagery; a second on crowdsourcing and social media data; and a third on combining the two. “We wanted to determine which features contributed most to wealth prediction,” explains Karsai.


Using the models, the researchers created maps detailing the poverty levels of communities in Sierra Leone and Uganda. Together with Liuhuaying Yang, they also created an interactive map tool where users can compare wealth inferred in both countries and explore the machine learning models.


“The performance of our models is comparable to the state-of-the-art and, in some cases, outperforms it,” ponders Karsai. “Importantly, we estimate not only the economic status of people living at a given location but also local fluctuations. Moreover, we find that different methods are needed for estimating high income (where satellite images provide good results) and poor regions (where metadata-based methods are the best).”

Covid-19 pandemic

For three decades, the World Bank has documented a decline in the number of people living in extreme poverty, defined as living on less than $2.15 a day. This trend was disrupted by the Covid-19 pandemic in 2020. The number of people living in extreme poverty has increased by 70 million to more than 700 million.


According to the World Bank, the recovery has been uneven since then. And most of the people still living in extreme poverty are in Sub-Saharan Africa, in conflict-affected areas, and in rural areas.


Consequently, more accurate, timelier, and cost-effective ways of measuring poverty are urgently needed, points out Kertész. “By combining traditional household survey data with nontraditional data sources, we can map poverty at a higher resolution and scale. This can help policymakers make better evidence-based decisions when designing programs that reduce poverty and inequality,” says Kertész, who’s also an external faculty member at CSH.

Using real data from an online chess platform, scientists of the Complexity Science Hub and the Centro Ricerche Enrico Fermi (CREF) studied similarities of different chess openings. Based on these similarities, they developed a new classification method which can complement the standard classification.



  • It is possible to checkmate an opponent in chess in two moves.
  • Mathematically, there are more possible chess games than there are atoms in the observable universe. This is the Shannon number, which represents all possible move variations in chess. It is estimated there are between 10111 and 10123 positions (including illegal moves) in chess.

“To find out how similar chess openings actually are to each other – meaning in real game behavior – we drew on the wisdom of the crowd,” Giordano De Marzo of the Complexity Science Hub and the Centro Ricerche Enrico Fermi (CREF) explains. The researchers analyzed 3,746,135 chess games, 18,253 players and 988 different openings from the chess platform Lichess and observed who plays which opening games.


If several players choose two specific opening games over and over again, it stands to reason that they will be similar. Opening games that are so popular that they occur together with most others were excluded. “We also only included players in our analyses that had a rating above 2,000 on the platform Lichess. Total novices could randomly play any opening games, which would skew our analyses,” explains Vito D.P. Servedio of the Complexity Science Hub.

Researchers at the Complexity Science Hub find new classification of chess openings


In this way, the researchers found that certain opening games group together. Ten different clusters clearly stood out according to actual similarities in playing behavior. “And these clusters don’t necessarily coincide with the common classification of chess openings,” says De Marzo.


For example, certain opening games from different classes were played repeatedly by the same players. Therefore, although these strategies are classified in different classes, they must have some similarity. So, they are all in the same cluster. Each cluster thus represents a certain style of play – for example, rather defensive or very offensive. Moreover, the method of classification that the researchers have developed here can be applied not only to chess, but to similar games such as Go or Stratego.



  • Currently Magnus Carlsen from Norway (born in 1990) is ranked as the best chess player worldwide (rating: 2853; according to FIDE).
  • According to the Federation Internationale des Echecs (FIDE) there are currently 1785 grandmasters worldwide (ten of them in Austria)
  • Of course, nobody knows how many people play chess worldwide. To give an approximate impression: FIDE currently has 160,684 active players registered. Estimates say that around 600 million adults worldwide play chess regularly.


The opening phase in chess is usually less than 20 moves. Depending on which pieces are moved first, one speaks of an open, half-open, closed or irregular opening. The standard classification, the so-called ECO Code (Encyclopaedia of Chess Openings), divides them into five main groups: A, B, C, D and E.


“Since this has evolved historically, it contains very useful information. Our clustering represents a new order that is close to the used one and can add to it by showing players how similar openings actually are to each other,” Servedio explains. After all, something that grows historically cannot be reordered from scratch. “You can’t say A20 now becomes B3. That would be like trying to exchange words in a language,” adds De Marzo.

July 20 is World Chess Day. It marks the date of the founding of the International Chess Federation (FIDE) in Paris in 1924.


In addition, their method also allowed the researchers to determine how good a player and how difficult a particular opening game is. The basic assumption: if a particular opening game is played by many people, it is likely to be rather easy. So, they examined which opening games were played the most and who played them. This gave the researchers a measure of how difficult an opening game is (= complexity) and a measure of how good a player is (= fitness). Matching these with the players’ rating on the chess platform itself showed a significant correlation.


“On the one hand, this underlines the significance of our two newly introduced measures, but also the accuracy of our analysis,” explains Servedio. To ensure the relevance and validity of these results from a chess theory perspective, the researchers sought the expertise of a renowned chess grandmaster who wishes to remain anonymous.


The study “Quantifying the complexity and similarity of chess openings using online chess community data” has been published in Scientific Reports.

Global supply shortages of antibiotics are currently on the rise. A study by the newly founded Supply Chain Intelligence Institute Austria (ASCII) in collaboration with the Complexity Science Hub now deciphers the causes of the shortage: First and foremost, the increasing concentration of production in two countries – China and India. Investing in a well-developed data infrastructure can help remedy the situation.

In the interactive visualization, created by Liuhuaying Yang of the Complexity Science Hub, you can explore the connections and interdependencies for yourself. This video shows how it works.

“Our data shows a clear trend of increasing concentration of production on a few countries, namely China and India,” explains Peter Klimek of the Complexity Science Hub and the Supply Chain Intelligence Institute Austria (ASCII). A trend that accelerated even further during the pandemic.


These dependencies are not apparent in direct imports but become clear as one traces the value chain further back. Intermediaries and active pharmaceutical ingredients (APIs) are more concentrated than unpackaged and packaged products. 76% of intermediaries manufacturing sites and 59% of API producers are located in China and India. “If a shortage occurs here, potential substitute products are also more likely to be affected by shortages,” Klimek said. In line with this observation, the estimated number of shortages that could be addressed through substitution has halved from 2018 to 2020.


The measures taken to combat the SARS-CoV-2 pandemic also reduced the circulation of other pathogens. As a result, both demand and consumption of antibiotics in the general population and in hospitals fell dramatically – by about 20% – during the pandemic. In 2022, when many countries stopped taking measures to contain SARS-CoV-2, antibiotic consumption increased accordingly. Combined with geographically concentrated production systems, shortages occurred in large parts of the world.


One of the few European production facilities for antibiotics is located in Kundl, Austria. Despite this (according to the Austrian Federal Office for Safety in Health Care, as of March 9, 2023), several products are on the list of antibiotics not available in Austria. “Our data show that Austria is most closely linked to Spain in terms of supply chains,” Klimek said. However, due to indirect dependencies, China is right behind it in second place, he added.


To better quantify and predict drug shortages in the future, both short-term and long-term measures are needed. “One of the most important things will be to invest in data, planning and forecasting infrastructure to measure, understand and forecast the demand for antibiotics,” Klimek explains. The focus should be on those drugs for which substitutes are also lacking.


Furthermore, greater emphasis should be placed on security of supply and appropriate measures taken, he adds. After all, a well-designed market should internalize the risk of disruption. “Today’s antibiotics market structures are highly internationalised. This ensures low prices, which is desirable. Unfortunately, security of supply has repeatedly proved to be a weak point. Consideration should be given to a market design in which health policy does not need to intervene and which still guarantees security of supply”, says co-author Klaus Friesenbichler (ASCII and WIFO). Providing market incentives for the development of new antibiotics will be particularly important as resistance to existing products increases. A failure to develop new and improved antibiotics therefore poses a risk to society.

Why does the world seem to suddenly be fighting over antibiotics?

A new study around Peter Klimek of the Complexity Science Hub and the ASCII provides insights.

In the interactive visualization, created by Liuhuaying Yang of the Complexity Science Hub, you can explore the connections and interdependencies for yourself. This video shows how it works.

Antibiotics Shortage © Complexity Science Hub/Yang

The concentration of the market on China and India

1. There is a clear trend of increasing concentration of production in a few countries, namely China and India: The development of production concentration accelerated during the pandemic. Systemic trade risk indicators for China and India show sharp increases after 2018. These dependencies are not directly visible in import relationships but can be unravelled by tracing the production processes backwards along the antibiotics value chain.


2. The production concentration is more pronounced in upstream stages of the value chain (intermediaries and active pharmaceutical ingredients, APIs) rather than in downstream stages like unpackaged and packaged products. 76% of the manufacturing sites of intermediaries are located in China and India. 59% of API (active pharmaceutical ingredients) producers are situated in these countries.


3. Due to a higher production concentration in intermediaries and APIs, shocks affect APIs more strongly than packaged products. Hence, it becomes harder to find suitable substitutes when confronted with a shortage. In line with this observation, the estimated number of shortages that could be resolved by substitution halved in 2020. Negative impacts on patient care increased.


4. The data suggests a tendency towards market segmentation in which firms in European and North American countries developed an increasing dependence on Chinese suppliers. Indian producers trade mostly with local neighbours in Asia, Oceania and African countries.

SARS-CoV-2 led to reduced need for antibiotics

1. When the pandemic hit, non-pharmacological interventions aimed at curbing the spread of SARS-CoV-2 also reduced the circulation of other pathogens. Consequently, both community and hospital demand and consumption of antibiotics dropped sharply during the pandemic (by approximately 20%).


2. Reducing contagion risks, hospitals restricted their services to non-COVID patients. Drug shortages in hospitals nearly halved during the pandemic compared to the frequency of shortages in 2018. This uncovered improvement potential in inventory management and demand forecasting.


3. In 2022, in an increasing number of countries, SARS-CoV-2 related containment measures ceased. As societies by and large “returned to normal” so did antibiotic consumption. Volatile demand and geographically concentrated production systems led to simultaneous shortages of antibiotics across many parts of the world.


1. Supply disruption of medications in general, and antibiotics in particular, are not a new phenomenon. Since 2014, they have steadily increased in frequency and severity. In the vast majority of cases, it was possible to resolve these shortages by finding suitable substitutes and thereby reduce negative impacts on patients.


2. The consumption of antibiotics declined in many European countries between 2011 and 2018. This is welcome as it reduces the build-up of antimicrobial resistance.


3. These trends in antibiotics demand coincided with structural transformations in the antibiotics production system. Overall, there is a clear trend of increasing geographic concentration of production. Production sites in China and India have benefitted from this trend.

Policy considerations to avoid antibiotic shortages in the future

1. There are short-term and long-term remedies. Short-term remedies include improvements in the data, planning and forecasting infrastructure. This will require additional investments. Supply chain disruptions can reflect market structure problems. Long-term policies address the market structure and the international division of labour.


2. Existing supply networks arise from market processes that reflect a competitive combination of qualities and prices. Deviating from market results comes at a cost which can be interpreted as an insurance premium that health agencies need to incur to avoid impacts on patients. Additional costs require appropriate financing. Ideally, a well-designed market should internalise the risk of disruption, e.g., through appropriate contracts with the adequate incentive structures.


3. Antibiotics shortages need to be considered against the backdrop of the global risk of antimicrobial resistance. Ideally, policy remedies should address both issues.


1. Understand the scope of the problem. Data availability is an issue. Efforts need to be undertaken to not only track and forecast drug shortages, but to focus on shortages of non-substitutable drugs.


2. Demand forecasting and stable supplier relationships. Health authorities need to better understand the demand developments for antibiotics in the population. Evidence-driven demand planning could form the basis for building stable supply relationships, e.g., through multiyear contracts with producers that contain robust provisions in case of non-delivery.


3. Capacity markets and excess inventory. In case of emergencies, add-on production capacities that timely provide the drugs in question may address arising shortages from the outset.


4. Single Market. The European Union provides a powerful tool to mitigate supply risks across multiple players through the Single Market. Coordinated and more centralised EU inventories can also help to reduce overall safety stock and thereby avoid inefficiencies.


5. Bargaining power. Countries, regions, or health agencies might consider deeper cooperation and pursue joint forecasting and joint procurement strategies. Adverse effects on the market structure of suppliers and possibly anti-competitive behaviour of suppliers need to be monitored.


6. Diversification of supply. Diversification requires internationally competitive producers. Hence, the debate about broadening the supplier base is embedded in a wider discussion about competitiveness and structural change.


7. Subsidised procurement prices. Reimbursement models that delink development costs from unit sales have been proposed to tackle drug shortages.