Networks are everywhere. Of course, today we are used to social networks, where information and ideas can spread at an extremely rapid rate. If someone sends a meme to two friends, and they send that meme to two friends, and so on, the idea gets to thousands (or more) people with little effort. Social scientists call these “network effects”.

But this isn’t an article about modern-day social media. It is an article about networks in history. Although our ancestors did not have the technology to make information spread quite as quickly as we do today, historical networks were still mightily important for all sorts of economic, social, and political processes.

Clearly, it is hard to recreate personal networks, as we need a lot of information about the people in question. Who did they know? How well did they know them? Where were their contacts located? This is hard enough information to get for people in the contemporary world (unless you work for Facebook). It is next to impossible to get for most historical figures.

Recreating Martin Luther’s Network

The Protestant Reformation in the 16th century was a classic situation in which networks might really matter. Initially, adopting the Reformation was costly. Protestants were burned alive, and religious warfare was endemic. This is where networks may matter. When adoption is costly, knowing that your neighbors adopted is useful — at least you will have allies. Networks also help facilitate the spread of information, which is incredibly important for a movement steeped in ideology.

Fortunately, Martin Luther (1483–1546), the leader of the Protestant Reformation, left enough traces where it is possible to recreate his network. In a recently-published paper with Yuan Hsiao, and Steven Pfaff, two sociologists, we attempt to understand the role that Martin Luther’s network played in the spread of the early Reformation. In one sense, we were lucky we were able to do this. Luther was an important person who has fascinated historians for centuries. As such, there has been much work done by a panoply of historians documenting numerous aspects of Luther’s life.

Many of the letters he wrote survived and have recently been digitized. For the sake of reconstructing Luther’s spatial network, we coded every city in which Luther sent a letter (see below).

Due to the meticulous work of Georg Buchwald, we know each city that Luther visited and when he visited it. From the work of Hyojoung Kim and Steven Pfaff, we know the location of each of Luther’s students at the University of Wittenberg. While these data of course do not tell us every person Luther ever met, it does give us a pretty good indication of whom he had some influence with.

Luther’s Network and the Spread of the Reformation

Luther was influential. Whom he was connected to may have therefore mattered. We can conceive of the process as leader-to-follower, originating with Luther and flowing to local elites through personal ties. Luther played the role of a global opinion leader based in Wittenberg. He had ties with local elites in towns across Central Europe, who, in turn, exerted influence in their towns.

But the Reformation may have also spread independently of Luther. Because it was costly to adopt, towns were only likely to adopt if they were connected to another town, via a network, that also adopted. We could therefore envision the Reformation as a “virus” that spread out of Wittenberg. It is certainly a possibility we cannot discount without evidence.

So, which was it? Was Luther completely unimportant to the spread of the early Reformation? Would any “heretical” movement coming out of Wittenberg have lit the anti-papal fuse, even in the absence of Luther?

This is unlikely. As we show below, Wittenberg was not very well connected within the Holy Roman Empire (via trade routes). It was, in Luther’s words, “on the edge of civilization.” So maybe the trade network was not enough to spread the Reformation.

But was Luther’s network enough, on its own, to account for the spread of the Reformation? Here, our analysis also finds little evidence in favour. We conduct a simulation analysis and show that Luther would have had to have been improbably “infectious” for his network to fully explain the spread of the Reformation. That is, way too many nodes in Luther’s network would have had to adopt the Reformation because of Luther (and not a myriad of other causes contributing to the success of the Reformation) for this to be the dominant explanation.

So, did Luther’s network matter at all? In short, yes. We argue — and our simulations and regression analyses support — that personal/relational diffusion via Luther’s network combined with spatial/structural diffusion via trade routes. It was the combination of the two diffusion processes that helped Protestantism’s early breakthrough from a regional reform movement to a general rebellion against the Roman Catholic Church.

The figure below illustrates this idea. Multiplex ties point to how Luther as an opinion leader mobilized his personal network through an ensemble of letters, visits, and student relationships. It also shows how Luther’s network blends with the spatial (trade) network to create complex contagion processes operating at the intersection of information flow and social influence.

The Value of Interdisciplinary Work

The two of us conceived of something along the lines of this paper a number of years ago over lunch. We both have studied the Reformation in depth from a social science perspective. We agreed that insights on Luther’s role in its spread were sorely lacking. We decided to write our friend, Steve Pfaff (a sociologist), to see if he would have interest in pursuing a project like this. Steve is also an expert on the Reformation and has worked on similar topics in the past. He wrote us back immediately, saying he had already started collecting data for a similar project. He brought on board one of his graduate students, Yuan Hsiao, an expert on both networks and simulations. As we proceeded with the project, we soon realized that we were not only making a contribution to the literature on the Reformation, but also to the much larger literature on networks and diffusion. We were able to show using a real-life network that multiple diffusion mechanisms can reinforce each other to facilitate spread within a network, even if each mechanism by itself is not sufficient.

This is not an insight we expected to find when starting this project. It was the fruit of an interdisciplinary effort that leaned heavily on sociology, economics, and network theory. We doubt any of us would have been able to write this paper or convincingly test its insights without the help of all of the other co-authors. We think the interdisciplinary nature of our work really made things possible that would not have been possible had we remained in our economics or sociology silos.

The Authors

Sascha O. Becker is Xiaokai Yang Chair of Business and Economics at Monash University, Melbourne, part-time Professor at the University of Warwick, Research Associate at CAGE, and Principal Investigator at SoDa Laboratories, Monash University.

Jared Rubin is a professor of economics at Chapman University. He is an economic historian interested in the role that Islam and Christianity played in the long-run “reversal of fortunes” between the economies of the Middle East and Western Europe.

The Paper is available open access: https://doi.org/10.1177/0003122420948059

Becker, Sascha O., Yuan Hsiao, Steven Pfaff and Jared Rub (2020) “Multiplex Network Ties and the Spatial Diffusion of Radical Innovations: Martin Luther’s Leadership in the Early Reformation” American Sociological Review, 85(5): 857–894.

How Martin Luther’s network helped him make the Protestant Reformation a success was originally published in Monash SoDa Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

Monash IP Observatory

Hundreds of thousands of people globally have marched in protests against police brutality and systemic racism, spurred by the killing of George Floyd in Minneapolis on the 25th of May. In the USA, over 450 protests have flared up across every state since the first demonstration in Minneapolis on the 26th of May.

The Black Lives Matter movement has since taken social media by storm, enabling the rapid sharing and accessing of information via these platforms. One social media trend, the controversial #BlackoutTuesday, saw over 23.3 million posts featuring plain black squares on the 2nd of June. Initially organised by the entertainment industry as a gesture of solidarity, this trend was widely criticized for clogging up social media and preventing the spread of useful information about the protests and the movement more widely.

Coordinated gatherings of the scale seen in the United States have the potential to locally warp internet speeds as masses of people create, post and upload content about activities happening on the ground. The demand for bandwidth can lead to a surge in signal latency, which serves as a proxy measure for local internet pressure.

At the Monash IP Observatory, our technology has previously provided internet regulators in the United States with comfort around the stability of US internet infrastructure under rolling COVID-19 lock-downs.

But it turns out that the Black Lives Matter movement sent ripples through the internet in ways that COVID-19 hadn’t (so far).

Here we document what we saw.

Southern California

Across several Southern CA cities, we observed large spikes in internet pressure, i.e. slower immediacy of internet speeds, on the 31st of May and 2nd of June. Los Angeles, San Diego, Bakersfield and Oxnard were the biggest hit on these dates.

The 31st of May marked the end of the first weekend of nationwide protests, as well as the implementation of curfews in many larger cities in response to looting and rioting.

Interestingly, the spike on Tuesday the 2nd of June occurred around the same time as Blackout Tuesday, a.k.a. #TheShowMustBePaused. With a huge, temporary increase in online activity on this day, increases in internet pressure are not unexpected, especially in entertainment industry hot-spots like LA.

These spikes could reflect changes in internet usage on the ground, infrastructural limitations in the region, or a combination of both.

With a rise in protest coordination on the 30th of May, the first Saturday since Floyd’s death, it is possible that there was added pressure on the demand for bandwidth as people shared and organised these events.

Los Angeles

LA is the largest city in CA, with a population of over 4 million people. Protests in LA began on the 27th of May, 2 days after Floyd’s death. After protest crowds started to become violent on the afternoon of Saturday the 30th, a curfew was introduced to cover the following week of demonstrations.

San Diego

San Diego sits about 190km south of LA, next to the Mexican border. In San Diego, protests began on the 29th of May. While most protests in San Diego were peaceful, some turned violent, with over 100 arrests made on Sunday May 31st, spurring curfews starting from 8pm that night.

Bakersfield

Bakersfield is a Southern Californian city sitting North of LA, with a population of ~380,000.

Oxnard

Oxnard is West of LA and South of Bakersfield, with a population of ~200,000.

Other cities of interest

Tampa Bay area, Florida

The Tampa Bay area is a coastal metropolitan region in Western Florida which includes the city of Tampa, St Petersburg and Clearwater and has a population of ~2.7 million.

Many protests in the Tampa Bay area remained peaceful over the first weekend (30–31st of May), but crowds became violent at nighttime, with several arrests reported in Tampa.

Protest crowds remained strong throughout the first week, with police beginning to take more aggressive measures to contain rioting crowds from the evening of Tuesday the 2nd of June.

There was a large spike in internet pressure in the Tampa region between the 31st of May and the 5th of June.

Denton, Texas

Denton is a small city in Texas, around 60km North-West of central Dallas.

Protests around Denton county courthouse square drew thousands over the weekend of the 30th of May and continued well into the following days and weeks.

With an evening curfew introduced from the 31st May, police were asking for voluntary compliance with crowds that were largely peaceful.

There was a period of high internet pressure from the 31st of May to the 5th of June with similar characteristics to the spike observed in Florida.

Anchorage, Alaska

Peaceful protests in Alaska’s largest city began on May 30th and have continued for several weeks. Internet pressure was sustained at a higher level between the 31st of May and 5th of June.

Springfield, MA

Demonstrations in Springfield have continued since the weekend of the 30th of May, with protesters gathering in large numbers outside police headquarters on June 3rd (Pictured). Around this date we also see a rise in local internet pressure.

Minneapolis

Minneapolis, the epicentre of nationwide #BLM protests, has remained relatively stable in regards to internet pressure over this turbulent time period.

Washington, DC

Tens of thousands of DC residents have turned out to protest the death of George Floyd over the past month, with the plaza running in front of the White house being officially renamed “Black Lives Matter Plaza”. Despite this, the local internet infrastructure has had no problem handling the added demand.

Our Methodology

To generate the data behind these observations, we combine a commercially available geo-located IP database with our powerful scanning technology which measures the online or offline status of millions of Internet addresses globally every hour.

Our observational methodology uses the most basic Internet messaging protocol that is widely used billions of times a day to establish routes for your email, tweet, or share. After developing a carefully selected set of Internet addresses (IPs) to measure, we periodically send them one of these tiny messages, essentially asking, ‘Are on you online?’. These online/offline answers form the basis for our ‘connectivity’ indicators.

In addition, we also receive back from these measurements the responsiveness, or latency, of the IP (measured in round-trip-time, or rtt). Latency is a reasonable proxy for the experienced speed of connection, especially for any user who is interacting with a major social platform where even basic chat activities to other users nearby must travel to a server well beyond national borders (and back again).

Importantly, the IP Observatory has no access to any content being shared, viewed, visited, or generated by a user at a given IP, and all IP Observatory activity works in aggregates of thousands of randomly sampled measurements across geo-spatial sub-regions.

Acknowledgement: Hayley Lock assisted with background research and analysis in the preparation of this article.

The mission of the Monash University IP Observatory — ‘internet insights for social good’ — is to monitor the availability and quality of the Internet during critical events such as elections, natural disasters or conflicts. The IP-Observatory is fully compliant with the EU’s General Data Protection Regulation (EU-GDPR). The IP-Observatory does not collect, hold or process personal data. The IP Observatory was founded by Klaus Ackermann, lecturer in Econometrics and Business Statistics, and Simon Angus, and Paul Raschky, Associate Professors in Economics. The observatory is a project of SoDa Laboratories at the Monash Business School, and tweets @IP_Observatory.

Originally published at https://medium.com on July 20, 2020.

How the BLM Protests Shook US Internet Infrastructure was originally published in Monash SoDa Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

While social scientists can run experiments to learn about the effects of a broad range of treatments, it is impossible to randomise political systems. Yet, understanding the effects of communism, democracy, and autocracy on people’s lives is of great importance. Not surprisingly, there is a wide literature exploring how political systems persistently affect the economy and population preferences, with a particular focus on different legacies between capitalist and socialist societies (for a review, see Simpser et al. 2018).

Studying persistent East-West differences after German division and reunification

German division and reunification has attracted the interest of many social scientists as it seems quite close to an experimental setting. After WWII, two previously united parts of the same country were assigned to two opposing political regimes, a capitalist West and a communist East. Reunification in 1990 brought the two parts under the same political system once again.

In terms of overall economic outcomes, income per capita did not differ widely between East and West Germany before WWII (see Alesina and Fuchs-Schündeln 2007). But by the time East Germany collapsed, East German GDP per capita was less than half of that of West Germany. After reunification, labour productivity in East Germany was at a third of the Western level, putting the East somewhere between Mexico and Chile. The communist system had ended in economic failure.

Given salient differences between the political and economic systems of West and East Germany, a large literature has argued that the communist experience had enduring effects on the population in the East, including their economic outcomes, political attitudes, cultural traits, and gender roles (e.g. Alesina and Fuchs-Schündeln 2007, Campa and Serafinelli 2019, Laudenbach et al. 2019, Goldfayn-Frank and Wohlfart 2020, Lippmann et al. 2020).

In a recent paper (Becker et al., forthcoming), we take a fresh look at the German case. Were East and West indeed similar before World War II? Did the war and the subsequent occupation affect the two parts of the country in the same way? What about migration between East and West from 1945 until the building of the Berlin Wall in 1961? And what does all this imply for our understanding of the effects of communism?

East Germany can be spotted before it even existed

The location of the border between the German Democratic Republic (GDR) and the Federal Republic of Germany (FRG) is not the random outcome of where American, British, and Soviet tanks stopped at the end of WWII in 1945. Instead, in anticipation of the defeat of Nazi Germany, the three allied forces had agreed in 1944 on a division of post-WWII Germany into Soviet and Western occupation zones that followed the pre-WWII borders of the German Empire states and the provinces of the largest state, Prussia (with a few very minor exceptions for geographic connectedness). As a result, the East-West border separated the populations of pre-existing regions with distinct histories and cultures.

Since the border follows pre-existing regions, we can explore pre-WWII county-level data to investigate whether West and East differed in relevant dimensions. A first dimension is the size of the working class, strongly emphasised by communist countries. Inspecting pre-WWII data, we see that the East Germany already had a substantially higher working-class share in 1925 (Figure 1), well before the area became communist. The difference between East and West in working-class share amounts to 12 percentage points. In fact, the working-class share jumps quite abruptly in several regions around the later inner-German border: it is significantly detectable when focusing on counties within 100 kilometres of the later border or on the counties that have a direct contact with the later border.

Figure 1 The working-class share in 1925: East-West differences before the GDR existed

Source: Becker et al., forthcoming.

East-West differences before WWII are salient in other dimensions as well. Figure 2 shows comparisons in economic outcomes relevant to the onset of communism. Among others, the manufacturing-employment share was significantly higher in the East, whereas the East’s share of the population that is self-employed was significantly lower (Fritsch and Wyrwich 2014).

Figure 2 East-West differences before WWII

Source: Own depiction based on Becker et al., forthcoming.

Interestingly, political preferences also already differed before WWII. The communist vote share in the East was double that of the West in 1924. When looking at left-wing parties more broadly (in particular, adding Social Democrats), the left vote share was 15 percentage points higher in the East.

Communism is often thought to have crowded out religion. But East Germany already had lower church attendance (in the Protestant church) at the beginning of the 20th century (Hölscher 2001). Furthermore, while the West was roughly balanced between Protestants and Catholics, the East was predominantly (91%) Protestant (Becker and Woessmann 2009).

Finally, the socialist GDR put much emphasis on getting women to work. Yet, already before WWII, female labour-force participation was higher in the East (Wyrwich 2019). In addition, East and West differed markedly in the incidence of extramarital births before WWII (Klüsener and Goldstein 2016).

To the extent that some of these pre-existing differences persisted through the communist period, they may well be an essential source of post-reunification heterogeneity between East and West Germans.

WWII and occupying forces affected East and West differentially

East and West were differentially affected by WWII and the occupying forces. Using data from the German Census jointly administered in all four occupation zones in October 1946, we show that the ratio of men to women was substantially lower in the Soviet zone. No such differences had existed in the last pre-WWII census, in 1939.

The larger decrease in the sex ratio in the Soviet zone may reflect a larger war-related male death toll, but potentially also mirrors sex imbalances in very early East-West migration. Whatever the source, the difference may have contributed to differences in several outcomes, such as female labour-force participation, gender roles, and even political views.

The East also suffered larger losses due to the dismantling of capital equipment by and paying reparations to the occupying forces during 1945–1949. This gave the GDR a worse starting position than the FDR (Sleifer 2006).

A selective fifth of the population left the East before building of Berlin Wall

The four occupation zones were established in 1945, and the GDR was founded in 1949. But it was possible, albeit increasingly difficult, to migrate between the two parts of Germany until the Berlin Wall was built in 1961. In fact, about one in five of East Germany’s population migrated West until 1961. While there is no data to compare these emigrants to those who stayed behind in the East, we show that individuals who moved from East to West differed from local West Germans in being more likely to be white-collar workers, self-employed, and better educated. Presumably, they were also less receptive to the communist doctrine (see also Bauernschuster et al. 2012).

What is sometimes overlooked is that also about half a million people migrated from the West to the East before 1961. GDR propaganda describes them as “not in agreement with the capitalist system”. We show that six of the 19 Politburo members in the early GDR (1949–1961) had been born in what became West Germany, including long-time GDR leader Erich Honecker. Taken together, the evidence suggests that there was selective migration and sorting by political preferences.

Caution warranted when interpreting evidence on ‘effects’ of communism

Considering these findings of pre-existing East-West differences, differential effects of WWII and the subsequent occupation, and selective East-West migration, is the German setting still useful to study the impact of communism? We think that the answer is yes, as it does offer some unique advantages.

However, we emphasise that each research question requires sensible consideration of the outlined challenges. The most convincing evidence of the effect of political systems probably stems from the post-reunification convergence of some economic behaviours, political preferences, and trust in the state. The experience of having lived in the communist system also seems to have permanently altered consumption patterns. In addition, the communist system likely shaped gender roles in terms of female labour-force participation or fertility preferences, but these also include a strong legacy component already visible before WWII.

The more general insight is that the development of political systems is hardly ever exogenous. For example, political systems become endogenous if political preferences are endogenous to prior experiences (Fuchs-Schündeln and Schündeln 2015). This idea is most apparent in revolutions started by groups frustrated with the current system. But effects of political systems must be carefully assessed even if a new political system is imposed by external players: regime changes may consider pre-existing conditions, and people dissatisfied with the new regime may simply choose to ‘vote with their feet’ and emigrate, leaving behind a population reasonably well aligned with the new regime.

References

Alesina, A, and N Fuchs-Schündeln (2007), “Goodbye Lenin (or not?): The effect of communism on people”, American Economic Review 97 (4): 1507–28.

Bauernschuster, S, O Falck, R Gold and S Heblich (2012), “The shadows of the socialist past: Lack of self-reliance hinders entrepreneurship”, European Journal of Political Economy 28(4): 485–97.

Becker, S O, L Mergele and L Woessmann (2020), “The separation and reunification of Germany: Rethinking a natural experiment interpretation of the enduring effects of communism”, Journal of Economic Perspectives 34(2), forthcoming.

Becker, S O, and L Woessmann (2009), “Was Weber wrong? A human capital theory of Protestant economic history”, Quarterly Journal of Economics 124(2): 531–96.

Campa, P, and M Serafinelli (2019), “Politico-economic regimes and attitudes: Female workers under state socialism”, Review of Economics and Statistics 101(2): 233–48.

Fritsch, M, and M Wyrwich (2014), “The long persistence of regional levels of entrepreneurship: Germany, 1925–2005”, Regional Studies 48(6): 955–73.

Fuchs-Schündeln, N, and M Schündeln (2015), “On the endogeneity of political preferences: Evidence from individual experience with democracy”, Science 347(6226): 1145–8.

Goldfayn-Frank, O, and J Wohlfart (2020), “Expectation formation in a new environment: Evidence from the German reunification”, Journal of Monetary Economics, forthcoming.

Hölscher, L (2001), Datenatlas zur religiösen Geographie im protestantischen Deutschland: Von der Mitte des 19. Jahrhunderts bis zum Zweiten Weltkrieg, 4 vols., Berlin: Walter de Gruyter.

Klüsener, S, and J R Goldstein (2016), “A long-standing demographic East-West divide in Germany”, Population, Space and Place 22(1): 5–22.

Laudenbach, C, U Malmendier and A Niessen-Ruenzi (2019), “The long-lasting effects of experiencing communism on attitudes towards financial markets”, working paper.

Lippmann, Q, A Georgieff and C Senik (2020), “Undoing gender with institutions: Lessons from the German division and reunification”, Economic Journa l, forthcoming.

Simpser, A, D Slater and J Wittenberg (2018), “Dead but not gone: Contemporary legacies of communism, imperialism, and authoritarianism”, Annual Review of Political Science 21(1): 419–39.

Sleifer, J (2006), “Planning ahead and falling behind: The East German economy in comparison with West Germany 1936–2002”, in Jahrbuch für Wirtschaftsgeschichte, Beiheft 8, Berlin: Akademie Verlag.

Wyrwich, M (2019), “Historical and current spatial differences in female labour force participation: Evidence from Germany”, Papers in Regional Science 98(1): 211–39.

Originally published at https://voxeu.org.

Discussion on A CORRECTION: A PODCAST

German division and reunification and the ‘effects’ of communism was originally published in Monash SoDa Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

When faced with intense public discontent, do country leaders engage in strategic foreign interactions to distract the domestic population? In a recent paper, I explore this question in detail, specifically in the context of the use of ‘diversionary foreign policy’ by leaders.

From the Argentinian military junta’s diversionary Falkland’s War in 1982, to Bill Clinton’s famous airstrikes against Sudan and Afghanistan in 1998 (ironically when the Monica Lewinsky scandal was at its peak), world history provides ample anecdotal evidence of how ‘unpopular’ leaders resort to violent inter-state conflicts to divert public attention away from domestic woes. More recently with the advent of social media, some eagle-eyed observers have noticed that diversionary foreign interactions can even take a more subtle, verbal form, which does not involve the significant costs and risks associated with violent conflict. From Emmanuel Macron to Donald Trump, many leaders have been observed as making statements targeting foreign countries that achieve the objective of diverting the attention away from pressing domestic issues.

Challenge 1: How can domestic turmoil and foreign interactions be quantified?

The main challenge for the empirical analysis of this relationship is the absence of consistent and quantified indicators of domestic turmoil and foreign interactions.

Political scientists have typically used economic indicators such as the rate of inflation or the rate of unemployment as proxies for the public’s discontent with the government (see, for example, Morgan and Anderson, 1999; Leeds and Davis, 1997; Miller, 1995). However, domestic turmoil is a broader concept that transcends mere economic hardships, and therefore, simple economic indicators are not ideal tools to gauge its intensity. Surveys on respondents’ perceptions of leader performance, such as those derived from the World Values Survey or the Afrobarometer survey, are potentially more suited to quantify domestic turmoil, but such measures are unavailable consistently, and at fine temporal resolutions, over long periods of time.

To overcome this data challenge, I generate a quantified measure of domestic turmoil (DT), using the public’s revealed preferences that materialize in the form of actual physical events targeting governments.

In particular, I peruse the GDELT database, which is a massive real-time collection of events occurring around world, as reported in print, broadcast and web media articles. For each event, GDELT reports a number of important attributes, including information on actors involved, the location, and a number of event features.

To identify events representing domestic turmoil I leverage on the Goldstein score, which identifies the theoretical impact an event can have on the political stability of a country. The score ranges from -10 (highly negative effect on political stability) to +10 (highly positive event on political stability). I classify all domestic events targeting the government, initiated by domestic parties, and recording a Goldstein score of less than -5, as contributing towards domestic turmoil. (The empirical results are robust to a range of alternative cut-offs of the Goldstein score.) Normalized based on the total number of domestic events targeted at the government, I then obtain a standardized index of domestic turmoil — the DT index — which can be used in my quantitative analysis. This index serves as the key explanatory variable of my study.

Figure 1 plots the DT index for a sub-sample of countries. I observe that in countries such as USA, UK, Canada and Australia, the distribution lies mostly between 0.2 — 0.6 range (on a scale of 0 to 1). The variation is significantly higher in Latin American countries and African countries, sometimes even reaching the highest possible value as per the scale, signalling extremely high levels of domestic turmoil. Interestingly, in Middle-Eastern countries such as the UAE, the score lies mostly at the lower end of the spectrum, sometimes even reaching zero, signalling extremely low levels of domestic turmoil. Although there may be a number of within-country factors affecting the DT index which need to be appropriately addressed in the design of the empirical identification strategy (Challenge 2), what Figure 1 reveals is the relatively large variation in the raw DT index within and between countries.

Figure 1: Distribution of DT in a sub-sample of countries

Next, I look at constructing the outcome variable of the study — a quantitative measure of governments’ foreign interactions. Due to the increased recent attention on verbal diversionary interactions, I am ex-ante agnostic about the nature of diversion. Accordingly, my objective here is to not only quantify the frequency of foreign interactions, but also their nature, ranging from cooperative to aggressive.

Again, I resort to the GDELT database to generate this measure. Within the GDELT database, all events are classified under four broad categories, based on the cooperative/aggressive nature of the act, at varying degrees of intensity. These categories enable me to identify four broad variants of the foreign interactions pursued by governments, i.e., verbal cooperation, material cooperation, verbal conflict, material conflict. Accordingly, I identify all foreign interactions initiated by governments under these ‘quad’ classes, and normalize them by the total number of foreign interactions, in order to arrive at standardized indicators that capture both the nature and frequency of governments’ foreign interactions. These four indicators act as the key outcome variables of this study.

Challenge 2: Identification issues

Now that I have obtained quantified indicators for foreign interactions and domestic turmoil, the next question that arises is on the direction of the causal relationship. Do high levels of domestic turmoil lead to governments engaging in diversionary foreign interactions? Or, do increased foreign interactions fuel domestic turmoil? Could there also be other unobservable factors that simultaneously influence both domestic turmoil and diversionary foreign interactions?

To address these concerns, and to ensure that I capture the true causal relationship between these two variables, I employ an instrumental variables (IV) strategy, where I focus on a specific `exogenous’ shock that directly, and only, affects domestic turmoil, i.e., football losses.

Many significant events throughout the history of sports exhibit how key sporting events and their outcomes shaped societies for years to come. The most prominent example probably is the 1954 FIFA world cup final, where West Germany, the underdogs, defeated Hungary, the favourites, and this ‘miracle of Bern’ is popularly viewed as a key turning point in West Germany’s socio-economic revival following the world war. Another example comes from the football war of 1969, where tensions between fans of El Salvador and Honduras following a FIFA World Cup qualifier led to the breakout of war between the two countries. More recently, Iraq’s win of the AFC Asian Cup in 2007 is widely believed to have unified the country despite many domestic political issues ranging from ethnic factionalism to invasion by the US military.

Complementing such anecdotal evidence is a growing body of literature which suggests that sports outcomes, and in particular, football outcomes, do drive people’s sentiments in a manner that leads to substantive changes in their behavior. For instance, Edmans, Garcia and Norli (2007) find evidence of a significant decline in stock market performance following soccer losses. Card and Dahl (2011) find that emotional cues related to football losses affect reported events of family violence, while Ge (2018) shows that sports outcomes affect the tipping behaviour of taxi commuters. More relevant to my work, Depetris-Chauvin, Durante and Campante (2020) find that football wins lead to the emergence of the national identity, overriding ethnic identity, in the context of Africa. My empirical strategy is complementary to theirs, focusing on the negative shock to public sentiment following football losses, instead of the positive effect following football wins.

One can argue, however, that in instances where a football match is played between a strong team and a weak team, it is possible to predict the winner beforehand, with a high degree of accuracy. To address any such concerns on the exogeneity of the IV, I restrict my focus to football matches played between teams of similar performance levels, as measured by World Football Eloratings, which is the most prominent ranking system of football teams across the world. Moreover, it is possible that the timing of the match influences governments to engage in foreign interactions, in which case the exclusion restriction of the IV may be threatened. I address this concern by considering the direct effect of the football loss on DT, conditional on the timing of the match.

Findings

In the next step, I apply the quantified indicators, along with the identification strategy, to a sample of 190 countries, over the years 1997–2014, and at the fine monthly temporal resolution, to explore whether and how domestic turmoil leads to diversionary foreign interactions.

Using close football losses as an IV, I find that high levels of domestic turmoil do in fact result in leaders engaging in more foreign interactions. In terms of the nature of such diversion, I do not find any evidence of leaders engaging in violent conflicts as a diversionary tactic. In fact, what I do find is support for the more subtle, verbal form of aggression that has become more manifest in recent times. Specifically, I find that a one standard deviation increase in domestic turmoil leads to approximately 8% increase in foreign interactions classified as verbal aggression. I also find that these diversionary effects are short-lived and not persistent over time. These results suggest that leaders do resort to diversionary tactics when faced with public resentment, and that they opt for verbally aggressive diversion as opposed to violent conflicts, demonstrating a preference for a short-term, low-cost, low-risk strategy to satiate the public.

Finally, using a range of inter-state connectivity measures, I explore whether leaders consciously target the victims of their diversionary tactics. I find that diversionary verbal aggression is typically targeted at culturally similar countries, as defined by countries closely connected via religious and linguistic ties. Apart from cultural similarity, I also find that leaders are less likely to target countries with whom they share a strong trade relationship, in terms of both imports and exports. Moreover, `weak’ countries, i.e. those with low levels of military capability and low levels of population, are more likely to be targeted. These signals on target selection further illustrate that diversionary foreign interactions are largely exercised by leaders in a manner that does not lead to significant risks of economic or military retaliation.

Accordingly, applying new data to quantify a societal aspect which has largely escaped the scrutiny of researchers, I provide interesting insights on how country leaders engage in the strategic manipulation of the masses. The evidence strongly suggests that verbally aggressive foreign interactions are systematically utilized as a short-term, low-cost diversionary tool, to distract the public from their resentment towards the government.

References

Amarasinghe, A. (2020), “Diverting domestic turmoil”, Unpublished Manuscript.

Card, D. and G. Dahl (2011), “Family violence and football: The effect of unexpected emotional cues on violent behaviour,” Quarterly Journal of Economics 126, 103–143.

Depetris-Chauvın, E., Durante, R. and F. R. Campante (2020), “Building nations through shared experiences: Evidence from African football,” American Economic Review 110, 1572–1602.

Edmans, A., Garcıa, D. and O. Norli (2007), “Sports sentiment and stock returns,” Journal of Finance 62, 1967–1998.

Ge, Q. (2018), “Sports sentiment and tipping behavior,” Journal of Economic Behavior & Organization 145, 95–113.

Leeds, B. and D. Davis (1997), “Domestic political vulnerability and international disputes,” Journal of Conflict Resolution 41, 814–834.

Miller, R. (1995), “Domestic structures and the diversionary use of force,” American Journal of Political Science 39, 760–785.

Morgan, T.C. and C. Anderson (1999), “Domestic support and diversionary externalconflict in Great Britain, 1950- 1992,” Journal of Politics 61, 799–814.

On the politics of distraction was originally published in Monash SoDa Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

Using remotely collected, alternative data, we can begin to document the month of darkness in more detail.

Here, we estimate the extent of infrastructure degradation at 12 important sub-national municipios using data from the Monash University IP Observatory (@IP_Observatory).

Our methodology uses the most basic internet messaging protocol, widely used billions of times a day to establish routes for your email, tweet, or share.

By sampling the online or offline status of a large sample of known geo-located internet addresses (IPs) in each municipio, we are able to build up a national, or sub-national picture of the electricity and internet infrastructure in near real time.

The national view

Nationally, Venezuela has endured 7 major outage events beginning at around 5pm on the 7th March (indicated in panel A, below). From a baseline of around 90% connectivity, the average index we obtain from a week of unaffected measurements in early February, the two major outages on the 7th and 26th took national connectivity below 6%.

Recovery in both cases was protracted, and interrupted by significant set-backs.

Documenting the dark times at sub-national level

Whilst the national picture is dismal enough, diving deeper to the sub-national level reveals the true impact of these outages.

In the second panel (B, above), we use our statistical anomaly detector to mark in black on the timeline where any one of 12 important municipios suffered a significant degradation to their connectivity.

What the figure immediately reveals is the grinding, slow recovery to ‘normal’ experienced by a majority of the municipios in the sample. Whereas an outage of more than a few minutes is considered ‘major’ in a modern economy, the first outage caused degraded connectivity for an average of 5 days.

Urdaneta (Miranda state), on the other hand, we estimate endured a mind-boggling degradation period of 253 hours, from 8pm, 7 March to 9am, 18 March, 10.5 days later. It is truly impossible to imagine the impact of this kind of degradation on the ground.

Indeed, if we tally up the total periods of degradation affecting each municipio since 7 March, we find eight municipios affected for more than 10 days, and two (Girardot, Aragua state; Libertador, Distrito Capital state) affected for more than 16 days.

Tallies for all municipios are provided to the right of the timelines in panel B.

To be clear, our statistical detector is not measuring complete outages in these locations, but rather, periods of anomalous connectivity —representing smaller or larger pockets of the region with no connectivity, or intermittent service affecting many households, or a combination of both.

What of the future, Dark April?

Panel A, at national level, shows the nation reaching up towards normal connectivity. Panel B, tells the richer story that a number of significant areas are still struggling with degraded infrastructure.

Our latest national reading, at 5am, 4th April 2019 local time in Venezuela, just a couple of hours ago, shows a connectivity index of 77.3.

The longest that Venezuela has been able to keep the lights on at that level since March 7 was a precious 13 days.

But of course, it isn’t just about the lights. When the power goes down, Venezuelans lose access to a fundamental human right: their online freedom.

At a time such as this, that freedom is at a premium.

Let’s hope April is a month of light in Venezuela, in all senses.

The mission of the Monash University IP Observatory — ‘internet insights for social good’ — is to monitor the availability and quality of the Internet during critical events such as elections, natural disasters or conflict to provide. The observatory was founded by Klaus Ackermann, lecturer in Econometrics and Business Statistics, and Simon Angus, and Paul Raschky, Associate Professors in Economics. The observatory is a project of SoDa Laboratories at the Monash Business School, and tweets @IP_Observatory.

Maduro’s Land of Darkness: documenting a month of degradation with remote, alternative data was originally published in Monash SoDa Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.