Environmental Shocks and their Effects on Imperial Rome’s State Capacity

Droughts of Dismay: Rainfall and Assassinations in Ancient Rome

By Cornelius Christian (Brock University) and Liam Elbourne (St. Francis Xavier University)

Abstract: We find that lower rainfall in north-central Europe (Gaul/Germania) predicts more assassinations of Roman emperors from 27 BC to 476 AD. Due to agricultural pressures on Germanic tribes, low precipitation caused more barbarian raids. These raids, in turn, weakened the Empire’s overall political stability, and reduced the costs of assassinating an emperor. We buttress our empirical analysis with case study evidence.

URL: http://d.repec.org/n?u=RePEc:brk:wpaper:1703&r=his

Circulated by nep-his on 2017/10/08

Review by Fernando Arteaga (George Mason University)



Was Imperial Rome’s political stability disturbed by environmental shocks? If so, what were the transmission channels? These are the two fundamental questions the authors aim to answer. Their thesis is straightforward: As any pre-industrial society, rainfall levels predicted agricultural output in Roman times. A lack of rain affected food availability, especially in the underdeveloped regions where Northern Germanic Tribes resided, making these societies more prone to raid Roman towns across the border. The incursions then created political conflict among the Romans themselves.


The text relies on econometric analyses and a couple of case studies to back up the argument. The main statistical test is simple: they regress Roman political stability on rainfall data. The main variable they use as a proxy of political unrest is the assassination of Roman emperors (as presented by Scarre 1995): the more emperors were killed, the less stability in the Empire. Alternatively, they also employ an index of inflation and new governmental infrastructure investment as a proxy for stability  (larger inflation and less imperial projects imply improved stability).  The rainfall variable comes from Buengten et al. (2011) own estimations on precipitation levels across France and Germany for all the period under study. Figure 1 displays the main data points used in the analysis. The authors find that negative rainfall shocks are both associated with more emperor’s being killed (Figure 2) and with having larger inflation rates and fewer investment projects. A decrease of one standard deviation in precipitation caused an 11.6% standard deviation increase in assassination probability. The regression is empirically valid because there is no possibility of reversal causality; precipitation is not a factor that may be influenced by Roman politics [1].



Figure 1: Roman Gaul/Germania (Yellow). Rainfall Datapoints (Green). Emperor Assassination Locations (Red)




Figure 2: The red line indicates the precipitation level, while the blue is the amount of Roman Emperors assassinated.


But how exactly does lack of rain destabilized Roman society? The paper’s hypothesis relies on the Germanic raid linkage: Germanic tribes attacked Rome when they had a poor harvest of their own, which then created unrest in Roman interior stability.  To test such assertion, they regress Germanic/Gaul incursions on the rainfall levels. The raid data they use comes from Venning (2010), who reviewed the many times the Roman Empire suffered raids through its history. The authors find a negative correlation: a decrease in one standard deviation in rainfall is associated with a 4% standard deviation increase in a number of raids. They corroborate the results by doing some robustness tests: 1) a placebo test, in which they regress non-Germanic raids on precipitation levels, which they find that had insignificant impact (which means that precipitation mattered only in Germanic zones, because they were the only ones that really suffered from a lack of rain); 2) an instrumental variable where the relationship between Roman instability and Germanic incursion is instrumented by rainfall. They find that “a standard deviation increase in the raid dummy [the presence of raiding] causes a 29.3% increase in the probability of assassination.”

To give more weight to their results, they present a brief recapitulation on the reigns of two assassinated Roman emperors: Severus Alexander (208-235) and Gallienus (218-268). The key insight is that both emperors faced important challenges on the Eastern and Northern borders, however only the latter had a relevant impact on Roman internal politics. On the East, the Roman Empire frequently collided with the Sassanid Empire (the other larger state in the area), but notwithstanding the severity of the clashes (at some point they even captured a Roman Emperor, the father of Gallienus) it never caused great civil unrest in Rome. However, on the North, Rome bordered the Germanic tribes (scattered non-organized societies) that did affect Rome’s stability.  The conclusion we get from the narrative is that the Germanic border was important/special precisely because it was very susceptible to environmental shocks, which then led to constant raiding; unlike the Sassanid border, in which the Romans faced a cohesive society that could successfully resist bad crops or confront military bravados on non-environmental factors.



I enjoyed reading the paper very much. It made me re-realize why I find Economic History fascinating: it deals with topics that are interesting in themselves (the politics of the Roman Empire! What is not to like about it?), that remain relevant for today’s problems (we still seek to understand the relationship between nature and political conflict very much), and it treats the issues under study with great care and humility (there is no grand universal theory, but a careful attempt to attain a reliable empirical finding- however small that is).

My main concern with the paper is that the authors never clarify the relationship between Northern Rome’s lack of state capacity and the barbarian incursions. The main narrative maintains that the Germanic raids were the source of Roman political unrest (that is the way I summarized the argument in the preceding section). But at several instances across the paper, the authors hint that Roman political complications in Gaul were themselves a precursory factor that made the Germanic incursions more menacing.

The problem is present in both the econometric analysis and in the case studies. If I understood it correctly (by looking at figure 1), the regressions they asses rely on data that captures rainfall in both Roman Gaul/Germania and non-Roman Germania. The argument is that lack of rain affected Germanic independent tribes more because they were less prepared than the Roman borderline towns. Intuitively, this sounds right. However, the assertion does not imply that alternative transmission channels could not matter too. Yes, Roman towns were better prepared to endure bad harvests than their Germanic neighbors, but that doesn’t imply that bad agricultural output in Roman towns could not be the cause of political instability in them. There may be a relevant omitted variable bias problem in the empirical specification. [2]

The problem seems clearer when we consider the conclusions the authors get from their case studies: in them, they compare the level of relevance local border town problems in Germania/Gaul and in Syria had on larger Roman politics. The Germanics were a constant thorn on Rome, but the Syrians weren’t. Why? The authors explicitly stress that Roman Gaul/Germania had lower state capacity than Roman territories next to Syria, and so it was easier to subdue unrest in Syria than in Gaul. However, if that is so, then we are led to beg the question of what causes what? Is weak state capacity due to raiding, or is raiding due to weak state capacity? The paper’s narrative emphasizes the former linkage (all of the quantitative estimations rely on that sole mechanism too) while, at the same time, it recognizes that the latter mattered too. Unfortunately, it never sets to disentangle the underlying causality. [3]


Buengten, Ulf et al. (2011)  “2,500 years of European climate variability and human
susceptibility.” Science, 331(6017), pp. 578-582

Scarre, Chris (1995) Chronicle of the Roman Emperors: The Reign-by-Reign Record of the Rulers of Imperial Rome. Thames & Hudson: London.

Scheidel, Walter (2015) “Orbis: the Stanford geospatial network model of the Roman world.” http://orbis.stanford.edu/

Venning, Timothy. (2010) A Chronology of the Roman Empire, Bloomsbury Academic: New York.


[1] The authors confirm this by regressing rainfall at time t on lagged t -1 political stability.  It is interesting to note that this obvious observation may not be true for current events. Climate change is indeed affected by the domestic politics of some countries.

[2] I also remain confused about what data was used for some of the alternative estimations. For example, on the placebo test, they regress non-Germanic raids on precipitation levels. I assume they are using non-Germanic precipitation levels too. Otherwise, it would mean they would be testing how rain in Germania affects raids in non-Germania, which would make no sense.  However, they don’t clarify.

[3] My two cents on the Syrian/Gaul distinction is that geography and travel times may explain it. ORBIS (A project that reenacts the geospatial framework of the Roman Empire) allows us to estimate the times and cost of regular trips to different cities in the Roman Empire. A trip from Rome to Cologne would last 32 days on the fastest route and 63 days on the cheapest. A trip from Rome to Palmyra, on the other hand, would last 28 on the fastest route and 42 on the cheapest. This can provide a benchmark of the cost of mobilizing resources across regions: moving a Roman army could be 1/3 cheaper if it had to go to Syria rather than Germania. This significative figure implies that the costs of subduing unrest in Germania were larger and so more difficult.

This entry was posted in Analytic Narratives, Analytical Narratives, Economic History, Europe and tagged , , , , , on by .

About Fernando Arteaga

I am PhD Economics Student in George Mason University. My research interests lie in the fields of Economic History, New Institutional Economics and Agent-Based Modeling. I am also interested in banking and monetary economics; my undergraduate dissertation centered on the evolution of the debates over the "management of money" (free banking and other schemes). My graduate dissertation (which I'm working on right now) dwells on the institutional properties that create incentives for political union and fragmentation (why some nations are large and some small); I focus on the case of the Spanish Empire in the 16th-19th century (and the process of fragmentation that led to the creation of the Latin American countries)

3 thoughts on “Environmental Shocks and their Effects on Imperial Rome’s State Capacity

  1. Willem Jongman

    The problem is interesting, but not as new as the authors suggest. Climate change has been the subject of a few recent interpretations of Rome’s economic history, and its decline in particular. Apart from Kyle Harper’s recent book, see e.g. my ‘Re-constructing the Roman economy’. In L. Neal, & J. Williamson (Eds.), The Cambridge History of Capitalism, vol 1 (Vol. 1, pp. 75-100). Cambridge: Cambridge University Press.
    As the reviewer also points out, the issue is the mechanism by which climate change created social change. My argument was that lower agricultural output translated both directly into economic problems, and also indirectly through the outbreak of large epidemics, the first of which, the so called Antonine Plague from the mid 160’s onwards, may well have killed a third of the Empire’s population. This set in motion a chain of fateful developments like de-urbanization, the collapse of long distance trade, a reduction in standard of living of the mass of the population, the decline of civic culture, a growth in repression, and also political instability and debasement of the coinage (which had always been vulnerable to bad weather induced harvests, as I already argued some years ago). In a sense, the most remarkable thing was that the new Severan dynasty managed to delay the political collapse for another 70 years. After that, it became pretty dangerous indeed to be an emperor.
    In short, the story is rather richer, even if this particular slice of it is interesting.

  2. Harrison Searles

    I think your post entirely neglects the fact that Sasanid Persia, not the Germanic nations, were the ancestral enemy of the Roman Empire. They were a threat that diverted the strength of Mediterranean civilization, east, not northwards. In addition, I think you also neglect that when the “Barbarian” influx came, it came across the Danube, in the form of the Goths, not the Rhine.
    You write: “The Germanics were a constant thorn on Rome, but the Syrians weren’t. Why?”
    I think this is a terribly specified question that is motivated by modern political-geography, rather than the political-geography of the Roman empire.
    Syria, understood as part of a wider Mesopotamian theater, was a massive thorn in the Roman Empire. Here, we really have to resist looking at the Roman Empire from the point of view of our modern geography. Syria didn’t really exist in the form we think of it now, really it was a part of Mesopotamia, where Rome would be fighting against thee Persians from the time of Trajan to Justinian. Syria wasn’t a nation, so much as a borderland between the Romans and the Sasanid Persians, and it was fought over frequently.
    In comparison to Syria and Armenia, the Rhine river was relatively a docile front, and for good military reasons: It’s an incredibly difficult river to cross. Any garrison on either side can neutralize an army on the other side, hence the reason why Rome maintained such a stable frontier there across the centuries (really until the arrival of the Huns).
    Syria and Mesopotamia is where the action for most of the Roman Empire happened. For example, then Julian sought to win a great victory, he didn’t look towards the West, he looked towards the East, to the Sasanid Empire, and its capitol in Mesopotamia, Ctesiphon. It’s a bias in the way that we look at history, as modern descendants of Western Europeans, that we see the frontier along the Rhine river as more important as the frontier along the Danube river, Mesopotamia, or Armenia. However, when we look at Roman history, the Goths came across the Danube, the Romans fought the Sasanid Persians in Mesopotamia for centuries, and Armenia was also a theater of the Roman-Sasanid conflict. It’s not really until the invasion of the Huns that the Rhine becomes an active theater of war for the Romans.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s