By Kent R. Kroeger (January 17, 2019)
The observation that Middle East oil money and levels of worldwide terrorism are related is not new, though typically the inquiry is focused on how terrorism impacts oil markets.
In an earlier essay, I reviewed some of the news reporting and qualitative research on the role of Gulf State oil money and the funding of terrorist groups. In this essay, I take a more quantitative approach to understanding the relationship.
If it looks like a duck…
Someone merely needs to look at a time-series chart of oil price and terrorism-related deaths to see there is something going on between them (Figure 1). Do changes in oil prices cause changes in terrorism deaths? Is it the other way around? Or is the relationship non-recursive (i.e., goes both ways)? Or is there some confounding factor acting on both of them that makes them look causally related?
Figure 1: Islamic Terrorism-related Deaths and the Price of Oil, 2001 to 2018
The following essay documents a simple empirical test into that question using terrorism data from The University of Maryland’s Global Terrorism Database (GTD), “an open-source database including information on terrorist events around the world from 1970 through 2017.” As the GTD ends in 2017, I have supplemented the GTD with terrorism-related death data from TheReligionOfPeace.com (TROP) for 2018 numbers.
I filtered the GTD down to terrorism deaths occurring in Middle East or North African countries, in countries that are significantly engaged in military actions in the Middle East (U.S., U.K., Russia), and countries where there are a significant number of refugees from the Middle East (France, Germany).
Some may be concerned I’m using the TROP database, an unvetted, non-academic and most clearly biased data source. But in working with both sources, I found the two terrorism fatality databases to be highly correlated (Pearson’s r= 0.88) and, therefore, I am comfortable using both.
The data for the following analysis are available at NuQum.com (here).
A methodological note
In our everyday language, we tend to speak of things causally as if we knew the true relationship between them. We’ve all heard a workmate say: “I’m a(n) witch/a**hole until I get my morning coffee.” But it is quite possible, if we systematically measured the reality in a controlled experiment, once this person gets their coffee they just become a high-strung witch/a**hole.
Humans are wired to see patterns and relationships (even when they don’t exist). That is why merely inspecting a graph such as Figure 1 (above) is risky, even while it is instructive.
It is not clear whether changes in oil prices precede changes in terrorism deaths (or vice versa). Luckily, there are a number of statistical methods to make sense of this.
Nobel Prize Economist Clive Granger argued that the nature of causal relationships only can be asserted once we can show (1) the cause happens prior to the effect, and (2) the cause has unique information about future values of its effect (i.e., we can rule out other factors). This has become known as Granger causality.
Building on Granger, the statistical technique of vector autoregression (VAR) was developed in the 1990s and is commonly used to determine the linear interdependencies among time-series variables where little is known a priori about these relationships.
VAR has limitations. It fails to account for latent confounding effects and does not capture instantaneous or non-linear causal relationships. So, as with any statistical technique, it should only be used when the likely benefits outweigh the technique’s flaws.
In the case of the oil price and terrorism relationship, researchers are constrained by another serious problem. The data are relatively sparse.
The anecdotal evidence on how (Saudi/UAE, etc.) oil money ends up in the hands of terrorists suggests causal timelines are measured in years and multiple months, not days or weeks. An instantaneous relationship between oil prices and terrorism is improbable.
More likely, when oil prices are anticipated to rise/fall, futures contracts are bought/sold, proceeds are realized and distributed to terrorist organizations, and only then used to purchase weapons and pay salaries. Any subsequent operation funded by these proceeds may not occur for months after that. For example, the majority of the planning and financing for the 9/11 attacks occurred over a year before the actual attacks.
That means measurements at the daily or monthly level will likely be too noisy for any analytic value; yet, if we use annual data to assess relationships, we are left with no more than two or three decades of data (20 to 30 data points).
A middle-ground solution is to use quarterly oil price and terrorism deaths data. If the third quarter of 2001 is our starting point, that leaves us with 70 data points for analysis — a sufficient (though bare minimum) for employing Granger causality tests within a VAR analysis.
Occam’s Razor applies here
Contrary to the wisdom of 14th-century Franciscan friar William of Occam, the simplest explanation is not always the best explanation, but it is certainly the best starting place.
If oil prices cause terrorism in a substantive way (or vice versa), the relationship should reveal itself with minimal statistical effort. Nuanced or minor relationships are not interesting.
Figure 1 is the first indication that the oil/terrorism relationship is real and meaningful. But it tells us nothing about the direction of this relationship or even its relative effect size.
For that answer, we turn to Granger.
Figures 2 shows the results of a simple VAR testing the causal relationship between oil prices, as measured by the price for West Texas Intermediate Crude, and Islamic terrorism-related deaths, as measured by the GTD. Lagged values of 1-Quarter and 3-Quarters were used for both variables.
The table results (Granger causality Wald tests), while not overwhelming, are indicative of a significant relationship (p = .0382) where changes in oil prices predict changes in terrorism deaths, not the other way around (p = 0.3797). Apart from transitory shocks to oil prices when large-scale terrorist attacks occur (think 9/11), there is no evidence in the quarterly data (2001 to 2018) suggesting terrorism impacts oil prices.
Figure 2: Granger causality test (Oil Prices and Islamic Terror-related Deaths)
The significance level of oil prices (lagged one and three quarters) on terrorism deaths (p = 0.0382) does not blow the lid off my kettle. Still, I am encouraged by the model’s fit given that the GTD and TROP terrorism data are both probably over-counts of terrorism deaths connected to Gulf State-funded militant groups. Unfortunately, our ability is limited to discern funding sources for terrorist attacks spread across the globe from Germany to Yemen. We must, instead, accept the blunt measurements of the GTD and TROP and assume, if the relationships are strong, they will nonetheless bubble to the top in the empirical results.
More interesting is the impulse response function in Figure 3 which shows what a one-standard-deviation shock in oil prices (~$26 dollars per barrel) does to the number of terrorism deaths.
Initially, nothing happens. In the first two quarters after an oil price shock, there is no change in terrorism deaths. However, around the 3rd Quarter, terrorism deaths start to increase and peak around 500 additional deaths by the 8th Quarter and remain elevated thereafter. To put that in perspective, in the most recent Quarter (4th Quarter 2018), we experienced 2,700 terrorism-related deaths (most in war zones such as Syria, Iraq, Afghanistan and Pakistan).
Figure 3: Impulse Response Function between Oil Prices and Terrorism Deaths (Simple Model)
Five hundred additional terrorism deaths after a shock in oil prices seems plausible. But perhaps something else is confounding this relationship? Do changes in factors impacting oil prices affect its relationship to terrorism deaths?
Many additional factors were tested (e.g., world GDP growth, U.S. Dollar index, U.S. presidential approval), but two were consistently significant across the various models tested: (1) oil supply/demand ratio (OECD Oil Stocks divided by OECD oil consumption), and (2) the number of U.S. troops deployed to Afghanistan, Iraq and Syria. The supply/demand ratio data was obtained through the International Energy Agency. The U.S. troop deployment data for 2001 to 2007 was obtained through the U.S. Congressional Research Service. For deployment numbers in 2018, press reports were used.
Figure 4 shows the Granger causality tests for a model including oil prices, terrorism deaths, an oil supply/demand ratio, and the number of U.S. troops deployed to the Middle East. As in the simple model above, lagged values of 1-quarter and 3-quarters were used for all variables in the following full model.
The table in Figure 4 shows the significance of the causal relationship between oil prices and terrorism deaths (p = 0.0422) is virtually unchanged with the inclusion of the other two factors.
Figure 4: Granger causality tests (Oil Prices, Terrorism, Oil Supply/Demand and U.S. Troop deployments in the Middle East)
Again, the impulse response functions for this full model are most instructive (Figure 5). In particular, the first panel in the second row (Oil Price => Terrorism Deaths) shows the impact of a one standard-deviation in oil prices on terrorism deaths. As in the simple, two-variable model, a shock in oil prices eventually leads to about a 500-person increase in terrorism deaths by the 8th Quarter, and while this impact fades over time, it doesn’t completely disappear until around the 12th Quarter (3 years!).
Figure 5: Impulse Response Functions for Full Model
Additionally, one standard deviation shocks in oil prices and in the oil supply/demand ratio lead to lower levels of U.S. troop deployments in the Middle East (last panels in rows 2 and 3).
The full model also reveals that changes in the levels of U.S. troop deployments in the Middle East do not impact terrorism deaths. Once you control for oil prices (our proxy for terrorism funding), the presence of American troops does not seem to change the realities on the ground. Terrorism exists independent of American military involvement.
If you believe in the importance of U.S. military involvement in the Middle East, this finding should make your kettle boil over.
Of course, you could look at it from the other perspective and say, the presence of U.S. troops doesn’t make things markedly worse.
When you are irrelevant, anything can make you seem relevant
West Texas Intermediate (WTI) crude is selling for around $52-a-barrel as of 3:00 p.m. (January 17), up from about $42-a-barrel in mid-December 2018. If the model presented here is correct and the current WTI oil price trend is sustained, we should observe a noticeable increase in terrorism in around nine months (+250 deaths per quarter). If the current price trend reverses, however, then we should not witness a substantial increase in terrorism.
We will know soon enough.
What doesn’t seem to matter, in terms of terrorism, is whether or not the U.S. leaves, stays, or increases its military presence in Syria and other parts of the Middle East.
How can that be? Haven’t we seen the rollback of ISIS-held territory since the increase in U.S. troops in Syria just a few years ago? Only four years ago, ISIS controlled its own gas and oil wells.
But that is what is so confounding about confounding factors.
Proponents of U.S. military interventions ignore the increase in the Russian presence in Syria at the time ISIS’ stranglehold over nearly half of Syria began to erode. We also cannot ignore Iran’s role in assisting Bashar al Assad’s and the Iraqi government’s efforts to defeat ISIS.
Before anyone declare the U.S. as decisive in the fight against ISIS-related terrorism, they must reconcile that the U.S. military intervention in Syria hit its peak in 2016 and 2017 while the decrease in terrorism deaths had started almost two years prior. In truth, the initial drop in terrorism deaths in 2014 was preceded by a sharp dive in oil prices (see Figure 1 above).
Oil industry analyst Greg Depersio doesn’t cite the Syrian civil war and the fight against ISIS as significantly impacting oil prices in 2014. Instead, he singles out the slowing economic growth in China, India and Brazil after 2010. At the same time demand was softening, the U.S. and Canada were increasing their oil and gas production through processes such as fracking. Canada has the world’s third largest crude oil reserves and as these are further exploited, that will keep some downward pressure on oil prices into the future.
Depersio also credits Saudi Arabia with facilitating the 2014 fall in oil prices.
“Faced with a decision between letting prices continue to drop or ceding market share by cutting production in an effort to send prices upward again, the Middle Eastern country kept its production stable, deciding that low oil prices offered more of a long-term benefit than giving up market share,” notes Depersio.
I cannot find an industry analyst that suggests the last three-and-a-half years of relatively low oil prices have been the product of Iranian, U.S. or Russian military interventionism in Syria/Iraq and the decline of ISIS.
Before the civil war in 2011, Syria was a minor oil producer. They were 29th in 2007 at around 404,000 barrels-a-day. In 2007, that put them in the company of Equatorial Guinea and Yemen.
No statistical analysis, particularly based on 70 data points, is going be definitive on a subject as complex as terrorism. Terrorists are loyal to an ideology, not a business plan or balance sheet. They don’t always adhere to predictable patterns of behavior.
And even if we conclude that there is a demonstrable relationship between oil prices and terrorism, it comes with caveats. Fluctuations in oil prices are merely a proxy for a far more complicated and intertwined network of causal factors impacting worldwide levels of terrorism. This analysis tells us little about what is going on inside the black box.
But perhaps it is what I didn’t find in the data that is most intriguing. U.S. troop deployments, itself a proxy variable for broader U.S. efforts in the Middle East, do not causally relate to changes in terrorism.
The mere possibility that this could be true should give any American significant pause in asserting that an increase or decrease in the U.S. military presence in Syria, Iraq and Afghanistan will change the realities for the region. This data does support such a conclusion.
Please send comments and nonperishable food stuffs to: firstname.lastname@example.org