Tuesday, February 11, 2014

Borrowing Analytic Techniques: Populations, Predictions And What Physics Tells Us About The Movement Of Alawites (Part 3 Of A Multi-Part Series)

Note: This is the third part in a multi-part series covering an extensive project on the human geography of Turkey and network analysis applications within the field of Intelligence Studies. Be sure to read the first two posts, My Conversation With The Free Syrian Army and Turkey Redrawn.


World Migration into Turkey
Source: International Organization for Migration Map
In the past, SAM has covered many different Structured Analytic Techniques (SATs), the zooming technique and Structured Role Playing among them. This post covers an analytic technique somewhat unfamiliar to the intelligence discipline, but extremely useful, especially within the realm of human geography and population movement. 

Using this technique, I was able to make predictions that read something like this:

"The ethnolinguistic populations within Turkey that will likely expand within the next 12 to 24 months are Bulgarians, Serbs, Iraqi Kurds, Iranian Arabs and Azerbaijanis."
and...
"The groups that are highly likely to impact Turkey within the next 12 to 24 months are Syrians and Kurds (from Syria, Iraq and Iran)."
Where did these predictions come from?

Gravity models.


I know what you're thinking, and yes, gravity models do have to do with gravity and yes, gravity models do happily reside within the domain of physics. 


Now, before you close the page, I am not going to spend the next four hundred or so words discussing Newton's Law of Universal Gravitation. Instead, on a one-stop flight from physics to intel, we will have a brief layover in the discipline of economics.  


I said, don't close the page!

Back in the 1950s, economists began to import gravity models into economics to predict trade flows and it looked something like this:



F_{ij} = G \frac{M_i M_j}{D_{ij}}

Where F is the flow of bilateral trade between country Mi and country Mj (measured in economic size). D is the distance between these two countries in kilometers squared and G is a constant. If it seems familiar, it should.  It is Newton's Law with the names of the variables changed.

In the 1970s, a man by the name of Peter Trudgill, a well-respected British sociolinguist, brought the gravity models to dialect geography. Trudgill, whose most recent book was published in 2011, holds many firsts within the field of linguistics, of which gravity models are one. He enacted a series of alterations to the gravity model in order to account for linguistic factors, the end goal being to create gradient zones of gradual shift from one phonetic marker to the next within geographic space (like geospatially representing the line in and around Boston that marks where people stop saying cah and start saying car). Trudgill's gravity model looked like this:


Trudgill's modifications to the original gravity model accomplish two goals:
  1. By adding the s variable measured on a scale of one to four, he factors in the well-known assumption in sociolinguistics that contact occurs more readily between groups that have "prior-existing linguistic similarity." This means that if the languages spoken by Population i and Population j are mutually intelligible (such as American English and British English), the s variable would be higher than for, say, American English and Italian). 
  2. By adding the second half of the equation, he gives the model directionality. The original gravity model was intended to predict bilateral trade, meaning it would predict the economic contact between one country and another. Trudgill's model predicts the influence country Pi will have on country Pj as opposed to the bilateral contact country Pand country Pj  will have.
Now, what does all this have to do with intel?

Within the context of my most recent project, mapping the Syrian refugee population in Turkey, gravity models provided a way to analyze the ethnolinguistic landscape in Turkey through a predictive lens. Instead of saying, "Here are where Armenians live in Turkey," I can say, "Here are where Armenians live and this is how likely there are to be more Armenians traveling to this region from Armenia in the next 24 months." 

In other words, it took the ethnolinguistic mapping project from descriptive to predictive

In order to achieve this, I added one more variable to Trudgill's altered gravity model: the c variable.


The c variable is a variable assessed on a scale of one to four that takes into account the trend line of migratory patterns from target ethnolinguistic groups over the past 10 years (data taken from the International Organization for Migration). In other words, to the degree that future migratory patterns follow historical trends for each ethnolinguistic group, the c variable takes these trends into consideration. 

What results from these models is an index score (usually a decimal) that is indicative of how likely or unlikely contact is to occur (or, in the case of some of the most advanced equations, how likely one region is to influence another region). This has the potential to  make a compelling predictive argument when translated onto a map (See Figure 1).
Figure 1. Results of gravity model analysis represented geospatially.
Green areas: Ethnolinguistic areas that are highly likely to expand in the next 12 to 24 months
Black areas: Ethnolinguistic areas that are likely to expand in the next 12 to 24 months
Click here for a map of all the ethnolinguistic areas of Turkey
These models have widespread application for predicting contact between any kinds of populations, but arguably the circumstance in which gravity models work best are when a generally homogeneous geographic region hosts pockets of ethnic diversity. Such is the case with Turkey, therefore gravity models provided a solid predictive approach. 

Gravity models were the backbone of my ethnolinguistic predictions regarding Turkey. It is the output of these models that got me to thinking about the incoming Syrian refugee population which, at the time, numbered close to 700,000 and today likely exceeds 1 million. 

With that in mind, don't miss the next post (by far the most interesting): Syrian Refugee Population Simulation: From *ORA to Istanbul.

5 comments:

Joseph Moynihan said...

That is a fascinating study. Though I am not a lover of equations and physics, I found that piece, and the supporting concepts, very interesting and a very logical approach to understanding/predicting population movements. Thank you for sharing.
Joe Moynihan, Dungarvan, Ireland.

J said...

Look at this one:
http://www.nature.com/nature/journal/v484/n7392/full/nature10856.html

Melonie Richey said...

J -

Thanks so much! This is exactly the kind of resource I am looking for as I start a new project using the same methods for a different region.

Anonymous said...

Facinating approach!. I wonder if it is possible to develop a predictive model for the spread of (offensive) technological capabilities between radical militant Islamic groups?

Melonie Richey said...

That is a very interesting concept. The next post will deal with the simulation part of the project (simulating refugee flows, which also has a predictive element). I used an information simulation to do that, there but there is a technology simulation feature that might have merit for the kind of project you are talking about.