📈 economics

Quantifying the Axis Shift: Agile Statecraft, Systemic Hedging, and Spatial Alignment of Turkey

By applying Neural Additive Vector Autoregression to 22 years of data, this study concludes that Turkey's geopolitical axis shift is a structurally self-reinforcing process driven primarily by path-locked spatial alignment, with visit frequency and export dependency identified as the only statistically significant non-autoregressive causal factors at the dyadic level.

Original authors: Hakan Mehmetcik

Published 2026-07-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hakan Mehmetcik

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Turkey's "Dance" on the World Stage

Imagine the world as a giant dance floor. For decades, one group (the West) led the music, and most countries danced in step with them. But now, the music is changing, and the floor is getting crowded with new leaders.

Turkey is a skilled dancer in the middle of this floor. It's not just standing still; it's constantly shifting its feet, sometimes dancing with the old group, sometimes trying out moves with new partners, and sometimes spinning on its own.

This paper asks a specific question: How does Turkey actually decide where to stand? Is it just talking about moving, or do its actions (like visiting other countries or signing deals) actually cause it to shift its position?

To answer this, the researchers built a "Geometric Audit." Instead of guessing, they treated Turkey's foreign policy like a complex machine and used a special computer brain (called NAVAR) to figure out which buttons, when pressed, actually move the machine.

The Three "Levers" of Turkey's Strategy

The researchers broke Turkey's behavior down into three main "levers" or tools it uses to navigate the world:

The Voice (Discourse): This is what Turkish leaders say. Think of it as the "warm-up." It's low-cost and easy to change. If a leader gives a speech criticizing a neighbor, that's a signal.
The Feet (Physical Travel): This is when the President or Prime Minister actually visits another country. This is like showing up to the dance floor. It costs money, time, and political capital. It's a much stronger signal than just talking.
The Scorecard (UNGA Voting): This is how Turkey votes at the United Nations. It's the official record of where they stand. It's like the final score on a game sheet that locks in their position for the record books.

The theory was that Turkey uses these three tools in a chain: Talk first $\rightarrow$ Visit next $\rightarrow$ Vote last.

The Experiment: Two Different Ways of Looking

The researchers ran the experiment in two different "tracks" to see if the theory held up.

Track A: The Solo Performance (Monadic)

The Setup: They looked at Turkey as a single unit over 22 years (2002–2024). They asked: "Does talking more lead to visiting more? Does visiting more lead to voting differently?"
The Result: The data was too fuzzy to say for sure.
- The Analogy: Imagine trying to predict the weather by looking at only 22 days of data. You might see a pattern (like "it rained after the clouds gathered"), but you can't be 100% sure it wasn't just a coincidence.
- The Finding: The computer saw some hints that talking leads to visiting, and that trade patterns lead to visits. But because there wasn't enough data (only 22 years), the computer couldn't prove these links were real. It was "suggestive" but not "confirmed."

Track B: The Partner Dance (Dyadic)

The Setup: They looked at Turkey's relationship with 264 different countries over those same 22 years. They asked: "What actually changes how close Turkey stands to a specific country?"
The Result: Here, the computer found clear, proven links.
- The "Momentum" Effect: The biggest factor was simply habit. If Turkey was far away from a country last year, it's likely to be far away this year. The position is "path-locked," meaning it's hard to break out of the groove once you're in it.
- The Real Drivers: However, two things did successfully push Turkey closer or further away from a specific partner:
  1. Visits: The more often the President visits a country, the closer the two nations' positions become.
  2. Trade: The more Turkey relies on exporting goods to a specific country, the closer their positions align.

The Key Takeaways (In Plain English)

1. The "Axis Shift" isn't a sudden flip; it's a slow slide.
Turkey isn't suddenly jumping from one side of the room to the other. The study found that its position is "self-reinforcing." Once it starts drifting away from a partner (like a Western nation), it tends to keep drifting unless something strong pushes it back.

2. Talking isn't enough; you have to show up.
While leaders talk a lot, the study found that visiting is what actually moves the needle. If the President goes to a country, the relationship physically shifts.

3. Money talks, but it's a quiet voice.
How much Turkey sells to a country (export dependency) also changes the relationship, but it's a smaller effect than the visits.

4. The "EU" Factor.
Interestingly, being in the European Union actually made countries vote less like Turkey in the UN. The study suggests that EU membership acts as a "brake" on how much Turkey and EU countries align their votes, even if they are close partners.

What the Paper Doesn't Say

It is important to note what this study did not find:

It did not find proof that Turkey's speeches alone caused it to change its voting or travel habits (at least not with the data they had).
It did not prove that Turkey is "anti-Western" or "pro-Eastern." It simply measured the distance between Turkey and others, regardless of who "others" are.
It did not predict the future. It only analyzed the past 22 years to see how the machine worked.

The Bottom Line

Think of Turkey's foreign policy like a heavy ship. The ship has a lot of momentum (it keeps going in the direction it's already heading). To change course, you can't just shout from the bridge (discourse). You need to turn the rudder (visits) and adjust the engine (trade). The study confirms that for Turkey, visits and trade are the rudder and engine, but the ship is so big that it takes a lot of effort to turn it.

Technical Summary: Quantifying the Axis Shift: Agile Statecraft, Systemic Hedging, and Spatial Alignment of Turkey

Problem Statement and Theoretical Gap
This study addresses the challenge of modeling Turkey's foreign policy trajectory within a structurally contested, post-unipolar international order. Existing literature typically categorizes Turkey's alignment through three distinct traditions: the normative-institutional approach (viewing divergence as a byproduct of autocratization), the hedging literature (framing mixed strategies as rational equilibrium), and the computational turn (applying NLP and spatial scaling to diplomatic data). However, these approaches share a critical limitation: none produces a unified causal architecture connecting discrete behavioral inputs to spatial alignment outcomes. The paper posits that Turkey operates not merely as a regime-type case but as a rational, utility-maximizing node within a multidimensional alignment space. The central research question is how discrete behavioral signals—specifically discourse, head-of-state travel, and institutional co-alignment—causally alter Turkey's spatial alignment relative to a Western hegemonic baseline.

Methodology: Neural Additive Vector Autoregression (NAVAR)
To resolve the methodological boundary of traditional linear frameworks (which impose symmetric proportionality and fail to capture the non-linear, state-dependent nature of "agile statecraft"), the study employs Neural Additive Vector Autoregression (NAVAR). This causal discovery architecture extends the Granger causality framework by decomposing target predictions into sums of source-specific learned functions, allowing for the isolation of distinct causal pathways without assuming a global linear functional form.

The research design utilizes a dual-track approach over a 22-year period (2002–2024):

Track A (Monadic Behavioral Interdependence): Analyzes Turkey as a unitary actor ( $T=22$ ). It models the internal causal feedback between three behavioral pillars: Discourse (sentiment-scored Turkish Foreign Policy Documents via XLM-RoBERTa), Physical Travel (Head-of-State itineraries from the COLT dataset), and Institutional Positioning (UNGA ideal points).
Track B (Dyadic Hedging Optimization): Analyzes bilateral relationships across a pooled panel of approximately 5,800 observations ( $264 \text{ partners} \times 22 \text{ years}$ ). It maps the non-linear causal effects of the three behavioral pillars on two spatial alignment outcomes: the absolute ideal-point distance from the Western baseline ( $abs\_ideal\_diff$ ) and bilateral UNGA voting coincidence ( $unga\_agree\_score$ ).

Data and Statistical Rigor
The study integrates diverse data sources, including the COLT dataset for travel, the TFPD corpus for discourse, Bailey et al.'s (2017) Bayesian IRT estimates for UNGA ideal points, World Bank WITS data for trade flows, and OECD migration statistics. Crucially, the analysis employs a rigorous statistical protocol to address the multiple-comparison problem inherent in testing hundreds of causal edges. The study applies the Benjamini-Hochberg (BH) False Discovery Rate (FDR) correction across 280 edges in Track A and 82 edges in Track B. Significance is determined via permutation bootstrap testing (2,000 iterations for Track A; 500 for Track B), with edges considered significant only if they survive the BH-adjusted $q < 0.05$ threshold.

Key Results
The findings revise the "agile statecraft" account by distinguishing between suggestive inter-pillar dynamics and confirmed dyadic drivers:

Monadic Level (Track A): At the level of the sovereign state, zero of the 280 simultaneously tested causal edges survive BH-FDR correction. While 12 edges pass the unadjusted permutation threshold ( $p < 0.05$ )—including suggestive cascades where NSC discourse precedes head-of-state visits ( $G=0.165$ ) and trade concentration drives agreement signings—none are statistically robust under joint correction. The authors attribute this to statistical underpowering given the limited temporal window ( $T=22$ ) and the strictness of the FDR correction.
Dyadic Level (Track B): In contrast, the dyadic analysis reveals a "spatial path-lock" structure. Seven of 82 edges survive BH correction.
- Path-Lock Dominance: The primary driver of alignment is autoregressive; the lag-1 value of ideal-point distance ( $abs\_ideal\_diff\_lag1$ ) is the dominant causal factor ( $G=0.420$ ), indicating that once a positional offset is established, it is self-sustaining.
- Confirmed Behavioral Antecedents: Within this path-locked structure, two non-autoregressive channels are confirmed as causal drivers of ideal-point distance: visit frequency ( $colt\_n\_visits$ , $G=0.051$ ) and export dependency ( $trade\_export\_share$ , $G=0.031$ ).
- Voting Coincidence: For UNGA voting alignment, the outcome is driven by autoregressive lags (1, 2, and 3) and a structural suppressor effect where EU membership ( $partner\_is\_eu$ ) significantly reduces voting coincidence ( $G=0.036$ ). No behavioral inputs (discourse or travel) achieve significance for this specific metric.

Significance and Claims
The paper claims to provide a "geometric audit" of foreign policy that moves beyond qualitative process tracing or linear modeling. Its primary contribution is the empirical demonstration that Turkey's "axis shift" is best understood as a structurally self-reinforcing process (path-lock) rather than a series of discrete, reversible maneuvers.

The study modestly concludes that while the internal causal mechanics of Turkey's monadic behavioral apparatus (how rhetoric translates to travel to voting) remain statistically inconclusive at FDR-corrected thresholds with current data, the dyadic alignment trajectory is robustly driven by structural inertia. However, the identification of visit frequency and export dependency as confirmed, albeit modest, causal drivers suggests that the axis shift is not entirely immune to behavioral intervention. Reversing this drift would require sustained bilateral engagement and trade restructuring over multiple periods to overcome the dominant autoregressive momentum, rather than single diplomatic intensifications. The framework offers a replicable, computational architecture for quantifying strategic autonomy and hedging in other middle powers navigating systemic transitions.