Chapter VIII. Science as Structural Empiricism
- john raymond
- Aug 25
- 3 min read
Analysts of President Trump, Vladimir Putin, the war in Ukraine, and the wider geopolitical field come from mixed backgrounds—history and law, philosophy and psychology, domestic politics, some game theory.
Many have little formal exposure to statistics. That gap is not cosmetic; it is the difference between commentary and science.
The task is not, however, to turn every analyst into a mathematician. It is to show that rigorous analysis is scientific because it is structural empiricism: data tied to mechanism, usually tested against null randomness, and scored by predictive accuracy.
The Main Formula of Science
Data is the empiricist’s tool. Mechanism is the structuralist’s tool. Together they constitute structural empiricism. Together they are modern science.
What Counts as Data
In geopolitics, data are the observable outputs of power: budget lines and force posture; sanctions, exemptions, and enforcement actions; logistics flows and depletion rates; alliance statements and votes; information operations cadence and reach; appointments, firings, back channels, and lawfare filings; battlefield outcomes and their temporal sequence.
One-off anecdotes are not data. Time series are.
What Counts as Mechanism
Mechanism explains why the data take the form they do: incentives, constraints, capabilities, and channels. Who benefits in regime-security terms; which asymmetric tools are cheapest and most available; which insiders translate intent into institutional output.
A mechanism is not a slogan; mechanism is a causal chain you can write down and test.
The Scientific Standard in Strategy:
Collect data across time and identify correlations Study processes, not snapshots. Build event logs and time series. Align observations to decision points: before, during, after. Compute simple, honest summaries: levels, changes, rates, lags. Correlations are descriptive, not dispositive; they tell you where to look harder.
Formulate mechanisms that explain incentives and capacity Propose channels that plausibly generate the pattern: money, manpower, media, lawfare, logistics. Map actors to levers and levers to effects. If you cannot say how a claim could be true, treat it as unripe.
Test against null randomness (usually) State the null explicitly: what would we expect to see if there were no alignment, no effect, no mechanism? Compare observed direction and frequency at leverage points against that baseline. If your pattern is what randomness would often produce, you have not learned anything.
Accept or modify the model based on predictive accuracy A model earns belief by risking failure. Write dated, measurable forecasts. Score them on arrival. Update the model or discard it. Calibration and accuracy are the only court of appeal.
Minimum Viable Math for Analysts
You do not need specialized notation to practice science, but you must respect its demands.
Baselines and deltas. Always compare to a baseline. Report both level and change. Separate signal from static noise.
Event studies. Define a window around a decision, track the series, and compare to matched controls. Do not cherry-pick the window ex post.
Forecast registration. Pre-state near-term predictions with dates, metrics, and what would falsify them. Keep a public ledger.
Scoring and calibration. Track forecast quality using simple scoring rules and a calibration chart. If your seventy-percent calls land only half the time, you are overconfident and must adjust.
Common Failure Modes
Selection bias and survivorship bias; narrative fallacy; look-ahead bias; overfitting headlines; game-theory theater that ignores measured constraints and channels.
A Worked Micro-Template
Question: did a sanctions step materially degrade an adversary’s operational capacity within sixty days?
H₀: no measurable degradation beyond normal variance.
H₁: measurable degradation consistent with the specified mechanism.
Data: an operational proxy over time (munitions-expenditure gap, spares availability, black-market price indices, exchange-rate-adjusted import flows), with controls.
Mechanism: identify enforcement pathway and chokepoints; specify why this step should bind where prior steps did not.
Test: pre-register a window and a threshold effect; compare to matched controls; report uncertainty.
Decision: accept, modify, or reject the mechanism; update the next forecast accordingly.
The Power Equation
Harm over time is the operative metric of modern war. Measure what policies and operations enable in cumulative effect, not what their authors claim. A move that looks small at t₀ may unlock compounding harm by t₁, t₂, t₃.
Structure your data and mechanisms to detect compounding.
Division of Labor
If you cannot do the math yourself, pair with someone who can, or at least adopt the minimum kit above with discipline. What you cannot do is pretend that science is optional.
Without baselines, nulls, and scoring, you will mistake noise for pattern and propaganda for truth.
Generating Real Results
When analysts adopt structural empiricism, they become scientists of geopolitics. They stop narrating and start explaining. They stop reacting and start predicting. They expose gambits before they land because their models bind data to mechanism and face the calendar.
That is how the web of truth is woven tight enough to resist industrialized deceit.
Comments