How the same institution wrote the doctrine, supplied the personnel, and graded its own homework.

That definition has survived every outcome it helped produce.

The institution that writes the strategy also evaluates the outcome. There is no external review requirement. There is no mandatory disclosure. This is legal, continuous, and seven decades old.

Most people have never heard of RAND Corporation. That absence is part of the mechanism.

January 2019. Washington. A report appeared with a title so dry it nearly disappeared into the paperwork: Overextending and Unbalancing Russia. Commissioned by the U.S. Army. Produced by RAND.

Three years later, Ukraine was at war.

RAND is a nonprofit research organization founded in 1948 with U.S. Air Force funding. Between 2014 and 2019, it received $1.029 billion from American defense sources 95.5 percent of all defense think tank funding in the United States. The client never changed.

RAND does not make weapons. It produces research, doctrine, frameworks, threat assessments then supplies the people who carry those frameworks into government. When the policies those people design require evaluation, RAND produces the evaluation. The assumption underlying the arrangement was that better analysis would produce better outcomes. The problem with that assumption only became visible when the same institution generating the data also assessed whether the outcomes were good.

Whoever controls the framework does not make decisions. It decides what decisions are possible.

Vietnam

In 1961, Robert McNamara arrived at the Pentagon with a specific problem. The American military was large, expensive, and, in his assessment, unmeasurable. He had run Ford Motor Company on data. He believed the Defense Department could be run the same way. He recruited a team to build the measurement system. He went directly to RAND.

Charles Hitch, RAND’s chief economist, became Pentagon Comptroller. Alain Enthoven, a RAND systems analyst, ran the office of Systems Analysis that would reshape how the military evaluated its own performance. They brought with them the methodology RAND had developed over the previous decade: systems analysis. The reduction of complex strategic problems to quantifiable metrics. The assumption that what could be measured could be managed, and what could not be measured did not need to be.

The methodology had worked for comparing the cost-effectiveness of weapons systems. McNamara applied it to counterinsurgency.

Body counts. Kill ratios. Sortie rates. Progress in Vietnam would be measured the way cost-effectiveness in defense contracts was measured: with numbers that could be tracked, compared, and reported upward. Units filed reports because reports were being requested. Reports showed progress because progress was what the measurement system was designed to detect. The war produced data. The data confirmed the strategy. The strategy continued.

The metric had become the objective. Commanders optimized for the number being measured, not for the outcome the number was supposed to represent. A system designed at RAND to bring rigor to defense planning had, when applied to a counterinsurgency in Southeast Asia, created the conditions for a war that could be shown to be going well while going badly.

RAND did not send soldiers to Vietnam. It sent the method that made the war continue.

By 1967, McNamara knew something was wrong. The numbers showed progress. The war showed no progress. He commissioned a classified historical study of how the United States had arrived at this point. RAND assembled much of the research team. One of the analysts brought onto the project was Daniel Ellsberg.

Ellsberg had spent a decade at RAND. He had a doctorate in economics from Harvard, had worked on nuclear war planning, and had spent time in Vietnam as a civilian attached to the State Department. He had watched the gap between the official narrative and the situation on the ground widen each year. He returned convinced that the war could not be won as described. He spent two years trying to bring that conclusion to Congress through official channels. Every channel closed.

The team produced 7,000 pages across 47 volumes. It documented that four successive administrations had systematically misled Congress and the American public about the origins of the war, its prospects, and the assessments their own officials had made in private. McNamara had commissioned the study. He did not share it with the White House. He did not share it with the State Department. He filed it. Fifteen copies were made.

In 1971, Ellsberg leaked the study to the New York Times. The Pentagon Papers.

The system recorded its own deception.

RAND built the doctrine that shaped the war’s conduct. A RAND analyst produced the study that documented the war’s deceptions. The methodology that extended the war and the researcher who exposed the lies behind it both came from the same institution, at the same address, funded by the same government.

The method did not fail. The war did.

Iraq

Donald Rumsfeld served on the RAND board of trustees for more than two decades, including as chairman. Condoleezza Rice was a trustee from 1991 to 1997. Both left to fill the two most senior national security positions in the U.S. government. The Secretary of Defense and the National Security Advisor had spent years inside the institution that produced the strategic frameworks through which they would now make decisions.

That proximity is not evidence of coordination. It is evidence of something more durable: a shared intellectual environment. RAND had spent the 1990s building the analytical vocabulary through which the post-Cold War security landscape was interpreted in Washington. It had produced threat assessments of regional instability in the Middle East. It had helped develop the concept of the Revolution in Military Affairs, the doctrine that a technologically superior force could achieve rapid, decisive victory with reduced troop numbers. That doctrine shaped how the invasion of Iraq was planned.

The people who made the decision to invade had not been told what to conclude. They had spent years inside an institution that determined which risks warranted serious responses and which could be dismissed, which strategic frameworks were rigorous and which were not, which questions were worth asking. By the time the intelligence assessments arrived in 2002 and 2003, the conceptual architecture through which those assessments would be read was already in place.

This is the mechanism. Not instruction. Preconditioning.

The invasion proceeded. The predicted rapid victory did not materialize. The post-conflict reconstruction failed. The counterinsurgency dragged on for years. At each stage, RAND produced the evaluations. When the planning failed, RAND analyzed why. When the de-Baathification policy collapsed the Iraqi state, RAND assessed the effects. When the surge was proposed, RAND analyzed its prospects. The institution that had contributed the intellectual frameworks within which the war became thinkable also produced the reports on what had gone wrong.

No external body reviewed whether the frameworks had been part of the problem.

The Revolution in Military Affairs doctrine, the intellectual foundation of the Iraq plan, was a RAND product. The post-invasion assessments of why the doctrine failed were also RAND products. The recommendations for what to do next were also RAND products. At no point in that sequence did an institution without a financial relationship to the Department of Defense provide the primary analysis. The evaluator and the evaluated shared an address.

The framework did not change between administrations. The personnel did.

Ukraine

In January 2019, RAND published Overextending and Unbalancing Russia. The client was the U.S. Army. The document was not classified. It was publicly available, indexed on RAND’s website, and read almost exclusively by people already inside the policy apparatus.

The report identified eight categories of measures that could strain Russian resources and exploit Russian vulnerabilities. One of them was the provision of lethal aid to Ukraine. The report’s executive summary described this option as capable of exploiting what it called Russia’s greatest point of external vulnerability. Those were the document’s own words, written for the client that paid for the analysis.

The report did not recommend this option without qualification. It assessed the likely Russian response: providing lethal aid to Ukraine, it noted, would risk provoking a significant Russian military response. The report named that risk explicitly. It evaluated it. It presented the option anyway as a viable instrument of strategic pressure, with costs the authors considered manageable from the American side.

The analysis was delivered to the U.S. Army in January 2019. Three years later, the United States was providing lethal aid to Ukraine. Russia had invaded. The escalation the report had described as the likely Russian response had occurred.

In January 2023, RAND published Avoiding a Long War: U.S. Policy and the Trajectory of the Russia-Ukraine Conflict. The report analyzed the costs and risks of extended conflict and recommended how the United States should calibrate its involvement. Later that year, RAND published additional analysis on Ukrainian war aims and the prospects for a negotiated settlement.

The institution that wrote the strategy evaluated whether the strategy worked. No step in that sequence required leaving the building.

The system did not expand. It repeated.

No external correction entered the system.

What this means concretely: a single institution identified the vulnerability, designed the pressure strategy, supplied the analytical vocabulary through which policymakers understood the options, and then, once the war began, became the primary source of guidance on how to end it.

The 2019 report and the 2023 report were produced by the same organization for the same client. The question of whether the 2019 analysis contributed to the conditions that made the 2023 analysis necessary was never formally asked. There is no institutional mechanism that would require it to be.

The Mechanism

This pattern does not require coordination. It requires only that the same institution supply the frameworks, the personnel, and the evaluation, and that no one is required to disclose the connection.

RAND’s budget depends on defense funding. Defense funding depends on threat assessments. An institution funded by defense budgets does not need to be told to identify threats. The incentive structure identifies them automatically. An institution that produces reassuring assessments loses relevance and eventually contracts. An institution that identifies serious threats, requiring serious responses, and necessitating continued serious investment in analysis, grows. This is not a description of corruption. It is a description of how organizations respond to the incentive structures they operate within.

That is not a conspiracy. It is arithmetic.

The architecture has one additional feature. Because RAND is a nonprofit research organization, not a contractor or a lobbying firm, its output carries the designation of independent analysis.

That designation provides access that commercial defense contractors do not have. It allows RAND researchers to testify before Congress as neutral experts rather than as advocates. It allows RAND reports to be cited in policy documents without triggering disclosure requirements. The financial relationship between RAND and the Department of Defense does not appear on the cover of the reports the Department uses to make decisions.

This is not a recent development. It has been the structural condition since 1948. What changes is only the domain.

RAND is currently the primary institutional voice on artificial intelligence policy in Washington. Between 2017 and 2024, RAND published more than two hundred research products on artificial intelligence, covering governance, national security applications, autonomous weapons systems, and AI risk assessment.

It has testified before Senate committees on AI governance, safety, and regulation. It has advised the Department of Defense on AI ethics frameworks and the operational doctrine governing autonomous systems. It has produced the threat assessments that determine which AI capabilities are designated as existential risks requiring regulatory response and which are not.

The primary funder of that research is the same institution it has always been: the Department of Defense.

The institution that defines which AI capabilities represent existential risks also advises on how those risks should be regulated. The institution that shapes the threat assessment also shapes the policy response the threat assessment is used to justify. The institution that designed the AI ethics framework for autonomous weapons will assess whether the framework worked. When the AI strategy requires evaluation, RAND will produce the evaluation.

The architecture is identical to 1948. The Air Force was the first client. The domain was nuclear weapons. Today the client is still defense. The domain is artificial intelligence. Seventy-six years of institutional structure, unchanged.

There is no stage in that sequence where an external party enters. No independent body reviews whether the frameworks are producing good outcomes. No disclosure requirement surfaces the financial relationship between the analyst and the client. The loop is closed before the public learns what question was being asked.

RAND does not win wars. It defines what winning means.

That definition has survived every outcome it helped produce.

Jerry writes forensic institutional analysis at The Manifest Archive. This article examines the structural relationship between strategic think tanks, defense funding, and policy outcomes.

Related from The Manifest Archive