Strip away the conspiracy and a documented structure remains: the part of government that holds power does not appear on any ballot, and a serious scholarly tradition has a name for it.
In 2008 a candidate ran for president of the United States promising to close the prison at Guantanamo Bay, end warrantless surveillance, and rein in the secret machinery the previous administration had built. He won decisively, served eight years, and left office with Guantanamo still open, the surveillance programs not only intact but expanded, and the drone-strike apparatus far larger than he had inherited. His successor, from the opposite party and temperament, campaigned against the same security establishment from the other direction and likewise found that much of it continued regardless. Two administrations, two parties, two mandates, and on a whole class of questions the policy barely moved. The faces on the ballot changed completely. The thing the ballot was supposed to control did not.
That pattern is the documented core of a phrase that has otherwise been buried under conspiracy: the deep state. Used loosely, the term has come to mean a secret cabal that rigs elections and runs the world from the shadows, and in that form it deserves the ridicule it gets. But underneath the lurid version is a structural fact that serious scholars have described carefully and without melodrama, and the fact is more unsettling than the cabal precisely because it requires no secrecy and no plot. The part of a modern government that actually exercises continuous power is largely permanent and unelected, and it does not change when the voters change their minds. That is not a theory about hidden men. It is an observation about the architecture of the state, and it is the only version of the deep state worth taking seriously.
The word and where it came from
The phrase has a specific and documented origin, and recovering it is the first step in separating the real phenomenon from the myth. It is a translation of the Turkish derin devlet, which entered common use in the 1990s to describe something concrete and genuinely sinister: a network within the Turkish state, drawn from the military, the intelligence services, and allied criminal and ultra-nationalist groups, that operated beneath the elected government and waged a covert dirty war against Kurdish insurgents and political enemies. It surfaced in public view through the Susurluk scandal of 1996, when a car crash near that town exposed four people who had been travelling together in a single vehicle, a combination that made no innocent sense: a senior police official, a member of parliament, a wanted ultra-nationalist hitman carrying a government weapons permit, and the hitman's companion. Three of them died in the crash; the member of parliament survived to be questioned. In one wreck the hidden seam where the legal state met an illegal parallel structure was laid open on a public road.
That was a real deep state, documented and investigated, and it matters here for two reasons. First, it shows the phenomenon is not imaginary; parallel structures that operate beneath elected governments have demonstrably existed. Second, it shows how much is lost when the term crosses borders. What described a specific criminal-security network in Turkey became, in American usage, a vague accusation thrown at any part of the government an elected official dislikes, and in that translation it lost its precision and gained its paranoia. To use it seriously again, it has to be returned to something definable, and there is a body of scholarship that does exactly that.
Double government
The most rigorous version of the idea is not found in conspiracy literature at all but in a 2014 book by Michael Glennon, a professor of international law at Tufts and a former legal counsel to the Senate Foreign Relations Committee. Its title is National Security and Double Government, and its argument is careful, sourced, and entirely free of secret cabals. Glennon borrowed his framework from the Victorian essayist Walter Bagehot, who in 1867 described the English constitution as a double government. There were, Bagehot wrote, the dignified institutions, the monarchy and the lords, which commanded public reverence and which most people believed ran the country, and there were the efficient institutions, the cabinet and the bureaucracy, which actually did. The dignified parts existed to secure the loyalty of the governed. The efficient parts existed to govern. The genius of the arrangement was that the public watched the dignified half and was reassured, while the efficient half did the work largely unobserved. Bagehot, it should be said, approved. For him the double government was not a scandal but the secret of English stability, a way of keeping the loyalty of a public that would never be moved by the dry machinery of administration while freeing that machinery to govern in peace. The unsettling step is Glennon's, which is to ask what it means for a republic founded on the principle that the governed control their government to have quietly grown a double structure of its own, in which the half the public can control is increasingly the ceremonial one.
Glennon's claim is that the modern United States has quietly developed the same double structure. The Madisonian institutions, the presidency, the Congress, and the courts, are the dignified part: visible, elected or appointed through visible processes, and believed by the public to be in charge. But on the vast and consequential terrain of national security, he argues, real policy is made by what he calls the Trumanite network, the several hundred senior officials of the military, the intelligence agencies, and the security bureaucracy who possess the expertise, the information, the continuity, and the institutional momentum that elected officials lack. The Madisonian institutions provide legitimacy and the appearance of control; the Trumanite network provides the actual decisions, which the elected officials then ratify and present as their own. No one needs to conspire. The structure produces the result on its own, because a president serves four or eight years and a security bureaucracy serves forever, and forever beats four years in any contest of will and attrition.
The apparatus that does not leave
Glennon's network did not arise by accident, and its origin is the most documented part of the story. The modern American security state was built in a single concentrated burst after the Second World War, above all by the National Security Act of 1947, which created the Central Intelligence Agency, the National Security Council, and the unified defense establishment in one stroke. The wartime intelligence apparatus, which might have been dissolved when the war ended, was instead made permanent and expanded, and the institutional culture that came with it, secrecy, autonomy, and a sense of standing above ordinary politics, became permanent too. As early as 1949 a government commission was warning that the Joint Chiefs of Staff had become virtually a law unto themselves, and in 1961 a departing president who had commanded the Allied armies in Europe, Dwight Eisenhower, used his farewell address to warn the nation against the unwarranted influence of a military-industrial complex whose total influence, he said, was felt in every city and every statehouse. Two years later the man who had actually created the central intelligence apparatus said much the same. In a column for the Washington Post on 22 December 1963, a month after the Kennedy assassination, former President Harry Truman wrote that he was disturbed by how far the CIA had been diverted from its original assignment, that the agency he had established in 1947 to coordinate intelligence had become an operational and at times a policy-making arm of the government. The two men most responsible for building the postwar security state, the general and the president, both lived long enough to warn that it had outgrown the control of the people who built it.
The point is not that these institutions are malign. It is that they are permanent, and permanence is itself a form of power that elected office cannot match. A career officer in an intelligence agency may serve through eight or nine administrations. The agency's institutional memory, its relationships, its ongoing operations, and its accumulated expertise all persist across every election, while the political appointees who nominally direct it arrive knowing little, stay briefly, and leave. Over time the permanent layer learns that it can wait out any transient occupant, and the transient occupant learns that fighting the permanent layer costs more than it is usually worth. None of this is written down as policy. It is simply the physics of an organization in which one part is replaced every few years by election and the other part is never replaced at all. The appointee arrives knowing little. The appointee stays a few years. The appointee leaves. The institution does none of these things, and the institution remembers.
The same policy, every administration
The clearest test of who actually controls a domain is to watch what happens when control formally changes hands, and on national security the answer has been strikingly consistent. The mass-surveillance apparatus that Edward Snowden exposed in 2013, the bulk collection of communications records and the programs reaching deep into the world's data, had been built under one administration and was running, expanded, under its political opposite, an administration that had campaigned in part against exactly that kind of secrecy. The legal authorities beneath it, the surveillance statutes enacted after 2001 and their successors, have been renewed again and again by both parties, through every change of president and Congress. Wars opened under one administration were continued under the next. Detention and drone programs denounced from the campaign trail were inherited and used by the people who had denounced them. The pattern is not that elected officials never change security policy; it is that the changes are marginal and the continuity is structural. And structural continuity across hostile transitions is precisely what you would expect if the decisive hand belonged to the permanent layer rather than the elected one.
Unelected by design
It is worth pausing on a part of this permanence that has nothing to do with spies and is openly defended, because it shows the structure is broader than the security state and not always sinister. The clearest example is the central bank. The Federal Reserve sets the price of money for the entire economy, and it is deliberately insulated from elections: its governors serve staggered, nonrenewable fourteen-year terms, longer than any president can reach, expressly so that monetary policy cannot be bent to the electoral calendar. That is unelected power by explicit design, defended on the ground that some decisions are too consequential and too technical to be surrendered to the passions of a campaign. The same logic built much of the regulatory state, the agencies that write the detailed rules governing food, drugs, air, water, finance, and communications, staffed by career officials who persist across administrations and whose decisions reach daily life more directly than most legislation does. None of this is concealed, and most of it is justified by competence and continuity. But it means that a very large share of the actual governing of a modern society is carried out by people no voter chose and cannot remove, and that the elected layer rides on top of a far larger permanent one whose existence is a matter of public record and settled law.
The government built to continue
There is a literal version of this permanence, and it is among the strangest documented programs in the modern state. Since the Cold War the United States has maintained an elaborate apparatus called Continuity of Government, designed to ensure that the state keeps functioning even if an attack kills the elected leadership. President Kennedy formalized the program in 1962, and the physical infrastructure built around it, hardened bunkers at Raven Rock and Mount Weather and beneath Washington itself, remains in service. In 1982 President Reagan established a secret office to run it, created by a classified executive order and known by the codeword Pegasus, with plans to reconstitute a functioning government, by some accounts including figures outside the constitutional line of succession, in the aftermath of a nuclear war.
Read literally, Continuity of Government is the deep state made concrete and given a budget. It is an explicit, documented arrangement for the state to survive the death of everyone the public elected, run by people the public never chose, according to plans the public has never seen. Its purpose is defensible and even necessary; a nation does need to survive a decapitating attack. But its existence states plainly the thing the rest of the argument only implies, which is that the continuity of the government and the continuity of the elected government are two different things, and the state has spent decades and fortunes ensuring the first while the second remains, by design, disposable. The apparatus is built to outlast not just any particular administration but the entire elected order, and it says so in its own planning documents.
The one time the curtain lifted
The permanence is not absolute, and the clearest proof is the single occasion on which the elected branches genuinely forced it open. In 1975 and 1976 a Senate committee chaired by Frank Church conducted the deepest investigation the American security state has ever undergone, and what it documented had unfolded across administrations of both parties over decades: CIA plots to assassinate foreign leaders, the FBI's COINTELPRO operations of surveillance and disruption aimed at domestic political movements, the CIA's secret opening of Americans' mail, and a National Security Agency program that had quietly copied vast volumes of private telegrams for years. Almost none of it had been authorized by any vote, and most of it had been concealed from the very committees nominally charged with oversight. The investigation produced real reform, the permanent congressional intelligence oversight committees that still exist and the surveillance court created in 1978. It is the strongest evidence for both sides of the argument at once. It showed how far the apparatus had operated beyond elected control, and it showed that elected control, when it finally rouses itself, can still reach inside. The permanent state is not invincible. It is only, on most days and most questions, unwatched.
What this is not
Because the term carries so much poison, the boundaries have to be drawn as firmly as the claim, and here the honest account has to turn against its own thesis and give the strongest objection real weight. The documented phenomenon is not a secret society. It does not rig elections, it did not stage the Kennedy assassination, and the lurid catalog of plots usually attached to the phrase, the staged terrorism, the puppet-master financiers, the single hidden hand directing world events, is unsupported and should be named precisely so it can be set aside. The structure described by Bagehot and Glennon needs none of it and is weakened, not strengthened, by association with it.
More important, the permanent state is in large part a feature rather than a defect, and any serious treatment has to say so. A professional civil service that does not turn over with each election is what prevents a government from collapsing into spoils and amateurism every four years. Career expertise, institutional memory, and bureaucratic independence are precisely what allow a state to resist corruption, to honor commitments across administrations, and to refuse an elected leader's unlawful order. This is the steelman, and it is formidable: the same permanence that frustrates a reforming president also frustrates a lawless one, and the very independence that looks sinister from one angle is, from another, the rule of law doing its job. It is no accident that the loudest deployment of the phrase deep state in recent years has come from elected leaders seeking to delegitimize the courts, the inspectors general, the civil servants, and the prosecutors who constrain them. The term is a weapon, and it is most often aimed at exactly the institutional checks that a healthy republic most needs. A reader who takes the structural argument seriously must hold it alongside this one, because both are true, and the tension between them is the actual subject.
The accountability question
What survives both the conspiracy theory and its debunking is a narrower, harder claim, and it is the one Glennon actually makes. It is that on certain domains, national security above all, the machinery of decision has drifted so far from the elected branches, has accumulated so much secret information and permanent expertise, that democratic control over it has become substantially thinner than the civics-textbook account admits. The voter chooses the dignified government and believes, as Bagehot intended, that he is choosing the efficient one. He is not, entirely. He is choosing the part that legitimizes, while much of the part that governs continues underneath regardless of his choice, staffed by people of real competence and usually good faith who were never on any ballot and cannot be voted out. Whether that is tyranny or merely the unavoidable shape of a complex modern state is a genuine question, and the Manifest's position is not to answer it but to make the structure visible, because a power that cannot be seen cannot be judged at all.
The deep state, properly understood, is therefore not a room with no windows and a hidden cabal inside it. It is the simple, documented, and slightly vertiginous fact that the government you vote for and the government that governs you are not quite the same government, that one of them stands for election and the other one does not, and that the one that does not is the one built to last. The part you elect does not fully govern, and the part that governs you do not elect. You can change the president. You cannot, at the ballot box, change the state, because the state was designed, deliberately and for reasons that are not all sinister, to be the thing that does not change.
Evidence Map
Facts, interpretations, forecasts, and disconfirming signals.
Core claim. In a modern security state the elected institutions hold legitimacy and visibility while a permanent, unelected apparatus holds continuity, expertise, secret information, and momentum, so on a class of questions the apparatus effectively sets policy the elected branches ratify, with no conspiracy required.
Evidence level. Facts: high (the Turkish origin of derin devlet and the 1996 Susurluk scandal, the 1947 National Security Act, Eisenhower's 1961 military-industrial-complex address, the Continuity of Government program, Bagehot's 1867 double-government framework). Interpretation: medium (Glennon's documented thesis that the Trumanite network sets national-security policy while Madisonian institutions provide legitimacy, applied to policy continuity across opposed administrations). The conspiratorial version, a cabal rigging elections or staging assassinations, is excluded.
What would confirm this. Security and surveillance policy showing strong continuity across changes of party; elected officials finding core national-security machinery resistant to redirection; the phrase deep state deployed by elected leaders against the institutional checks that constrain them.
What would disprove this. A change of administration producing deep, durable reversal of the security state's core programs by elected direction; or the permanent apparatus proving reliably subordinate to elected control on exactly the questions where the thesis predicts autonomy.
Watchlist. Structural and ongoing, tested at each transfer of power by how much of the security and administrative state actually changes.
Jerry van der Laan writes The Manifest Archive, a continuous investigation into how institutions, language, and systems shape what people are permitted to see as reality. He does not report events. He traces the structures beneath them.
Related from The Manifest Archive
