The drawer nobody opened for forty years
You are three months into a new role at a central bank's research department, and someone has handed you a question that should be answerable: why did the institution's predecessors hold the policy rate steady through a particular contraction when every contemporary model would have called for a cut? The minutes exist. The governor's correspondence exists. But the internal working papers the relevant committee circulated before each meeting, the ones that would show the analytical assumptions baked into the decision, are not on the digital finding aid. They are not in the catalogued boxes either. Eventually an archivist mentions, without drama, that a batch of files from that era was stored in a sub-basement annex on a different site during a building renovation, cross-referenced only in a card index that was never migrated to the database, and that several boxes were damaged when a pipe failed sometime in the 1980s. The answer to your question is, in a practical sense, gone.
Not exotic. The ordinary operating condition of institutional memory inside central banks, and it matters far more than most monetary economists will admit. The physical layout of an archive, which materials are stored where, how they are indexed, what environmental controls they receive, and how the retrieval system connects across sites and formats, does not merely affect convenience. It determines, structurally, which parts of an institution's past are reconstructible at all.
Paper, proximity, and the hierarchy of retrieval
Central bank archives are not monolithic. A typical large institution accumulates several distinct categories of record: formal policy committee minutes and voting records; governor-level correspondence with finance ministries and peer institutions; internal research and briefing documents; operational records from payments, supervision, and foreign exchange intervention desks; and staff personnel files. These categories are almost never stored together. They are generated by different departments, retained under different legal schedules, and often housed in physically separate locations.
Proximity to the main building matters more than it sounds. Records kept in the primary archive, with trained staff, climate control, and a live catalogue, get consulted. Records warehoused at an off-site facility run by a third-party contractor, accessible only by written request with a five-day turnaround, effectively do not exist for a researcher on a six-month project. Think of it as the difference between a book on your desk and a book in a storage unit two towns over: technically available, practically absent. The Bank of England's archive in Threadneedle Street is physically integrated into the institution's working building, which has shaped the relative richness of scholarship on British monetary history compared with institutions whose historical holdings are scattered across provincial storage depots. That is not an accident of intellectual culture. It is partly an accident of floor plans.
The deeper mechanism is what archivists call the finding aid problem. A finding aid is the roadmap to a collection: box numbers, folder titles, date ranges, sometimes item-level descriptions. When an archive grows by accretion over decades, with different departments depositing records at different times and under different internal naming conventions, finding aids accumulate inconsistencies. A folder labelled "FX Operations Correspondence, 1973" in one deposit might contain materials that would answer questions a researcher is asking under the heading "Exchange Rate Policy," but if the finding aid does not bridge those conceptual categories, the folder stays closed. Physical layout makes this worse: if the FX operations deposit lives in a different building from the monetary policy deposit, the archivist who might notice the connection between the two is less likely to be the same person.
What survives is not random
Here is the part that makes economic historians quietly uneasy. The survival of records inside a large bureaucracy follows patterns that are anything but neutral.
Consider two categories that routinely go missing. First: informal analytical work. Before a major policy decision, a great deal of the real intellectual labour happens in working drafts, annotated spreadsheets, and email chains (or, in earlier eras, handwritten memos passed between floors). These are rarely scheduled for permanent retention. They are generated by mid-level economists who do not think of themselves as creating historical records, and they live on personal drives, in desk drawers, or in departmental filing cabinets cleared during office moves. When those economists retire or the department reorganises, the material evaporates. What remains in the formal archive is the cleaned-up version: the final briefing note, the committee minute, the published statement. This creates a systematic bias in the reconstructible record toward consensus and away from the contested analysis that preceded it. The institution's official past is, by default, airbrushed.
Second: records from operational rather than policy functions. A central bank's foreign exchange intervention desk generates enormous volumes of transactional and analytical records. So does the payments oversight function, and the bank supervision division. These operational records are often retained on shorter schedules than policy records, because they were never conceived as having historical value. But they are frequently the only contemporaneous evidence of how a policy decision was actually implemented versus how it was described in the committee minute. Lose the operational records and you are left with the intention; the execution disappears.
Take a concrete case. Two researchers, call them Petra and David, are both trying to reconstruct the same institution's response to a credit squeeze in the early 1990s. Petra finds the governor's letters to the finance ministry, the published monetary policy reports, and the formal committee minutes: a coherent, well-documented narrative of deliberate countercyclical action. David, by chance, also locates a surviving set of internal liquidity-desk logs misfiled under a departmental code for "operational testing," which is why they escaped a culling exercise. Those logs show that the institution was, for roughly six weeks, rationing overnight credit to a subset of commercial banks while the public record shows no such constraint. Same archive. Different physical luck. Two incompatible histories.
The index as intellectual argument
A finding aid is not neutral description. It is an argument about what matters.
When an archivist decides to describe a collection at the box level rather than the folder level, they are making a judgment that researchers do not need item-level granularity. That judgment is usually correct for routine administrative records. It is potentially catastrophic for collections where the significant item is a single annotated draft buried inside a box of routine correspondence. The Federal Reserve System's historical archive has grappled with exactly this tension: the FOMC transcripts, once they began being released after a five-year embargo, were item-level indexed and became some of the most-cited primary sources in monetary economics. The operational records of the same period remain far less granularly described, and are correspondingly underused.
Digitisation has changed the arithmetic but not the underlying logic. Scanning a box of documents and uploading the images is not the same as making those documents retrievable. Without optical character recognition of sufficient quality, without subject tagging, without integration into a finding aid that researchers already know how to search, a digitised document is marginally more accessible than a physical one in a distant warehouse. The Riksbank, Sweden's central bank, undertook a substantial digitisation programme for its pre-war records and paired it with structured metadata entry. The result was a measurable increase in citations in academic monetary history, precisely because the intellectual architecture of the finding aid caught up with the physical preservation. Preservation without retrieval is just slower decay.
What people get wrong about this
The common assumption is that the survival problem is primarily about age. Old records are fragile; recent records are fine. This is wrong, and consequentially so. Records from the 1970s and 1980s that were created on paper and filed in physical cabinets are, if they survived at all, in reasonably stable condition and can be found by a determined researcher. Records created between roughly 1995 and 2010 exist in formats that present entirely different problems: WordPerfect files on 3.5-inch disks, proprietary spreadsheet formats, email systems whose archives were never migrated when the institution changed platforms. This is the era of peak institutional amnesia for many central banks, and the irony is that it covers some of the most consequential periods in modern monetary history: the Asian financial crisis, the dot-com contraction, the early years of inflation targeting as a formal framework.
Ask yourself: if the analytical record of those decisions is already degrading, what exactly are contemporary policymakers learning from when they invoke historical precedent?
The physical layout problem for digital records is a question of server architecture and migration policy, not shelf space. But the structural logic is identical: materials that are not actively maintained in a retrievable state, with a finding aid that connects them to the questions researchers will ask, become invisible. The medium changes. The mechanism does not.
The consequence that actually matters
Institutional memory is not just a scholarly concern. Central banks learn from their own histories. The frameworks used to stress-test contemporary policy decisions are built, at least partly, on reconstructions of how past decisions played out. If those reconstructions are systematically missing the contested analysis, the operational friction, and the informal judgments that shaped actual outcomes, then the lessons extracted are lessons from a tidied-up version of events. The institution learns from its own press releases rather than from what happened.
Archivists at institutions like the Bundesbank and the Bank of Japan have argued, with some success, for integrated preservation strategies that treat operational records and policy records as a single intellectual unit. The argument is not sentimental. It is that the gap between a committee's stated intention and the desk-level execution of that intention is precisely where the interesting and replicable knowledge lives.
The drawer nobody opened for forty years is not a metaphor. It is a filing cabinet in a sub-basement, and what is in it would change the story. Whether anyone ever reads it depends less on the curiosity of future economists than on decisions being made right now about shelf space, server migration budgets, and the granularity of a finding aid that no one in the building thinks of as consequential. That last part is the real problem.