Clerk Turnover and Institutional Knowledge: What Actually Breaks When a 20-Year Clerk Leaves

Every clerk's office has a version of the same story. Someone retires after eighteen, twenty-five, thirty-two years. At the farewell lunch, somebody says the line: "Nobody knows where anything is but her." Everyone laughs, because it is funny, and then everyone goes quiet, because it is also true.

The standard conclusion follows automatically. Turnover is a crisis. Institutional knowledge is walking out the door. Performance is going to fall off a cliff. That narrative is repeated so consistently across local government that it has stopped being an argument and become an assumption.

The first half of it is measurably correct. Turnover is real, it's rising, and the largest wave is still in front of us. The second half is where it gets interesting, because we now have unusually good evidence on what actually happens after the long-tenured official leaves, and it does not say what almost everyone expects.

The turnover is real, and now it's measured

Until recently, "clerk turnover is up" was an impression. It's now a dataset. In "Does Leader Turnover Degrade Local Government Performance? Evidence from Local Election Officials" (September 25, 2025), Joshua Ferrer and Daniel M. Thompson of UCLA traced the tenures of nearly 20,000 elected and appointed chief local election officials across 6,290 counties and municipalities in all 50 states, covering every even-year general election from 2000 through 2024. That's over 80,000 jurisdiction-year observations, and it is by a wide margin the most complete picture of local administrative turnover anyone has assembled.

One definitional note matters before the numbers, because it's easy to misread them. The authors define turnover as a change in a jurisdiction's chief election official since the general election held four years prior. They use a four-year window deliberately, because many of these officials are elected on four-year midterm cycles and a two-year measure would just pick up the election calendar. So a 41% turnover rate does not mean 41% churn annually. It means that in 2024, four out of ten jurisdictions had a different person in the chair than they did in 2020.

The authors fit a regression on the 2004 to 2020 data and asked what turnover should have been in 2022 and 2024 if the existing trend had simply continued. Both years came in about four percentage points high, and both are statistically distinguishable from that trend. The rise is not an artifact of the normal cycle. Something changed.

And the wave hasn't crested

If the last four years felt like the hard part, the workforce data says otherwise. MissionSquare Research Institute's 2025 State and Local Government Workforce Survey, released July 8, 2025 and fielded March 4 through April 7, 2025 with 382 state and local HR professionals, found that 46% of governments expect the most significant wave of anticipated retirements is still to come.

Set that next to the two numbers that follow it in the same survey, because the pairing is the actual story:

• 50% of governments rate succession planning as very important.
• 61% report not having a succession planning process in place.

That gap is not a knowledge problem. Nobody needs to be persuaded that succession planning matters; half of them already say it's very important and most of the rest presumably don't think it's worthless. The gap is that succession planning, as usually conceived, is a project. It's a binder someone has to sit down and write, competing against a Tuesday meeting that has to be posted by Friday. Projects that compete with deadlines lose to deadlines, permanently.

The part almost everyone gets wrong

Here is where the Ferrer and Thompson paper stops confirming the conventional wisdom and starts contradicting it. Having documented the turnover surge, they went looking for the damage it was supposed to have caused. They measured six performance outcomes: voter turnout, the residual vote rate (ballots cast with no vote at the top of the ticket, a standard proxy for ballot and equipment problems), the provisional ballot rate, voter-reported problems voting, wait times at the polls, and voter confidence.

Their finding, in their own framing: local election official turnover does not noticeably degrade performance. It held across the outcomes they measured and across the subsets of the data they tested, with one possible exception we'll come back to.

They then pushed on the obvious objection: maybe offices held up because the replacements were all seasoned insiders. Partly true, and worth knowing. Of the new officials they could research, more than 60% had prior professional experience in elections, and nearly 80% had prior experience in government. That's a stronger experience profile than newly elected mayors or members of Congress.

But they then tested the harder case. When they restricted the analysis to incoming officials with no prior elections or government experience — the true cold starts, the exact scenario the panic narrative is about — performance still didn't fall off. As they put it, incoming officials without that experience performed about as well as those with it.

So what was holding the office up?

The authors offer two candidate explanations, and the second is the one that should interest anyone who runs a records function. Either selection is good enough that replacements are consistently capable, or the individual leader simply is not what determines performance. On the available data they can't fully separate the two.

The theoretical grounding they lay out for that second possibility is worth stating directly, because it reads like a description of a well-run clerk's office. Performance survives leadership change when institutional knowledge is distributed across many people rather than concentrated in one, when lower-level staff stay in place through the transition, and when the organization is built with formal rules and standard routines that buffer against leadership transitions. Offices don't run on a person's memory. They run on procedure, on documentation, and on the people one rung down who never left.

Which reframes the farewell-lunch joke considerably. "Nobody knows where anything is but her" is not a tribute. It's a defect report. In the offices that came through turnover intact, that sentence was false, and its being false is precisely why they came through.

The exception, and why it's the interesting one

One outcome in the study didn't come through as cleanly: wait times at the polls. The authors flag it as the possible exception to their null result.

That is a suggestive place for the exception to land. Turnout, residual vote rates, provisional rates — those are downstream of rules, forms, equipment, and procedure, the parts of the operation that live in documents and don't care who's holding the clipboard. Wait times are downstream of live judgment under load on a single high-pressure day: how many machines to put where, when to open a second line, how to react when something goes sideways at 4:45 p.m.

The codified layer transferred. The real-time judgment layer transferred less well. For a clerk's office, the analogue is not hard to see. The agenda template, the retention schedule, and the posting procedure are codified and portable. Knowing that this particular board always tables an application when the applicant isn't in the room, or that the chair says "without objection" and means a vote was taken, is real-time judgment. That is the layer that walks out the door, and it's also the layer that determines whether the minutes of a messy meeting come out right.

What actually leaves with a long-tenured clerk

Take the finding seriously and the question sharpens usefully. The risk is not that a departing clerk takes "the knowledge." Most of the knowledge is in statutes, codes, bylaws, templates, and the files themselves, and none of that moves. The risk is the specific subset that was never written anywhere because the person who knew it was always in the building.

In a records function, that undocumented layer is remarkably consistent from office to office:

• Which body actually approves what. The formal delegation is in the code. The practice — what goes to the full board versus what the commission disposes of, what has been handled on consent for eleven years — often is not.
• The numbering and naming conventions. How ordinances, resolutions, and minute orders are numbered, when the sequence resets, and what happens to the number when an item is tabled and returns three months later.
• Where things get posted, and where else. The statutory location plus the bulletin board by the side door that the newspaper checks, plus the person who has always been emailed the packet as a courtesy.
• The unwritten motions practice. How this chair handles amendments, whether abstentions are recorded with reasons, how a failed motion is written up, what "so moved" means when nobody restated the motion.
• Where the recording lives and who controls it. Which account holds the video, what the retention clock is, who has the password. This one is a genuine single point of failure and it is almost never documented.
• The correction history. Which past minutes were amended and why. New clerks inherit approved minutes with no memory of which ones were fought over.

None of that is exotic. All of it is the kind of thing that gets discovered by a new clerk in week three, usually at the worst possible moment, usually in public.

The deadlines don't care that you're short-staffed

Here is the structural cruelty of the transition period: open meetings law is almost entirely blind to staffing. The obligations arrive at full strength on a new clerk's first day.

Illinois requires an agenda posted 48 hours ahead, and as we covered in our breakdown of 5 ILCS 120/2.02, that rule carries no population threshold — a village of 400 and the City of Chicago are on the same clock. Ohio's Sunshine Law, examined in our post on R.C. 121.22, applies its minutes standard and its notice-rule requirement to every public body in the state regardless of size, with a mandatory injunction behind it. Washington's OPMA special-meeting notice is the same 24 hours whether you have a records division or a part-time clerk who also handles utility billing.

Nowhere in any of these statutes is there a grace period for a new clerk. There is no provision that reads "unless the clerk started last month." A body that misses the notice requirement during a transition is exposed exactly as much as one that missed it through indifference, and in several states the consequence is that the action taken is void or voidable.

So the transition is the moment of maximum legal exposure and minimum institutional slack, and it is arriving at more offices, more often, with 46% of governments saying the biggest wave is still ahead.

Treat the record system as the succession plan

The practical implication of the research is genuinely encouraging, and it's this: the thing that protects an office through turnover is not a heroic knowledge-transfer project. It's the ordinary, boring degree to which the work already lives outside anyone's head.

That's a much better problem to have, because it's addressable incrementally, during normal work, rather than as a binder someone has to find three months to write. A few things that convert tacit knowledge into transferable knowledge as a byproduct of doing the job:

1. Make the process the documentation. If the agenda is built by filling in a defined structure rather than by editing last month's file, the structure itself teaches the next person the rules. A workflow that only works if you already know the conventions is a workflow that cannot survive a transition. This is one of the underrated arguments for moving off Word and email: shared files carry no rules, only precedent that has to be interpreted.
2. Write the conventions down where the work happens, not in a manual. A note attached to the item type, visible at the moment someone creates that item, gets read. A binder in a drawer does not. The test of documentation is whether it is in the path of the work.
3. Keep the source material attached to the record. When the recording, the transcript, and the packet stay linked to the meeting rather than scattered across a drive, a personal inbox, and a vendor portal, a new clerk can reconstruct what happened without needing to ask anyone. This is the same reason AI drafts need the audio kept attached for verification: an unverifiable record is a record that depends on someone's memory.
4. Audit the single points of failure now. Make a list of every system where exactly one person has the credentials or the knowledge. Recording platform, posting account, website CMS, archive. Fix the ones that only have one name on them. This is an afternoon, not a project, and it's the highest-value afternoon in the list.
5. Cross-train one rung down, deliberately. The research's most plausible mechanism is that staff who stayed kept the office running. That only works if someone besides the clerk has actually done the task. A deputy who has watched minutes get built is not the same as a deputy who has built them.
6. Capture judgment calls when they happen. The reasoning behind an unusual ruling is available for about a day and then it's gone. A one-line note on the item at the time is worth more than an exit interview two years later.

The through-line

The clerk turnover wave is not a myth. It's documented now, more rigorously than most local-government trends ever get: a steady two-decade climb, a sharp out-of-trend break in 2022 and 2024, and a retirement wave that nearly half of governments say hasn't arrived yet. Anyone telling you it's overblown is arguing with a good dataset.

What the same body of evidence undercuts is the fatalism that usually accompanies it. Offices did not collapse. Even the cold-start replacements, with no elections or government background, did not preside over meaningful degradation. The institution turned out to be more durable than the individual, which is the whole point of an institution and is easy to forget when the individual has been there for thirty years.

But that durability was not free and it was not luck. It came from distributed knowledge, from staff who stayed, and from formal rules and standard routines that carried the work across the gap. Where those were absent, there was nothing to carry it.

Which leaves a clerk's office with a concrete and fairly cheerful task. You cannot control when your most experienced person retires. You can control how much of what they know exists only in their head on the day they do. Every convention you move out of memory and into the system is one less thing that leaves with them. Do that steadily, and the farewell lunch is just a lunch.

Frequently asked questions

Sources: Joshua Ferrer & Daniel M. Thompson (UCLA), "Does Leader Turnover Degrade Local Government Performance? Evidence from Local Election Officials" (Sept. 25, 2025); verification materials at Harvard Dataverse, doi:10.7910/DVN/WML66X · Bipartisan Policy Center, "Election Official Turnover Rates from 2000-2024," with Ferrer & Thompson · MissionSquare Research Institute, 2025 State and Local Government Workforce Survey (July 8, 2025; n=382, fielded Mar. 4–Apr. 7, 2025). Turnover figures are four-year-change rates across 5,928 jurisdictions in 49 states, excluding Massachusetts for data availability. The turnover research measures election administration outcomes; its application to records and minutes functions is our inference from the mechanism, not a finding of the study. This article is general information, not legal advice; consult your attorney or state archives for your jurisdiction's rules.