THE POWER OF WALKING AWAY

 is Professor of Surgery at the University of Arizona and practicing surgeon in Arizona. He served as Minister of Health of the Republic of Kosova (2021–2022). Correspondence: rlatifi@arizona.edu

“Strike a half-empty pot, and it will make a loud sound; strike one that is full and hear the silence.”                      

- Kabir, an Indian Poet of 15th Century.

Resignation is the only instrument of dissent a single individual can execute without the cooperation of any other party. At least three distinct acts share the term. Type I, the declared resignation, is staged for an audience possessing the standing to act. Type II, the quiet resignation, is delivered to one recipient and may or may not generate a public record, but is powerful in its own right. Type III is not a resignation but a de facto firing, effected through institutional culture and subsequently characterized as the leader’s own decision. This essay sets out discriminating criteria for the three, drawing on Hirschman’s exit–voice framework and on cases from Celestine V to the 2025 resignations of United States federal prosecutors. One Type II resignation from national office is reported first-hand. It argues that academic medicine enumerates leadership departures without classifying them, and that burnout and the non-disparagement settlement convert Type III into Type II in the record.

Keywords

resignation; abdication; exit and voice; forced departure; institutional culture; academic surgical leadership; leadership turnover; non-disparagement agreements

“Walking away is older than the word for it. An animal losing a contest withdraws, and nothing follows from it. A marriage that has stopped working ends because one person stops staying. Nobody demands a statement; nobody keeps a record. The word appears only when the departure is attached to an office. A prime minister leaves a government, a chair leaves the department he built — and the same act acquires a name, resignation, and an obligation to explain it. Some discharge that obligation in public. Others move into silence and are not heard from again.”

The document discharging that obligation is the least reliable in the record. The resignation letter is

short by convention, composed under conditions that reward omission, and drafted with an awareness of who will read it later. Mine ran to half a page and accounted for none of the eleven months preceding it. The evidentiary base for this subject is a class of documents designed not to explain, and the question that follows is evidential rather than moral: what else must be established before a departure can be classified?

The act can be executed by one individual without the cooperation of any other party, and once tendered it is complete. The ambiguity is as old as the act: in Roman law abdicare denoted first the disowning of a son and only later the renunciation of office, and the standard definition notes that where an abdication is claimed there is frequently an evident element of constraint — a display of willingness produced to avoid what would otherwise be classified as deposition.1

Three cases fix the range. In 1294 Celestine V, finding that canon law had never contemplated a pope departing alive, first legislated the possibility and resigned a week later; he was detained by his successor until his death in 1296, and the act was contested from the outset as either his own decision or one Boniface VIII induced.2 In 1783 George Washington returned his commission while commanding a victorious army, registering no grievance — which shows that efficacy is independent of protest and greatest where the person forfeits the most.3 In April 1814 Napoleon, deposed by the Senate on 3 April, signed unconditional terms on the 6th; Fontainebleau granted him Elba, an annuity and his imperial title.4 The signature followed the removal rather than producing it, and the settlement is recognizably a severance agreement. He returned in 1815 and abdicated again after Waterloo, but never revoked the first abdication and was restored by no institution.5 A resignation cannot be withdrawn, only overthrown.

A Three-Part Taxonomy

The word is applied to at least three acts differing in initiator, audience, documentary residue and consequence. Type I, the declared resignation, is staged to be observed; its objective is witnessed departure and the transfer of cost to a second party. Type II, the undeclared resignation, is delivered to a single recipient without publicity; the motive may be identical, the method and consequences are not. Type III is not a resignation but a de facto dismissal, effected through institutional culture and characterized afterward as the leader’s own decision.

Three parties have an interest in the conflation. The institution benefits because one undifferentiated term allows a Type III removal to be recorded as a Type I or Type II departure: the act enters the record, the initiator does not. The departing leader often benefits too, ambiguity being more tolerable than either protest or defeat, and in medicine frequently a contractual condition of the settlement. The result is a classification error with downstream effects: protest is recorded as failure, removal as choice, and institutional dysfunction as individual exhaustion.

Exit, Voice, and the Timing of Leverage

Hirschman’s 1970 formulation supplies the mechanism: members of a deteriorating organization have two responses, exit and voice, with loyalty determining the proportion.6 Resignation is the only act combining them. Exit alone transmits no information; voice alone is deniable, the organization absorbing the objection and retaining the objector to absorb the next. One consequence governs every case below: leverage exists only while the act is unexercised, evidence only once it is exercised. An officeholder who threatens to resign has influence, one who resigns has proof, and the two are never simultaneously available.

Type I: The Declared Resignation

On 20 October 1973 Attorney General Elliot Richardson declined an instruction to dismiss the Watergate special prosecutor and resigned; William Ruckelshaus declined and resigned within minutes.7 Neither delivered a statement: the act was the statement. Cyrus Vance wrote his resignation as Secretary of State and delivered it three days before the Iran hostage rescue operation he had opposed, recording that he would resign irrespective of the outcome.8 Where the act follows the outcome it cannot be distinguished from retrospective judgment. Shevardnadze in 1990,10 Cook in 200311 and Hailemariam Desalegn in 201812 register the same act in three different systems.

In February 2025 the acting United States Attorney for the Southern District of New York resigned rather than dismiss the corruption prosecution of the mayor of New York;13 six further federal prosecutors resigned over the same directive, in what was reported as the largest mass resignation at the Department of Justice in recent memory.14 What made the acts Type I is that the reasoning entered the public record and third parties acted on it. Simultaneity modifies the calculation: one departure is absorbed, seven are not.

Four variables determine whether a Type I act functions: timing, which decides whether it is separable from the outcome; venue, which decides who must respond; the documentary record; and an audience with standing to act. Where one is absent, the act remains protest to the person performing it while functioning externally as an ordinary departure. The discriminating test is whether cost transfers to the party being left. Howe’s did. Thatcher’s longest-serving minister resigned on 1 November 1990 and said nothing for twelve days; his statement of 13 November removed the assumption on which her position depended, that no senior colleague would stand against her. She fell four votes short in the first ballot and left office on 28 November after eleven years.9,15 Fifteen days separated the statement from the change of government.

Type II: The Undeclared Resignation

Type II has almost no literature because it generates almost no evidence. The attached report is the Type II substitute for a public statement — a record transferred to the main recipient rather than to the public.16

A hypothesis follows, offered as a hypothesis and not a finding: Type II resigners may be disproportionately people who accepted the position to complete a specified body of work rather than to hold the office. At least four competing explanations predict identical behavior — temperament; structural isolation, since a resignation is amplified only where an amplifier exists; option preservation in a small state and a small profession; and genuine uncertainty as to whether public protest was warranted. Discriminating among them requires that the reason for accepting the post be established independently, before the manner of departure is examined.

Type II terminates complicity and preserves the capacity for subsequent work, which for a surgeon is material rather than rhetorical. Its principal limitation is that it is externally indistinguishable from acquiescence, and institutions depend on that: silence intended as restraint is received outside as consent. It generates little or no entry in the record, which is why the category remains unquantified.

Type III: Removal Presented as Resignation

Institutions maintain a stable lexicon for the third category: stepping down to pursue other opportunities; returning to the faculty; attending to elderly parents; concentrating on research and patient care. Each formulation obscures a specific antecedent, and each is accepted at face value by the literature that then classifies the departure as voluntary.

Corporate governance addressed the equivalent problem thirty years ago. Parrino’s 1997 classification treats the official announcement as a hypothesis rather than a conclusion: a departure is reclassified as forced where the surrounding record indicates removal or unspecified policy differences.17 The analogy is close — a dean occupies the position of a board, faculty that of shareholders — but the verification apparatus does not transfer: academic medicine generates no filings carrying a penalty for a materially misleading account.

The available data are descriptive and silent as to cause. Mean tenure of internal medicine chairs at United States medical schools fell from approximately 5.3 years in the 1970s to approximately 4.0 in 2000–2006, while mean interim chair tenure rose from about 1.4 to 2.4 years.18 Annual chair turnover in United States obstetrics and gynecology departments approximately doubled between 1981 and 2005, from 6.0% to 12.7%.19 Among 178 United States surgery department chairs, women held the position a median of 5.3 years against 7.0 for men.20 Among 1,166 medical school deans appointed between 1959 and 2019, median tenure was 4.3 years.21 Each series enumerates departures; none classifies them. The field knows how often its leaders leave and has no estimate of how often they were removed.

Burnout completes the reclassification. Shanafelt and Noseworthy established that burnout is a system-level problem institutions persistently treat as an individual one.22 Read in reverse, the finding specifies a mechanism: an organization that withholds resources, reverses decisions after they are taken and excludes a leader from the forums where decisions are made will reliably produce exhaustion, which is then advanced as the cause of the departure. The diagnosis is correct and the causal direction inverted.

The record is then closed. In the United Kingdom, 98 higher education institutions reported issuing close to 11,000 non-disclosure agreements between 2014–15 and 2018–19.23 At one large United States university, student journalists documented ten settlement agreements with departing employees over six months, carrying non-disparagement provisions and payments exceeding $1.26 million, none appearing in public documents.24 The instrument that resolves the dispute eliminates the evidence.

 Consequences and Limitations

Four consequences follow. For the individual, selecting the wrong type expends the act without effect. For the institution, a forced departure recorded as voluntary forecloses learning: the successor inherits the same withheld resources and parallel structures, and the sequence recurs under a different name. For the evidence base, a discipline that would not accept self-reported outcomes in a surgical trial accepts them without qualification in its own leadership data. And for patient safety, the leader most exposed to quiet removal is the one pressing hardest on staffing, safety or resources.

Four limitations bound these conclusions. The historical cases come from the published record, which over-represents Type I by construction; the essay reasons from visible cases to invisible ones, and that inference is not secured by the evidence presented. The account of my own resignation is a self-report by an interested party, offered as illustration and not as evidence. No denominator exists: the frequency of Type III in academic medicine is unknown. And the Parrino method is proposed for transfer here, not demonstrated.

An Evaluative Framework

Dobel argued that the moral case for resignation rests on three dimensions of integrity — the promise made in accepting office, the competence to perform it, and the capacity to be effective in it — failure in any one generating strong moral reasons to resign.25 Six operational questions follow. What was the appointment intended to accomplish, and does that remain achievable? Is the obstacle a disagreement, which has identifiable counterparties, or a design, which exhibits a pattern of resources withheld and exclusion from the forums where decisions occur? Would a Type I act alter anything, which requires both an audience with standing and a cost that transfers? What capacity will remain the following day? What is being signed, and under what time constraint? And what will the departure demonstrate to those who remain?

Conclusion

The resignation letter cannot classify itself. Classification requires external evidence — the conditions surrounding the letter rather than its contents — and that work has not been undertaken systematically in medicine or in government. The power of resignation does not reside in the departure but in its availability: a leader who cannot conceive of resigning has forfeited the only leverage that does not depend on another party’s permission. The option need not be exercised to function; it is required only that it be real. An institution that can engineer a departure and reclassify it has established, for every leader remaining, that the option is not real — that the departure, when it comes, will be characterised by another party, in language chosen by that party.

Key Points

•   Three acts share the term resignation and are separable by initiator, intended audience, documentary residue and who expends the leverage.

•   Leverage exists only while the act is unexercised and evidence only once it is exercised; the two are never simultaneously available.

•   Type I efficacy depends on four concurrent variables: timing, venue, documentary record, and an audience with standing to respond.

•   Type II is externally indistinguishable from acquiescence, which is why the category remains unquantified.

•   Burnout and the non-disparagement settlement convert Type III into Type II in the record and eliminate the confirmatory evidence.

References

1. Abdication. Encyclopaedia Britannica. Available at: britannica.com/topic/abdication

2. Papal resignations: the case of Celestine V. Lillian Goldman Law Library, Yale Law School. Available at: library.law.yale.edu (Abdication of Pietro da Morrone, Pope Celestine V, 13 December 1294; the enabling constitution and the disputed circumstances.)

3. Resignation of military commission [George Washington, Annapolis, 23 December 1783]. Digital Encyclopedia, George Washington’s Mount Vernon. Available at: mountvernon.org (Includes George III’s reported remark to Benjamin West.)

4. 4, 6, and 11 April 1814: Napoleon I’s three abdication declarations. Napoleonica. La Revue, 2023. Available at: shs.cairn.info (The conditional abdication of 4 April, the unconditional act of 6 April, and the Treaty of Fontainebleau of 11 April 1814.)

5. The Hundred Days. World History Encyclopedia. Available at: worldhistory.org (Departure from Elba, February 1815; Waterloo, 18 June 1815; second abdication, 22 June 1815; exile to Saint Helena.)

6. Hirschman AO. Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Cambridge, MA: Harvard University Press; 1970.

7. October 20, 1973: The Saturday Night Massacre. ABA Journal. Available at: abajournal.com

8. Editorial Note [Vance resignation letter, 21 April 1980]. Foreign Relations of the United States, 1977–1980, Volume II, Document 199. Office of the Historian, US Department of State. Available at: history.state.gov

9. Howe G. Personal Statement [resignation from the Cabinet, 1 November 1990]. Parliamentary Debates (Hansard), House of Commons, 6th series, 13 November 1990, vol 180, cols 461–465. Available at: api.parliament.uk

10. Shevardnadze E. Speech to the Fourth Congress of People’s Deputies, 20 December 1990. Seventeen Moments in Soviet History, Michigan State University. Available at: soviethistory.msu.edu

11. Cook R. The Point of Departure. London: Simon & Schuster; 2003. (Resignation from the Cabinet and speech to the House of Commons, 17 March 2003.)

12. Ethiopia prime minister Hailemariam Desalegn resigns. Al Jazeera, 15 February 2018. Available at: aljazeera.com

13. Acting US attorney in Manhattan resigns after directive to drop case against Mayor Eric Adams. ABC7 New York (WABC), 13 February 2025. Available at: abc7ny.com

14. Fallout from the Eric Adams case continues at the Justice Department. NPR, 14 February 2025. Available at: npr.org (Seven federal prosecutors in New York and Washington resigned; reported as the largest mass resignation at the Department in recent memory.)

15. MT’s private files for 1990 — (7) uneasy peace then leadership election. Margaret Thatcher Foundation archive. Available at: margaretthatcher.org (Howe’s resignation and speech; Heseltine’s challenge; the first ballot of 20 November 1990 and Thatcher’s withdrawal on 22 November.)

16. Latifi R. A Conversation with Dr. Rifat Latifi [self-interview on leaving the Ministry of Health of Kosova]. Dielli — The Sun, 2025. Available at: gazetadielli.com

17. Parrino R. CEO turnover and outside succession: a cross-sectional analysis. Journal of Financial Economics. 1997;46(2):165–197.

18. Ringenbach JL, Ibrahim T. Tenures for department chairs: how short is too short? The American Journal of Medicine. 2008;121(2):163–168. Available at: amjmed.com

19. Rayburn WF, Schrader RM, Cain JM, Artal R, Anderson GD, Merkatz IR. Tenure of academic chairs in obstetrics and gynecology: a 25-year perspective. Obstetrics & Gynecology. 2006;108(5):1217–1221. doi:10.1097/01.AOG.0000240140.27246.93. PMID: 17077245.

20. Kennedy GE, Bergstresser SL, Rakestraw SL, Novak Z, Corey B, Chen H, Sutzko DC. Does chair of surgery gender influence divisional or residency program director gender diversity? Journal of Surgical Research. 2021;267:224–228. doi:10.1016/j.jss.2021.05.026. PMID: 34157491.

21. Chatterjee A, Chin LS, Connolly H, et al. Length of tenure of U.S. medical school deans who started in their position from academic years 1959–1960 to 2018–2019. Academic Medicine. 2025;100(9):1044–1050. doi:10.1097/ACM.0000000000006028. PMID: 40080627.

22. Shanafelt TD, Noseworthy JH. Executive leadership and physician well-being: nine organizational strategies to promote engagement and reduce burnout. Mayo Clinic Proceedings. 2017;92(1):129–146. doi:10.1016/j.mayocp.2016.10.004.

23. Stokel-Walker C. UK universities issue 11K non-disclosure agreements in five years. Times Higher Education, 11 April 2019. Available at: timeshighereducation.com

24. Buying Silence: settlement agreements between the University of Michigan and former employees, November 2018 – April 2019. The Michigan Daily, 10 December 2019. Available at: digital.bentley.umich.edu

25. Dobel JP. The ethics of resigning. Journal of Policy Analysis and Management. 1999;18(2):245–263. doi:10.1002/(SICI)1520-6688(199921)18:2<245::AID-PAM3>3.0.CO;2-H.