Part I — Situation overview

NATO member states report the size of their armed forces to the alliance every year, and NATO has also published the main figures since 2014 (HVG). In October 2026, a new government reported for the first time in 16 years. According to HVG’s article of 26 October, the latest NATO statistics not only contain this year’s figure but also retroactively correct earlier years. The previous Ministry of Defence (HM), led by Kristóf Szalay-Bobrovniczky, reported 20,100 personnel for 2023, but according to the new leadership only 17,600 soldiers served in 2023. That is a gap of 2,500 people, or 12.4 per cent, roughly the strength of a medium-sized brigade (HVG, 24.hu). According to Telex’s summary, the 2024 headcount was cut back by 2,200 (10.5 per cent) and the 2025 headcount by 1,700 (7.8 per cent). A retroactive correction of 7–12 per cent is the third largest in relative terms within the alliance, after Turkey (−19 per cent) and Finland (−16 per cent).

The figures in themselves are estimates that member states can refine later; according to HVG, after two years they no longer count as estimates but as final data. The previous ministry leadership did not refine them before the end of its term. The HM told the paper that the headcount fell significantly in 2023 because of the previous government’s “rejuvenation”, and that the data were also affected by the transfer of the Defence Forces Hospital to the Ministry of the Interior. The decline is therefore partly a real loss of personnel and partly a change in the scope of what is counted, but the proportion of the two factors is not known. Since compulsory military service was suspended in peacetime in 2004 (HVG), the headcount today directly measures the capacity to recruit and retain. On the same day, after a lawsuit that went all the way to the Kúria, 24.hu obtained the construction contracts for an official military villa; this shows a similar deficiency on the procurement side and is the subject of a separate analysis.

MIAK’s reading: the headcount reported to NATO is not an internal statistic but common baseline data for alliance planning and for domestic parliamentary scrutiny. Its credibility cannot depend on whether someone looks back after a change of government. The correction is right, but it is a one-off step; without an institutional guarantee, the next discrepancy may also come to light only years later.

Part II — Grounding in the literature

According to the organisational behaviour model in Essence of Decision, the analysis of the Cuban missile crisis by the American political scientist Graham Allison and the American historian Philip Zelikow, large organisations operate along routines, and information is filtered at every level of the hierarchy. A piece of data can be “in the system” without the leadership seeing it, so a figure once reported can be passed on uncorrected. Carl von Clausewitz, the Prussian military theorist, warns in his main work On War that most military reports are contradictory, inaccurate or doubtful, so the decision-maker must treat them with methodical criticism. This approach also applies to peacetime headcount data. In his book World Order, the American diplomat and historian Henry Kissinger describes how allies base their security on an assessment of the real balance of power, not on formal commitments. Alliance credibility therefore also depends on the reliability of data. The detailed discussion of the literature — author by author, with quotations — is in section 6.4 Literature in detail.

Part III — MIAK’s concrete proposal

MIAK proposes three measurable steps that turn the current one-off correction into a lasting data-reporting regime.

3.1 A defence revision log and publication of the methodology (by 31 December 2026)

The HM should publish the headcount and expenditure time series reported to NATO back to 2014, together with the counting methodology: which personnel groups are included (professional, contract, voluntary reservist, defence civilian employee), and which institutions fall within the scope. Every retroactive correction should be entered in a public revision log recording the date, size and reason of the correction — indicating separately whether it is a methodological change (for example, the transfer of the Defence Forces Hospital) or a real change in headcount. The HM should break down the 2023 gap of 2,500 into these two factors. The proposal transfers to defence a principle already formulated for budgetary data: MIAK’s analysis of 2 October called for a similar log for revisions of the EU deficit reports. According to the lesson of organisational behaviour (see 6.4.1), an error inherited through routine can be broken only by a mandatory, public review point. The proposal extends programme point HV2 (transparency of defence spending) to headcount data and follows the logic of the decision log in KI8.

3.2 Independent ex-post data verification (annually, for the first time by 30 June 2027)

The annual audit plan of the State Audit Office (ÁSZ) should include a comparison of the headcount data reported to NATO with the data of the military payroll and the personnel records. The ÁSZ is an audit body: it submits its findings in a report to the National Assembly, and drawing conclusions is the task of Parliament. The public part of the report should contain the size and reason of any discrepancy; classified data should be discussed by the National Assembly’s Defence Committee in closed session. The principle of “public unless justifiably classified” already appears in programme point HV2; here it must be applied to the fact that the national aggregate headcount is not secret, since NATO itself publishes it. Clausewitz’s warning (see 6.4.2) says precisely this: it is not enough to accept a report; it must regularly be tested against an independent source. The independent verification runs every year regardless of changes of government, which makes it party-neutral.

3.3 Measurable recruitment and reservist targets (together with the 2027 budget bill)

When submitting the 2027 budget, the Government should publish headcount targets by personnel group for the end of 2027 and for 2030, and the HM should report quarterly on their achievement. For the reserve forces, MIAK’s programme point HV10 envisages 20,000 voluntary reservists by 2030; for this, the starting position must also be recorded publicly, using the same methodology as in the NATO report. Retention must also be measured: how many soldiers leave before their contract expires, and for what reason. Without this, the success of recruitment campaigns cannot be judged, and the headcount target could be met on paper only, just like the 2023 report. According to the alliance logic described by Kissinger (see 6.4.3), partners weigh actual capability, so a measurable target also strengthens the credibility of Hungary’s commitments.

The three proposals are linked by one principle: defence data are a common resource on which allies, Parliament and voters alike rely. If the counting method is public, corrections are logged and the reported data are verified every year by an independent body, then the next discrepancy will come to light not years later, after a change of government, but in the same year.

Part IV — Expected effects and risks

Dimension Expected effect Risk
Alliance relations more credible Hungarian data reporting; capability planning rests on real figures further corrections damage the country’s image in the short term
Defence the personnel shortage becomes visible; recruitment and retention become more targeted publishing the real figures can also be read as a vulnerability
Public administration a single methodology for the NATO report and the domestic report additional administrative burden on the HM’s personnel organisation
Budget personnel expenditure and headcount become comparable meeting the recruitment targets may require additional funds
Public trust party-neutral verification repeated every year the correction may become the subject of political skirmishing

The main question to weigh is the relationship between openness and national security. The national aggregate headcount is not secret; NATO itself publishes it. The breakdown by unit and readiness data, however, are genuinely sensitive. The proposal therefore makes public only the aggregate and the methodology, and leaves the details to closed sessions of the Defence Committee. The other risk is political: if the correction is used solely as an argument against the previous government, the same could happen in reverse at the next change of government. This is exactly why the proposal does not ask for a one-off investigation, but for an annually repeated procedure that applies equally to every government.

Part V — Measurability and summary

5.1 What is worth tracking? (proposed KPIs)

MIAK proposes tracking the following key performance indicators (KPIs):

  • Revision log: by 31 December 2026, the headcount and expenditure time series reported since 2014 is public, with the methodology and a breakdown of the 2023 gap by cause.
  • Gap between the two reports: from 2027, the headcount reported to NATO and the one in the domestic report are based on the same methodology, and the gap between them is zero.
  • Size of ex-post corrections: future retroactive corrections do not exceed 2 per cent of the reported headcount in any single year.
  • Recruitment and retention: the headcount target and the actual headcount by personnel group, as well as the share of departures before the end of contract, are published quarterly.

5.2 Summary

The key message: the current correction is right, but it will only become a lasting result if an institutional rule follows it. MIAK asks the defence ministry to publish the methodology for calculating the headcount and the log of corrections by the end of the year. It asks the State Audit Office to include a comparison of the reported data in its annual audit plan, and the Government to publish measurable recruitment targets together with the budget.

Two of MIAK’s foundational values are at stake. Data-drivenness, because defence planning, burden-sharing within the alliance and parliamentary scrutiny all rest on these figures — if the baseline data are false, every decision built on them is distorted. Accountability, because the proposal is not aimed at holding a former member of the government to account, but at a procedure that applies equally to every future leadership.

Part VI — Reasoning and further sources

6.1 Press framing by spectrum

The topic was brought to light by HVG in its subscriber section, with strong framing: according to the headline, Fidesz may have reported “cosmetically adjusted” data, and the lead suggests that this may not be a one-off discrepancy. The paper highlights the two-year refinement deadline and the fact that a new government reported for the first time in 16 years. Telex, citing HVG, strikes a more matter-of-fact tone: it uses the conditional (“may have reported”), and also publishes the international comparison (Turkey, Finland) and the HM’s full explanation, so the reader also sees the methodological reasons.

24.hu personalises the responsibility (“Szalay-Bobrovniczky and his team reported far more soldiers”), but it too quotes the HM’s explanation about rejuvenation and the transfer of the Defence Forces Hospital. On the same day the paper also wrote about the contracts for the official villa, so in its coverage the two cases fit into a single accountability narrative. Index, Portfolio and ATV did not put the topic on their front pages. The conservative segment (Magyar Nemzet) did not include the headcount correction among its main topics that day, so an interpretation reflecting the previous government’s perspective — for example, that the later refinement of estimates is standard practice in the alliance — did not appear in the press examined.

6.2 Facts and data

Indicator Value Source
Headcount originally reported for 2023 20,100 HVG, 24.hu
Corrected headcount for 2023 17,600 HVG, Telex, 24.hu
2023 correction −2,500 (−12.4%) HVG, Telex
2024 correction −2,200 (−10.5%) Telex
2025 correction −1,700 (−7.8%) Telex
Larger relative corrections in NATO Turkey −19%, Finland −16% Telex
NATO’s public headcount statistics since 2014 HVG
Refinement period, after which data are final 2 years HVG, 24.hu
Compulsory military service in Hungary (suspended in peacetime since 2004) 1868–2004 HVG (exhibition of the Museum of Ethnography)

The numerical breakdown of the two causes named by the HM — the 2023 “rejuvenation” and the transfer of the Defence Forces Hospital to the Ministry of the Interior — is not publicly known; proposal 3.1 asks for it to be published.

6.3 Policy dimensions

  • Defence (programme points) — public, verified publication of defence data (HV2), measurable expansion of the reserve forces (HV10), taking organisational routines into account in decision-making (HV13);
  • Transparency and anti-corruption policy (programme points) — the logic of the public money dashboard applied to personnel expenditure (A1), transparency of procurement by HM companies as a parallel thread (A2);
  • Public administration and e-government (programme points) — the order of responsibility for data reporting to international organisations and the decision log (KI8).

6.4 Literature in detail

6.4.1 Graham Allison and Philip Zelikow: Essence of Decision

Allison and Zelikow analyse the Cuban missile crisis with three models; according to the second, the organisational behaviour model, a large share of government decisions are not the choice of a rational leader but the result of the standard procedures of large organisations. According to the authors, information does not automatically reach the top of the organisation: “Facts can be in the system without being available to the head of the organization”, because information has to be filtered at every level of the hierarchy. Organisations also cling to familiar patterns, and major organisational change typically comes after a crisis, not before it. In the case of Hungary’s headcount data, an estimate once reported can be passed on uncorrected if it is nobody’s job to look back — and the correction was again brought about by a “crisis”, the change of government. The revision log replaces this dependence on crisis with regular, mandatory review.

📖 Source: Graham Allison – Philip Zelikow: Essence of Decision: Explaining the Cuban Missile Crisis

6.4.2 Carl von Clausewitz: On War

In his chapter on information in war, Clausewitz is unsparing: “Great part of the information obtained in War is contradictory, a still greater part is false, and by far the greatest part is of a doubtful character.” The officer’s task is therefore the ability to discriminate, which requires knowledge and judgement. In his view, the most dangerous situation is when reports reinforce and magnify one another, and the decision-maker acts on a picture that appears consistent but is wrong. Translated to peacetime defence data reporting: it is not enough to pass on an internal report; it must be tested against an independent source, for example the payroll. The annual ÁSZ comparison in proposal 3.2 is the institutional form of this principle.

📖 Source: Carl von Clausewitz: A háborúról (English edition: On War)

6.4.3 Henry Kissinger: World Order

Writing about alliance systems, Kissinger stresses that states base their security decisions on a sober assessment of the real material and psychological balance of power, not merely on formal guarantees. A country’s strategy depends on how credible it considers its partners’ policies and how it sees the actual balance of forces (paraphrase). An alliance therefore works on the credibility of commitments, and this credibility depends on measurable capabilities. The headcount reported to NATO is exactly such measurable capability data: if a member state’s figures shrink by 7–12 per cent after the fact, partners weigh its other commitments more cautiously too. Hungary’s interest is therefore not a favourable figure but an accurate one.

📖 Source: Henry Kissinger: World Order

6.5 International comparison

According to the NATO comparison published by Telex, retroactive correction is not a Hungarian peculiarity: Turkey corrected its earlier data by 19 per cent and Finland by 16 per cent, and refining estimates is a standard procedure in the alliance. The difference lies in how transparent a correction is: whether it becomes clear what methodological change or real movement in headcount lies behind it. The model of programme point HV2, the annual performance report of the Swedish Defence Materiel Administration (FMV), shows that regular, public reporting of defence data is compatible with national security considerations. This is precisely the practical counterpart of Clausewitz’s critical approach and Allison’s organisational lesson: regular publication accompanied by methodology, which turns a correction into a normal, explicable event.

Defence

  • HV2 — Transparency of defence spending
  • HV10 — Modernisation of the reserve forces
  • HV13 — Multi-model crisis management decision-making

Transparency and anti-corruption policy

  • A1 — Public money dashboard
  • A2 — Public procurement transparency

Public administration and e-government

  • KI8 — Drucker-style efficiency measurement in public administration

Proposed new programme point: Defence data-reporting revision log and annual independent data verification — for the Defence area.

6.7 List of sources

Press sources (MIAK press monitor, 26 October 2026 — topic 4):

Knowledge base references (books):

  • 📖 Graham Allison – Philip Zelikow: Essence of Decision: Explaining the Cuban Missile Crisis
  • 📖 Carl von Clausewitz: A háborúról (On War)
  • 📖 Henry Kissinger: World Order

MIAK internal materials:

  • MIAK policy area: Defence (programme points; programme point ID: HV2, HV10, HV13)
  • MIAK policy area: Transparency and anti-corruption policy (programme points; programme point ID: A1, A2)
  • MIAK policy area: Public administration and e-government (programme points; programme point ID: KI8)
  • MIAK press monitor, 26 October 2026 — topic 4, score: 76/100

Supplementary public data sources (if used):

  • NATO: Defence Expenditure of NATO Countries — annual publication of the headcount and expenditure data reported by member states

Generation metadata