Part I — Situation overview

On 30 June 2026 the Hungarian statistical and social-science profession turned jointly to the Central Statistical Office (KSH — the institution that publishes the official statistical data). Three committees of the Hungarian Academy of Sciences, the Hungarian Sociological Association and the Hungarian Economic Association asked, in a joint position, that the office reveal in detail the circumstances in which the income and poverty indicators that proved faulty had arisen, that for the duration of the investigation it temporarily remove the faulty indicators from the domestic and Eurostat databases, and that it build institutional frameworks that prevent similar errors in the future (Telex, Portfolio, 30 June 2026).

The core of the scandal lies in the Hungarian data of the EU-SILC survey (the EU income and living-conditions survey). Researchers had already demonstrated serious problems in early 2025: households were visibly, artificially clustered in the band just above the poverty threshold — this biased the measured poverty rate downwards — and a significant part of them showed uninterpretable tax data (a tax credit instead of a tax payment). The KSH launched a review in November 2025, then carried out a correction in March 2026; but according to the researchers, after the correction the income changed for almost 100 percent of households, while a large part of the original errors remained untouched or worsened further (Portfolio, 30 June 2026). In the background there is a paradox that also divides public opinion: the income-based and the consumption-based indicators paint a contradictory picture of whether Hungary is one of the EU’s poorest countries, or whether poverty has in fact fallen (HVG, 25 June 2026; the article was not publicly downloadable).

In MIAK’s reading this is not a one-off technical mishap but an institutional risk touching one of the cornerstones of data-driven governance. Faulty official statistics are, after all, not neutral: they distort social-science research, mislead policy, and undermine the public’s trust that decisions are born on the basis of facts — not mood. The character of the problem is therefore not that an office made a mistake, but that the finding, admission and correction of the error were slow and opaque.

Part II — Literature foundation

Before turning to MIAK’s proposals, it is worth fixing the scientific frame. The World Bank’s report World Development Report 2015 — Mind, Society, and Behavior points out: not only citizens but the decision-makers and the offices themselves are exposed to systematic biases, and therefore institutions must build in processes to filter out errors — the quality of measurement and the review of the measurement procedure are preconditions of evidence-based policy. The European Commission’s European Semester dossier Addressing Inequalities systematises the metrics of income and opportunity inequality (the S80/S20 ratio and the Gini index), and makes clear: the picture formed of poverty and inequality is worth only as much as the data behind it — an error in the metric directly misdirects the policy conclusion. The common lesson of the two sources is that statistics are not a passive mirror but an active decision infrastructure whose quality must be guaranteed institutionally. The detailed literature treatment — by source, with quotations — can be found in section 6.4 Literature in detail.

Part III — MIAK’s concrete proposal

MIAK proposes three measurable measures that together restore the verifiability of official statistics without harming the institutional independence of the KSH.

3.1 An independent data-quality council alongside the KSH (within 90 days)

MIAK proposes the setting up of an independent, professional data-quality council that is organisationally separate from the KSH but carries out a methodological review before the publication of the priority indicators (income, poverty, inflation, employment). The council’s members would be delegated by the academic committees, the Hungarian Sociological Association and the Hungarian Economic Association, with a rotating, public mandate. An important legal clarification: the KSH is an autonomous, professionally independent office — the council cannot instruct it or override the office’s decisions, but gives a public opinion and methodological observation to which the KSH must respond with a duty to give reasons. This is the direct institutional realisation of the “bias-reducing process” recommended by the World Bank’s WDR 2015 (see 6.4.1): it entrusts the quality of measurement not to the self-review of a single organisation but to an external, credible professional filter. The council extends the principle of G19 — the public reasoning and mandatory review of economic decisions — to data production itself.

3.2 A public error log and correction protocol (by Q4 2026)

MIAK proposes that the KSH keep a public error log: for every identified data error it should publish when and how it arose, when it was detected, what correction was made, and to what extent the correction modified the earlier time series. The present case shows precisely that correction in itself is not enough — if after a correction the income changes for almost 100 percent of households while the original errors remain, then the opacity of the correction itself causes a loss of trust. The error log is not a punishment for error but a tool of learning: it carries the ex-post impact-assessment logic of G20 — the comparison of the expected and the actual result — over to statistical production. The metric focus of the European Commission’s Addressing Inequalities dossier (see 6.4.2) becomes operational here: if the poverty rate is faulty, it must be flagged and corrected openly before it enters an EU comparison.

3.3 Open anonymised microdata for researcher verification (by 2027)

The third proposal is the preventive side of error detection: the KSH should make the anonymised microdata available in a secure, accredited researcher-access system. The present errors were noticed precisely by external researchers — the artificial clustering around the poverty threshold and the uninterpretable tax data — because they had access to the detailed data. The many-eyes principle works in statistics too: the more independent analysts see the anonymised microdata, the faster the systemic bias comes to light. This is the natural intersection of the open-data programme of D2 and the inequality monitoring of G7 — the balance of data-protection anonymisation and researcher openness is ensured by the differential-privacy standard.

The common principle of the three proposals is that the quality of official statistics is entrusted not to the office’s good will but to institutionalised, public and verifiable mechanisms — building in precisely the external filter that the literature frame identifies as a precondition of an evidence-based decision.

Part IV — Expected impacts and risks

Dimension Expected impact Risk
Economy More reliable income and poverty indicators; more precisely targeted social spending Time series dropping out for the duration of the investigation may cause uncertainty in planning
Society Restored trust in official data; a real picture of poverty In the short term the scandal may damage the credibility of all KSH data, even where there is no error
Public administration Institutionalised error detection, faster correction The independent council may become formal if there is no real duty to give reasons

The main consideration is the balance of independence and verifiability. The professional independence of the KSH must not be harmed — the office answers autonomously for the methodology, and this may be overridden by no external council or political actor. At the same time independence cannot mean unverifiability: the public review and the error log strengthen precisely the legitimacy of independence, because they prove that autonomy does not conceal opacity. The proposal tips to the risk side if the independent council becomes merely a rubber-stamping body without a real duty to give reasons, or if the error log is used for the destruction of trust rather than for learning — which is why it is critical that the tone of the error log stay professional and correction-oriented.

Part V — Measurability and summary

5.1 What is worth tracking? (suggested KPIs)

The following performance indicators (KPIs) will show in 6–18 months whether the direction is successful. These are proposals, not government decisions — it is worth tracking them:

  • Whether the detailed, public revelation of how the faulty EU-SILC indicators arose appears by the end of 2026;
  • Whether the independent data-quality council has been set up, and before how many priority indicators’ publication it carried out a methodological review;
  • Whether a public KSH error log is available, and whether the turnaround time of corrections (from detection to correction) falls;
  • Whether the number of open datasets available through researcher microdata access grows by 2027.

5.2 Summary

MIAK asks the decision-makers and the KSH to close the present scandal not with defensiveness but with institutional learning: an independent data-quality council, a public error log and open microdata — these are concrete steps that can be launched within 90 days. Correcting the faulty data is necessary, but not enough; the goal is that the next error come to light quickly and visibly.

This topic is at the core of two MIAK foundational values. Data-drivenness moves here because the whole MIAK governance philosophy is built on the decision resting on measurable, verifiable facts — if the measurement itself is faulty and the error stays hidden, data-driven governance turns into its own opposite. And transparency, because official statistics fulfil their role if they are publicly verifiable: the open error log and researcher microdata access do not weaken but authenticate the independence of the KSH.

Part VI — Justifications and further sources

6.1 Press framing by spectrum

Liberal-left and public-affairs band (Telex, HVG). Telex frames from the side of the professional organisations: the emphasis is on the united action of the research community and the concrete requests formulated to the office (revelation, temporary withdrawal, institutional guarantee). It highlights the harm to science and to evidence-based policy. The HVG /360/ (behind a paywall) material presents the topic in a broader, explanatory frame: it circles the apparent contradiction — are we the EU’s poorest country, or has poverty fallen? — that is, it focuses on the divergence of the income and consumption indicators, not primarily on institutional responsibility.

Economic band (Portfolio). Portfolio is the most technical: it names the two concrete data errors (the artificial clustering around the poverty threshold and the uninterpretable tax data), and places the events in chronological order (early-2025 signal → November-2025 review → March-2026 correction → 30 June 2026 position). The framing emphasises the united action of the profession (“the entire profession urges an investigation”).

Governing-party/conservative band (Magyar Nemzet, Mandiner). The conservative band pointedly did not bring the topic into the top focus on this day — which is in itself telling: the political reading of an official data-quality matter is band-dependent, and MIAK’s goal is precisely that the question be treatable ideology-free, as a technical-institutional problem.

6.2 Facts and data

  • EU-SILC is the common EU data collection of income and living-conditions statistics, whose Hungarian results also enter the Eurostat database — so the faulty indicator distorts not only the domestic but the EU comparison as well.
  • The signalled error types: (1) artificial clustering above the poverty threshold, which biases measured poverty downwards; (2) uninterpretable tax data (a tax credit instead of a tax payment) for a significant part of households (source: Portfolio, 30 June 2026).
  • The S80/S20 ratio (the quotient of the annual income of the richest and poorest 20 percent) and the Gini index are the main EU metrics of income inequality; the EU-average S80/S20 value was around 5.1 (2015) — the Hungarian figure is compared to these, which is why an error in the Hungarian value directly distorts the EU comparison (source: European Commission, Addressing Inequalities).

6.3 Policy aspects

  • Economy (programme points) — the data-driven budget and the radical transparency of economic decision-making stand or fall directly on the quality of the official data;
  • Social policy (programme points) — the precision of targeted support and of poverty monitoring depends on the reliability of the income and poverty data;
  • Digitalisation and AI regulation (programme points) — the open-data programme and researcher microdata access are the technical infrastructure of error detection.

6.4 Literature in detail

6.4.1 World Bank: World Development Report 2015 — Mind, Society, and Behavior

One of the report’s central theses is that not only citizens but experts and offices too are exposed to systematic cognitive biases, and therefore institutions must consciously build in processes to filter out errors. The report puts it thus:

„Development professionals and policy makers are, like all human beings, subject to psychological biases. Governments and international institutions […] can implement measures to mitigate these biases, such as more rigorously diagnosing the mindsets of the people we are trying to help and introducing processes to reduce the effect of biases on internal deliberations."

The KSH case illustrates precisely this: the error is not individual bad intent but systemic measurement bias, which only an external, institutionalised review — the independent data-quality council proposed by MIAK — can reliably filter out. Guaranteeing the quality of measurement depends not on the self-review of the single producer but on a built-in, credible filter mechanism.

📖 Source: World Bank: World Development Report 2015 — Mind, Society, and Behavior

6.4.2 European Commission: European Semester Thematic Factsheet — Addressing Inequalities

The dossier systematises the two dimensions of inequality — outcome (income and wealth) and opportunity inequality — and fixes the main metrics. Speaking of the inclusive character of growth, it notes:

„When the income produced in a country, as measured by GDP, is growing faster than the incomes received by that country’s households, this suggests that growth is not inclusive and that its benefits are not being felt by all households."

For the Hungarian EU-SILC matter this is the key: if the data measuring household income is itself faulty — an artificial clustering around the poverty threshold — then it is precisely the indicator needed to judge inclusive growth that becomes unusable. A reliable metric is not a luxury but a precondition of the policy conclusion; a faulty poverty rate, raised into an EU comparison, misleads several times over.

📖 Source: European Commission: European Semester Thematic Factsheet — Addressing Inequalities

6.5 International comparison

Several EU patterns exist for the institutionalisation of the independence and quality of official statistics. Eurostat’s European Statistics Code of Practice fixes the principles of professional independence, methodological soundness and publicity — MIAK’s proposal would strengthen domestic practice in line with these. France’s Observatoire des inégalités model shows that an independent, public inequality report makes the public-policy debate evidence-based; and researcher microdata access operates in several EU member states in the form of accredited “safe centres”, where anonymised data can be analysed without data protection being harmed. These practices confirm that openness and independence can be strengthened not at each other’s expense but together.

Economy

  • G1 — Data-driven budget
  • G19 — Radical transparency in economic decision-making
  • G7 — Wealth-inequality monitoring
  • G20 — Economic-policy impact-assessment system (Drucker audit)

Social policy

  • SZ1 — Targeted support
  • SZ9 — Public monitoring of wealth inequality

Digitalisation and AI regulation

  • D2 — Open-data programme

Suggested new programme point: An independent data-quality council alongside official statistics — for the Economy area.

6.7 Source register

Press sources (MIAK press monitor, 1 July 2026 — top-10 topics):

Knowledge-base references (literature):

  • 📖 World Bank: World Development Report 2015 — Mind, Society, and Behavior
  • 📖 European Commission: European Semester Thematic Factsheet — Addressing Inequalities

Note: the local file path of the books does not appear in the blog’s visible text — only the author and the title.

MIAK internal materials:

  • MIAK policy area: Economy (programme points; programme point ID: G1, G7, G19, G20)
  • MIAK policy area: Social policy (programme points; programme point ID: SZ1, SZ9)
  • MIAK policy area: Digitalisation and AI regulation (programme points; programme point ID: D2)
  • MIAK press monitor, 1 July 2026 — topic 5, score: 78/100

Additional public data sources (where used):

  • Eurostat — EU-SILC (EU income and living-conditions survey); European Statistics Code of Practice

Generation metadata