The Order and the Gap

In 2023, EU member states issued formal orders to leave to 484,000 third-country nationals. Fewer than one in five — about 91,000 people — were confirmed to have actually departed. The gap is not new, not a rounding error, and its causes are documented in the official record. But the figure also does not mean what it is usually taken to mean — and naming precisely what it does and does not mean is the task of this investigation.

1. The twenty-percent figure: what it actually is

The return execution rate is the ratio of confirmed effective returns to return orders issued in a given year across EU member states. Established The trajectory: approximately 15–16% in 2022, 19% in 2023, 24% in 2024, and a quarterly annualised pace above 30% in early 2026. The rate is rising — any account that freezes it at the 2023 trough misrepresents the trend.

Three caveats attach to the figure before it does any further analytical work. First: return orders are issued to all irregularly present third-country nationals — those who overstayed visas, those found without documentation, and rejected asylum applicants. Public Eurostat data does not disaggregate these populations. Second: of the returns that are executed, Eurostat Q1 2026 data shows 59.6% were voluntary departures — the person left following an order, without physical escort. Both voluntary and forced departures count in the headline statistic. Third: "effective return" explicitly excludes intra-EU Dublin transfers, where one member state sends a case to another for processing. A Dublin transfer does not remove anyone from EU territory.

With those caveats named: roughly 80% of return orders issued in 2023 — and still the large majority in 2024 — resulted in no confirmed departure. That is the confirmed finding. The investigation now asks why, and what the broader pipeline looks like.

2. The five-stage pipeline

The table below presents each stage of the EU asylum and return system as the primary sources record it. Each row carries its own denominator, year, and confidence label. Do not read across rows as a conversion chain. The wall between stages is named in Section 3.

Stage	Metric	Figure	Year	Source	Confidence
1 · Border detections	Irregular crossings detected, EU external borders	380,000–385,000	2023	Frontex / EUAA Asylum Report 2024	Established
↳ Spain: Canary Islands	Western African route detections	40,400 (2023) / ~47,000 (2024)	2023–24	EUAA / Frontex	Established
2 · Asylum applications	First-time applicants, EU27	1,048,900	2023	Eurostat DDN-20240325-1	Established
3 · First-instance decisions	Decisions issued; protection rate; rejection rate	674,125 total · 53% positive · 47% rejected	2023	Eurostat DDN-20240426-2	Established
4 · Appeal decisions	Appeal decisions issued; protection at appeal	191,530 total · 27% positive · 73% rejected	2023	Eurostat DDN-20240426-2; EUAA 2023 §4.14.1.2	Established (rates); appeal-uptake rate: data gap
5 · Return orders / execution	Orders issued; effective returns; execution rate	484,000 orders · 91,000 returns · 19% rate	2023	Eurostat Migration 2024 Interactive	Established

⚠ Definitional wall: each stage above has a different denominator and a different underlying population. They are presented side by side for scale and context; they cannot be divided against each other to produce a valid end-to-end ratio. See Section 3.

3. The definitional wall — why the stages don't chain

The most important structural fact about this pipeline is the one that prevents it from functioning as a pipeline analytically: no stage shares its denominator with any other, and no official agency has published a conversion factor linking them.

Start at the top. A Frontex border "detection" counts crossing events at the EU external border — not unique individuals. A person who attempts to cross three times is counted three times. The detection dataset and the asylum applications dataset are maintained by separate agencies in separate systems and are not linked. Established

The numbers make the gap visible. In 2023, first-time asylum applications (1,048,900) substantially exceeded Frontex irregular detections (380,000–385,000). This is not an accounting error. The majority of EU asylum applicants entered through regular channels — valid visas, family reunification, legal crossings — and applied subsequently. Applications in 2023 also include people who arrived in 2021 or 2022 but had not yet filed. Established

At Stage 3, a further mismatch: first-instance decisions issued in 2023 (674,125) do not correspond to applications lodged in 2023 (1,048,900). Processing backlogs mean the system runs with a multi-year lag; decisions in any year draw from applications filed over the prior two or three years.

At Stage 5, the universe shifts again. Return orders (484,000 in 2023) cover all irregularly present third-country nationals — including people who entered on a tourist visa, worked, allowed the visa to lapse, and were subsequently found. Such a person appears in Stage 5 data and does not appear in Stages 2–4 at all. The share of return orders specifically attributable to rejected asylum applicants is not disaggregated in public Eurostat releases.

The consequence: any sentence of the form "X% of people who arrive are eventually removed" would require dividing a Stage 5 figure against a Stage 1 figure — two populations measured by different agencies, in different systems, using different definitions, covering different years. The resulting number would not describe the EU asylum system. It would be an artefact of the arithmetic operation itself. This investigation will not perform it.

4. The nationality wedge

The EU-wide first-instance recognition rate of 51–53% is accurate. In the context of Spain's southern approaches, it is consistently misapplied.

The average is dominated by the three largest nationalities in 2024: Syrians (92% recognition), Venezuelans (89%), Afghans (81%). Established These nationalities account for a large share of total EU applications and structurally lift the aggregate rate. The nationalities prominent on Spain's Canary Islands and mainland southern routes — Algerians, Moroccans, Gambians, Senegalese — carry recognition rates substantially below the EU average. EUAA nationality-disaggregated data makes the spread unambiguous, though compiling per-nationality rates for each year requires the full EUAA Annex tables. Assessed

For Part 1: the EU-wide recognition rate does not describe the protection profile of the population most relevant to the series. Using it as a proxy for Spain's experience would be an error of application. Part 2 addresses the Spain-specific data directly.

5. Why orders go unexecuted

Of the ~393,000 return orders that went unexecuted in 2023, the EU's institutional record identifies a stable set of structural causes:

Readmission obstacles. Physical removal requires a functioning readmission agreement with the country of origin. Several major source and transit countries on the western routes have no such agreement with the EU, or have agreements that function poorly in practice. Bilateral negotiations are slow, politically sensitive, and often traded against other diplomatic interests.

Legal suspension. EU and member-state law mandates that removal be suspended while legal proceedings remain pending — first-instance appeals, further appeals, interim measures issued by national courts or the CJEU. A return order can be validly issued and simultaneously legally unenforceable for years.

Detention shortfall. Forced removal in most legal systems requires that the subject be held pending the removal. EU member states collectively lack detention capacity proportionate to the volume of outstanding orders. Some member states do not meaningfully detain at all beyond the voluntary-departure deadline.

Identity and documentation. Physical removal requires travel documents, which require establishing the subject's nationality. When a person arrives without documentation — common on sea-crossing routes — identification depends on source-country authority cooperation, a process that fails more often than it succeeds and can take years when it proceeds.

These causes are structural. They explain the slow upward trajectory in the execution rate — the EU has added bilateral readmission frameworks, return coordination capacity, and detention infrastructure incrementally since 2016 — and why the rate has not breached 40% in any recorded full-year figure.

6. What the data does and does not license

This investigation can state the following with primary-source support:

• ~19% of EU return orders in 2023 were executed; ~24% in 2024; 31.8% on a Q1 2026 quarterly basis. Established
• Roughly 80% of return orders in 2023 resulted in no confirmed departure. Established
• The EU-wide first-instance recognition rate of 51–53% is a weighted average that does not describe the nationality profile prominent on Spain's western arrival routes. Established
• The five pipeline stages are measured by separate agencies with separate denominators, in separate years, with no official published conversion factor between them. Established

This investigation will not state:

• An end-to-end arrival-to-removal ratio. The denominator populations are different and officially unlinked.
• What share of rejected asylum applicants is eventually removed. Return orders and asylum rejections are drawn from overlapping but non-identical populations that public data does not disaggregate.
• A verdict on whether European asylum policy is too permissive or insufficiently so. Part 1 has documented the system. The thesis requires more stages of the investigation.

Bottom line: The return execution rate — ~19% in 2023, ~24% in 2024, ~31.8% on a Q1 2026 quarterly basis — is the system's most precisely sourced figure, and it applies to return orders issued to all irregularly present third-country nationals, not to asylum applicants specifically. Roughly 80% of orders go unexecuted for documented structural reasons: readmission gaps, legal suspension, detention shortfalls, and identity failures. The EU-wide recognition rate of ~51–53% is real but does not describe the nationality profile on Spain's southern routes. The five pipeline stages cannot be arithmetically chained: each has a different denominator, a different year, and a different population, with no official link between them. These are not limitations of the investigation. They are the documented condition of the system. Part 2 turns to Ceuta, where the pipeline meets its most compressed form.

Prediction — logged in the Ledger

Part 2 of this series turns to the Ceuta crossing — where detection, pushback, application, and removal are compressed into a few hundred metres of border fence — and examines what the Spain-specific data shows about recognition rates and return outcomes on the western routes.