Executive verdict
Inburgering exam pass rates differ sharply by level and component. The latest partial-year table we found reports A2 language-component rates from 74.5% to 85.4%, with speaking lowest. These are published component results, not a national probability of completing the whole integration process [1].
The latest complete NT2 reporting year is 2025. At B1, listening had the lowest pass rate, 43%; at B2, reading was lowest, 53% [2]. Those findings answer the question about the hardest components more accurately than a single ranking across all integration exams.
The central research finding is a measurement problem. An exam sitting, a person passing a component and a person meeting their integration obligation are different outcomes. DUO's cohort statistics can count exemptions based on other diplomas within the category for meeting the obligation [3]. Reading that category as an examination success rate produces misleading comparisons.
Our verdict: use the figures to identify possible preparation bottlenecks, then test your own skills. The evidence does not support a universal first-time pass rate, a current national average number of retakes, or a prediction of your individual chances. The strongest practical conclusion is to prepare each required component separately and leave room to recover from a failed sitting.
What does an inburgering exam pass rate actually measure?
A pass rate is useful only when its exam, population and denominator are clear. Before comparing percentages, establish whether they concern exam attempts, first attempts, eventual component passes or completed integration obligations.
Inburgering means civic integration. The Wet inburgering 2021 (Civic Integration Act 2021) took effect on 1 January 2022, while the previous law continues to apply to people covered by its transition arrangements. The government describes separate language, education and self-reliance routes [4].
Naturalisation creates another population. IND's current guidance says applicants generally demonstrate integration through an examination at A2, subject to exemptions and waivers [5]. A candidate sitting an A2 paper may therefore have a different legal purpose from someone following the municipal B1 route.
DUO also distinguishes route completion from examination participation. The zelfredzaamheidsroute (self-reliance route) can end with a municipal certificate without central language exams. Under conditions, someone in the B1 route may take language components at A2 instead [6].
These distinctions explain why seemingly incompatible statistics can all be correct.
| Published measure | What the denominator represents | What it cannot establish |
|---|---|---|
| Annual NT2 component pass rate | Results for the relevant component and reporting year [2] | The proportion of new learners who eventually obtain a diploma |
| DUO cohort completion status | People grouped by when their integration obligation began [3] | The probability of passing an individual exam sitting |
| Municipal integration progress | People within the relevant legal regime and migrant categories [7] | A like-for-like comparison of examination difficulty |
| CBS integration status | The dashboard's specified population and observation period [8] | A first-attempt language-component success rate |
DUO's public reporting page describes progress figures by municipality and distinguishes asylum status holders, family migrants and other migrants [7]. That is useful administrative information, but the breakdown itself does not make it an exam-attempt dataset.
CBS's current dashboard covers the new integration law through December 2025 and labels those annual results provisional. Its technical notes explain that populations differ between dashboard sections [8]. A percentage copied without that population note loses an essential part of its meaning.
Our analytical rule is simple: compare like with like before interpreting a difference. More time to complete an obligation can raise a cohort's completion share even if examination performance is unchanged. A cohort containing learners who have barely started cannot fairly be judged against one whose participants have already finished.
For readers establishing their own route, our inburgering guide provides the wider framework. The report here examines what the published numbers can support.
What are the latest A2 inburgering exam pass rates?
The latest A2 figures located are a partial-year snapshot in parliamentary answers from September 2026 [1][9]. Speaking has the lowest reported rate.
The following table preserves the published precision. The same answer includes B1 and B2, allowing a comparison within that particular reporting snapshot [1].
| Language component | A2 | B1 | B2 |
|---|---|---|---|
| Speaking | 74.5% [1] | 54% [1] | 55% [1] |
| Reading | 85.4% [1] | 42% [1] | 56% [1] |
| Listening | 83.3% [1] | 43% [1] | 59% [1] |
| Writing | 83.5% [1] | 59% [1] | 71% [1] |
The official document record identifies the parliamentary material; we read the answer's PDF through a public parliamentary mirror and checked its table against another transcription [1][9]. This matters because the evidence is a government answer, rather than a practice provider's estimate.
What the A2 table supports
Our calculation from the published values puts the gap between A2 reading and speaking at 10.9 percentage points [1]. This suggests speaking deserves particular attention when allocating preparation time.
The snapshot also puts B1 reading and B2 speaking lowest at their respective levels. That differs from the complete-year NT2 ranking below, illustrating why the reporting period belongs beside every comparison [1][2].
It does not show that speaking will be your weakest skill. A learner who uses Dutch confidently in conversation but seldom reads formal messages might experience the reverse pattern. National results help set an initial hypothesis; an individual diagnostic should decide the study plan.
What the table leaves unspecified
The answer does not supply component counts or label these rates first-attempt-only. It also does not state an exact extraction date for its year-to-date figures [1]. We therefore retain its limited description rather than assigning a more precise denominator.
The absence of counts prevents us from calculating uncertainty around the displayed differences. Nor can a partial-year snapshot reveal the final annual outcome.
This is especially relevant when searching for older A2 results. We did not verify a matching national component series for the earlier years requested in common searches. Substituting NT2 or overseas-exam figures would create an apparent trend across different examinations.
What were the B1 and B2 pass rates in 2022 to 2025?
The full-year NT2 tables show a deterioration at B1, particularly in listening, while B2 follows a different pattern [2][10]. These are results for Staatsexamen Nt2 (Dutch as a second language state examination), which supplies the language exams at those levels.
The earlier rows below were checked in the annual report for 2024 and against the historical rows reproduced in the following annual report [2][10].
| Programme and year | Speaking | Listening | Writing | Reading |
|---|---|---|---|---|
| B1, 2022 | 57% [10] | 65% [10] | 66% [10] | 53% [10] |
| B1, 2023 | 54% [10] | 62% [10] | 66% [10] | 55% [10] |
| B1, 2024 | 57% [10] | 52% [10] | 70% [10] | 53% [10] |
| B1, 2025 | 53% [2] | 43% [2] | 62% [2] | 44% [2] |
| B2, 2022 | 72% [10] | 60% [10] | 64% [10] | 59% [10] |
| B2, 2023 | 57% [10] | 62% [10] | 65% [10] | 56% [10] |
| B2, 2024 | 57% [10] | 56% [10] | 66% [10] | 54% [10] |
| B2, 2025 | 59% [2] | 60% [2] | 68% [2] | 53% [2] |
Our calculation shows B1 listening fell by 22 percentage points between the endpoints [2][10]. This is a change in recorded examination outcomes. It is not, by itself, evidence that the listening test became harder.
The earlier report already highlighted the listening decline and said its cause was unexplained [10]. The later report again reported no explanation for the B1 deterioration, while noting an increased share of resitters [2].
There are several possible mechanisms. A different candidate population, a different balance of first sittings and resits, or changes in preparation could affect the observed rate. These are hypotheses to investigate, not findings established by the tables.
Why B2 can have a higher pass rate than B1
A higher examination level can record a higher pass rate because its participants are different. Candidates are not randomly allocated to levels.
The reports describe programmes serving different work and study purposes, and explain the expansion of NT2 examination participation under the new integration system [2][10]. Comparing their percentages therefore combines differences in the test with differences in who takes it.
For a learner choosing qualifications, the relevant question is which level their purpose requires and which they are ready to demonstrate. Choosing an examination because its national percentage is higher would mistake a population statistic for a personal shortcut.
The reported average across components is also distinct from a complete-diploma pass rate. The annual tables do not isolate first attempts. Separate component results cannot tell us how many of the same people passed everything [2][10].
Which inburgering component is hardest?
The weakest published component changes with level and reporting period. The evidence supports a level-specific preparation priority, rather than a universal claim that speaking, writing or listening is always hardest.
The latest A2 snapshot points towards speaking; the complete-year B1 and B2 tables point towards listening and reading respectively [1][2]. Here, “hardest” means the lowest observed pass rate among the listed language components. It does not measure difficulty for equally prepared candidates.
What DUO's speaking rubric actually rewards
DUO's current A2 speaking model assesses adequacy and comprehensibility, vocabulary, grammar, fluency, coherence and pronunciation. An answer receiving zero for adequacy and comprehensibility also receives zero for the remaining assessment aspects [11].
The practical implication is more specific than “speak better Dutch”. An answer must address the prompt and be understandable. A rehearsed response that does not answer the actual question can fail at the foundation of the rubric.
The pronunciation descriptors allow an accent when speech remains understandable [11]. Our inference is that learners should prioritise intelligibility and successful communication before trying to remove every trace of their first language.
A useful practice routine is to record a response, check whether it answers every part of the prompt, then listen for passages a listener would struggle to follow. Ask a teacher to identify recurring problems rather than simply supply a corrected script.
Why writing feedback should start with task completion
DUO publishes separate A2 models for forms and for free writing or correspondence. Both make adequacy central, with a zero there carrying through to the other aspects [12][13].
The form model includes spelling and vocabulary. The free-writing model also includes grammar and coherence [12][13]. That distinction gives learners a practical way to request focused feedback.
For a form, check that the requested information is present and recognisable. For a message, check that the reader can understand the purpose, the relevant details and the requested action. Then review language accuracy.
These are recommendations derived from the rubrics. The pass-rate tables do not show which marking criterion accounts for failures, so we cannot claim that missing information or pronunciation is the leading national cause.
How many people need to retake the inburgering exam?
Historical DUO-based research demonstrates substantial repeat participation, but we could not verify a current national distribution of retakes. The government explicitly said the requested current-year breakdown was unavailable within its response period [1].
The most useful historical finding
A commissioned study of the old integration system recorded 35,079 successful candidates, 44% of its successful-candidate group, with a single attempt at each of the five components analysed together. The analysis excluded Oriëntatie op de Nederlandse Arbeidsmarkt (orientation on the Dutch labour market), or ONA, from that comparison [14].
That is meaningful evidence about repeat participation among a defined historical group. It is not a first-time diploma pass rate for everyone entering integration.
The selection issue is decisive. Looking backwards from people who succeeded omits those still working towards the qualification, those who stopped and those who followed other outcomes. It can describe successful pathways, but cannot directly estimate the probability that a new starter will succeed.
Resits can serve different purposes
For the study's group receiving an AGI-ontheffing (waiver based on demonstrable effort after an integration course), fourth-attempt pass rates were 19% for listening, 5% for writing and 4% for speaking [14].
Those rates apply to a group selected by its eventual waiver outcome. They should never be presented as the chances of an ordinary learner passing a fourth attempt.
The research also shows why administrative rules matter. A sitting may be taken while pursuing an effort-based waiver rather than because the candidate expects to demonstrate the required language level.
A source discrepancy deserves disclosure: the study describes 52% of that AGI course group as showing no improvement from the first to the fourth attempt; the minister's subsequent letter says the second to the fourth [14][15]. We follow the study's stated comparison and do not treat the letter as an interchangeable description.
Waiver rules are not a maximum number of attempts
The regulation reduced the relevant old-system effort-waiver requirement from four attempts to three per unpassed component, effective 1 April 2023 [16]. This changed an eligibility condition for a particular waiver.
There is an official-source inconsistency: the introductory text in DUO's progress PDF dated 1 January 2026 still describes four attempts, although the published amendment requires three [3][16]. We follow the amendment and treat that introductory wording as outdated.
It does not mean an ordinary learner must pass by that attempt or automatically receives a waiver after repeated failure. Other conditions and the applicable legal route remain relevant.
Likewise, free attempts are a funding rule. DUO's portal manual provides eligible asylum status holders with two free attempts per component at the level recorded in their personal plan, while family and other migrants do not have that entitlement [17].
For NT2, the official resit guidance says a failed component can be retaken, with registration possible once its result is known [18]. Build recovery time around the actual result and booking process, rather than assuming an immediate second sitting.
What is the inburgering exam pass score?
A pass score is the threshold for an individual result; a pass rate is the proportion of results meeting the threshold. Confusing these concepts can turn a national statistic into a completely incorrect examination target.
A component pass rate does not say what proportion of questions an individual must answer correctly.
A2: why a universal raw percentage is unreliable
An official historical explanation shows that A2 speaking cut-offs were calculated per exam version using the difficulty of the questions. It describes a rounding error that wrongly failed 1,584 candidates. The minister said their results would be corrected [19].
That incident establishes a narrow but important point: assessment thresholds and their implementation matter. It also illustrates why a remembered rule about correct answers from an old version should not be treated as a permanent national standard.
The historical explanation concerned the then-existing receptive speaking section [19]. We do not transplant its question counts or cut-offs into today's examination.
For the current A2 speaking and writing exams, we verified assessment models but could not verify a universal published raw-score conversion or minimum number of correct answers [11][12][13]. A rubric explains what is assessed; it does not necessarily supply the final conversion into a pass result.
A learner seeking a “safe minimum” should therefore distinguish an officially documented threshold from a practice website's suggested target.
B1 and B2: the published threshold is a scale score
The official NT2 candidate FAQ states that a score of at least 500 is required. Raw points are converted to a scale score, and the possible score ranges differ by component and examination [20].
Consequently, that threshold cannot be read as a percentage of correct answers. It also does not imply that candidates compete for a fixed quota of passing places.
NT2's assessment page confirms that reading and listening are marked by computer, while speaking and writing are assessed by trained assessors, with two assessors judging those productive exams [21].
For preparation, this suggests different kinds of feedback: analyse incorrect choices in receptive tasks, and obtain rubric-based feedback on spoken and written responses. A single overall practice percentage can conceal a weak component that remains essential to the qualification.
Are overseas inburgering pass rates comparable?
No. The basisexamen inburgering buitenland (basic civic integration examination abroad) is a separate A1 examination, with a different population and component structure [22].
The DUO-based monitor reports the following explicitly first-attempt results [22].
| Overseas component | 2022 | 2023 | 2024 |
|---|---|---|---|
| KNS, knowledge of Dutch society | 96% [22] | 95% [22] | 96% [22] |
| Speaking | 73% [22] | 75% [22] | 76% [22] |
| Reading | 78% [22] | 80% [22] | 76% [22] |
The monitor changed its reporting definition to include first component attempts by candidates who had not yet taken every component. It also records a reading-format change on 1 May 2023 [22].
Both points matter when comparing these results with older monitors. An apparent disagreement may reflect a different group being counted or a changed examination structure.
The overseas exam's official guidance says each component is graded on a scale from 1 to 10, with at least 6 required in every component. Candidates retake only the failed component or components [23].
Those clearly published rules answer an overseas-exam question. They do not establish a raw-percentage threshold for domestic A2 papers.
Location searches also need care. Overseas results concern candidates taking the basic exam at embassies or consulates [22]. They cannot establish whether a domestic A2 exam is easier in a particular Dutch city, nor whether learners from a country have better intrinsic prospects.
Our recommendation is to keep overseas preparation and domestic integration preparation in separate folders, with the correct exam name on every resource. Similar terminology is not sufficient evidence that materials or statistics transfer.
What this means for learners
Use the national figures to decide what to check first, then let your own performance determine what to practise. Passing the required components is a sequence of separate demonstrations, so an average impression of being “good enough at Dutch” is a weak planning tool.
Confirm your actual requirement. Read your DUO record and, where applicable, your persoonlijk plan inburgering en participatie (personal integration and participation plan). Distinguish a municipal route from a naturalisation requirement, and check exemptions before booking [5][6].
Diagnose every required skill. Work under realistic conditions and record what went wrong: comprehension, vocabulary, incomplete answers, response organisation or handling the task. Use an unfamiliar exercise when repeating the diagnostic so that remembering the answer does not masquerade as improvement.
Give productive skills observable feedback. For speaking, save recordings and ask whether the answer is understandable and responsive. For writing, ask the reviewer to mark missing task information separately from language errors. The published A2 rubrics provide a concrete basis for that conversation [11][12][13].
Practise the official NT2 environment. The examination authority recommends its instruction videos and practice environment, which works properly on a computer or laptop. It also explains that real speaking and writing exams contain tasks beyond the published practice assignments [24].
Our NT2 preparation guide can help organise this work. If you need feedback on output, compare Dutch tutors by their ability to assess the relevant exam tasks, rather than a claim about a national success percentage.
Plan for the result before the resit. DUO does not allow NT2 registration for a component while its previous result is outstanding. Candidates needing adjustments for a disability or personal circumstances should identify that need when registering [25].
Change the preparation after a failure. Preserve a short error log, choose a specific weakness and practise it before repeating a full task. A new booking creates another opportunity; it does not itself correct the underlying problem.
Read providers' success claims critically. Ask whether the denominator is all enrolled learners, those entered for the examination or eventual completers. Ask whether repeat attempts are included, which components are covered and what period is measured. Without those answers, a school's percentage cannot be compared fairly with the national tables.
What we could not verify
We could not verify a complete national A2 component series matching the earlier annual searches, or an overall first-attempt diploma pass rate on a consistent current basis. The latest parliamentary component table lacks published component counts and an exact extraction date [1].
We could not verify a current national mean or median number of attempts, or a distribution showing how many people passed at each attempt. The government's answer explicitly left the requested resit breakdown unavailable [1].
We did not verify a universal current A2 raw-score threshold for speaking, writing, reading or listening. The assessment models and historical explanations should not be converted into an invented number of correct answers [11][12][13][19].
Finally, the evidence does not establish a causal explanation for the B1 decline, a reliable ranking of exam locations, or an individual success probability based on nationality or first language. The annual reports identify unresolved questions rather than a demonstrated cause [2][10].