A zero is an observed value. Missing data means the source does not provide a usable value for that observation. Replacing one with the other can create a fact that the source never asserted.
This is one of Cameralius’s core evidence rules. It sounds elementary, but public datasets routinely contain blank cells, confidentiality states, not-applicable observations, suppressed values and special symbols alongside genuine zeros.
The short answer
If an official dataset reports 0, the source is normally making a numerical statement: the measured quantity is zero under that source’s conventions.
If the value is missing, unavailable or withheld, the source is making a different statement—or declining to provide one. You cannot safely turn that absence into zero without source authority.
0 ≠ missing ≠ confidential ≠ not applicable.
These states can look similar in a table. Semantically, they are different.
Eurostat makes the distinction explicit
Eurostat’s current database conventions distinguish a real zero from unavailable data. Its coding conventions define 0 as a real zero, while the special symbol : means that data are not available. Eurostat also uses separate observation-status and confidentiality-status flags for states such as missing, provisional and confidential.
Eurostat even distinguishes a true zero from a very small non-zero quantity: 0n means a value smaller than half of the displayed unit but different from a real zero.
That is precisely why a parser or analyst should not infer meaning from appearance alone. The metadata and source conventions are part of the evidence.
A Belgian example: missing can mean confidential
Statbel’s be.STAT tables provide a concrete Belgian example. In its turnover and investment statistics based on VAT declarations, Statbel explicitly notes that missing amounts are confidential.
A missing cell in that dataset therefore does not mean “nothing happened” or “the amount was zero”. It means the value is not being shown because of confidentiality.
Why replacing missing with zero is dangerous
Suppose a table contains expenditure for ten institutions and one observation is missing. If that missing observation is silently converted to 0, several downstream claims can become wrong:
- a total can be understated;
- an institution can appear to have spent nothing when the source merely withheld or lacked the value;
- a trend can show a false fall to zero;
- averages, shares and rankings can change;
- a later publication of the missing value can look like sudden new activity rather than the resolution of an information gap.
The reverse error is also harmful. Treating a genuine zero as missing removes information that the source actually provided.
Missing is not one single state
“Missing” is often only the beginning of the question. A value may be unavailable, confidential, not applicable, not yet reported, not quantifiable or suppressed for another reason. Some systems encode these states separately; others collapse several meanings into one symbol.
Cameralius therefore preserves the source state whenever it is known rather than flattening every non-number into one generic null.
What this does not mean
It does not mean every printed “0” can automatically be trusted as a factual zero. Some source systems use special display conventions, rounding rules or temporary placeholders. Cameralius still has to read the documentation for the specific source.
The rule is not “zero is always true”. The rule is: do not change a source state without evidence.
The same principle appears elsewhere in Cameralius
This is the same evidence discipline behind our explanation of why a Belgian federal budget is not the same thing as spending, why federal budgetary expenditure does not necessarily mean the money was already paid, and why the institutional perimeter behind “Belgian government” must be stated explicitly.
See the broader Cameralius methodology for how source observations, facts, derivations and analysis are kept distinct.
The Cameralius rule
Cameralius will preserve a genuine zero as zero. A missing, confidential, not-applicable or otherwise unavailable value will remain distinct when the source provides that distinction. If the source itself is ambiguous, the ambiguity should be preserved rather than guessed away.
Scope
This explainer concerns interpretation of Belgian and European official data rather than one specific fiscal dataset. Individual source systems can use their own symbols and conventions, so the authoritative metadata for each source remains controlling.
Sources & evidence
- Eurostat — Database coding conventions. Current conventions distinguish real zero (0), small non-zero values (0n), unavailable data (:), observation-status flags and confidentiality-status flags.
- Eurostat — Available data formats. Explains that a colon in a data cell means no data value is available.
- Statbel be.STAT — Turnover according to VAT returns. The dataset note states that missing amounts are confidential.
- Statbel — Turnover in the services sector. Current publication notes that missing indices are confidential.
Evidence review: 4 September 2026. Source-specific conventions remain authoritative over any generic rule.
Editorial status
Country: Belgium
Publication type: Explainer
Topic: Evidence & comparability
Last evidence review: 5 September 2026