Reporting thresholds — four designs, four consequences

“Are we above the threshold?” is a question with four different meanings depending on which return you are asking about, and the four are not interchangeable. Some thresholds decide whether you report at all. Some decide how often. Some decide whether what you report is shared onward. And some are not thresholds but derogations a supervisor grants. Confusing them is one of the most reliable ways to end up either over- or under-reporting across a multi-market footprint.

1. The four designs

2. What the threshold is measured on

Just as important as the design is the measurement basis, and three variants recur:

• Per item. The German EUR 12,500 exemption applies to the individual payment.
• Aggregated at a defined level. The AnaCredit threshold is EUR 25,000 on the debtor’s commitment amount, calculated as the sum of commitment amounts for all the debtor’s in-scope instruments in relation to the observed agent — not group-wide and not per instrument. Aggregating at the wrong level changes the population rather than a field.
• Across a period, not at a point. The same regime tests the threshold on any reporting reference date within the reference period, so an exposure that peaked mid-period and fell away is still in scope. Testing only at period end produces a short population, invisible in the file itself.

3. Derogations are asked for, not assumed

The payment statistics regime is the clearest illustration of how a derogation differs from a threshold. National central banks may grant derogations guided by the principle of proportionality to payment institutions meeting the conditions in Article 32(1) and (2) of PSD2, to e-money institutions meeting those in Article 9(1) and (2) of the e-money directive, and to other providers meeting both — and expressly regardless of whether the firm has actually been waived or exempted from prudential requirements nationally.

Two things follow. The test is the conditions, not the status: a firm that is not a registered small institution may still meet the conditions. And whether a derogation is available depends on a national-materiality test — under Article 4(3), among the alternatives, where the total value contributed by all providers that could benefit does not exceed 5% at national level for each of the listed services. So the answer is market-specific and it is a question to put to the national central bank rather than to infer.

4. Building threshold logic that holds up

Facts: a firm operating in several markets implements one threshold module that filters items below a value before they enter the reporting population.

What goes wrong: it is correct for the German scope threshold and wrong everywhere else. In a frequency-band regime it removes items that should be reported at a lower cadence. In a disclosure-threshold regime it removes items the supervisor is entitled to receive even though other lenders never see them. And it cannot express an aggregate-at-a-level test at all.

What the practitioner does: models the threshold as an attribute of the obligation rather than a filter in the pipeline, with four properties recorded per return: the design, the measurement basis, the aggregation level, and whether it is tested at a point or across a period. Those four fields are enough to implement every variant above, and their absence is why a single filter function never generalises.

FAQ

Related: ECB payment statistics · The Spanish ETE bands · German AWV reporting