The Hidden Cost of an Inaccurate Claims Reserve
From the individual claim to the financial statements: why an estimate that fails to follow the risk can cost far more than the file containing it.
Take an €800,000 claim. Liability is uncertain, the amount sought decidedly ambitious and the file still young, but somebody must put a value on it, so the insurer establishes a €200,000 reserve.
Simply put, the reserve is an estimate of the claim’s ultimate cost. It is not the amount the claimant demands, nor necessarily what the insurer would pay today to settle. Above all, it is not a suitcase containing €200,000 sitting in a vault awaiting judgment. It is a forecast: the best possible on the information available at that moment.
A year passes—or two or three—and the file changes considerably. A court-appointed expert’s report seriously complicates the defence on liability but substantially reduces the loss. A supposedly decisive witness proves less memorable than expected. Another potentially liable party emerges. The claimant, still formally demanding €800,000, starts suggesting that, for a much lower sum, they might suddenly discover the virtues of settlement.
Yet the reserve remains €200,000.
It could be perfectly correct. But if so, it should be correct for different reasons from those of a year—or two or three—earlier.
This is the point I find interesting. Reserves have a remarkable ability to put down roots. Initially, €200,000 is an estimate. After a few months, it becomes “the reserve”. After a year, the number acquires almost an ontological dignity of its own. Curiously, changing it requires an explanation, while leaving it exactly where it is often seems the neutral choice.
It is not.
Maintaining a reserve means asserting, even implicitly, that everything learned in the meantime has not significantly changed the ultimate cost expected from that claim.
For an insurer, this is hardly marginal. One file matters relatively little; thousands make up the accounts. Reserves affect the presentation of the company’s financial position and performance, its underwriting result and solvency, and have tax consequences too. If systematically underestimated, reserve strengthening eventually follows—a rather elegant expression for saying that previous years’ claims are costing more than expected. If too high, reserves unnecessarily depress results and are later released when risk proves lower than anticipated.
Tax treatment is also less straightforward than sometimes suggested. Increases in non-life technical provisions affect taxable income under specific rules, including mechanisms deferring deductions for their long-term component. In other words, a reserve is not merely a technical assessment of a file: aggregated with others, it shapes how an insurer recognises today costs that will arise tomorrow.
That is why a reserve should be neither “prudent” nor “aggressive”. It should, as far as possible, reflect reality.
This is where the lawyer’s role becomes interesting.
We lawyers are accustomed to describing litigation in our own language: the expert’s report is favourable or unfavourable, our position has strengthened or weakened, sound arguments remain available to the defence. That last phrase, in particular, has accompanied generations of files with great dignity all the way to an adverse judgment.
Claims handlers, however, need something different.
Suppose that before the expert’s report we considered the probability of losing relatively low, but potential exposure in an adverse outcome could reach €500,000. The expert then concludes that the insured’s liability is difficult to contest, but the technically substantiated loss does not exceed €150,000.
Is the report unfavourable? In litigation terms, yes. Economically, the answer is much less obvious.
The probability of paying has increased, but potential severity has fallen sharply. Overall risk may even have decreased just as the legal position worsened.
Writing “unfavourable expert report” is therefore accurate but insufficient. The useful information is what that event changed in the distribution of risk: whether payment became more likely, realistic quantum fell, the worst-case scenario disappeared, potential recovery from a third party emerged, or settlement suddenly became more attractive.
This does not mean turning litigation into a mathematical model. I would regard with some suspicion a colleague who reported that, after the latest hearing, the probability of success had fallen from 63.2 to 58.7 per cent. Civil litigation retains a certain resistance to a second decimal place.
But between false precision and permanently retreating behind “litigation uncertainty” lies a very broad space. That is precisely where lawyers can produce information useful to those managing reserves.
Perhaps we should ask what this uncertainty actually contains.
Not just the judge.
It naturally includes how the judge will read the file, assess witnesses, interpret a clause or decide which account of the facts is more persuasive. But it also includes the court-appointed expert who sees what nobody else saw, the witness who remembers too much or too little, the document that emerges after we have built our theory of the case and, above all, a variable lawyers discuss with less enthusiasm: we may simply have got it wrong.
We may have overvalued evidence, misread a clause, placed too much weight on a precedent, underestimated a defence, imagined a witness would withstand questioning better, or devised a strategy that proves less brilliant in court than it seemed at our computer.
In other words, uncertainty is not just what we do not know about the judge. It is also what we do not yet know about the quality of our own assessment.
Perhaps that is why I find it unconvincing when success percentages are presented as external, objective facts, as though written somewhere in the file and the lawyer’s task were merely to read or correctly infer them.
That percentage contains us too.
It contains the quality of the available information, our ability to understand it, and the possibility that something tomorrow will show our reconstruction was wrong.
The useful question should therefore be not merely “Counsel, how do you see it?”, but “What has changed since the last assessment?”
There may be an even better question: “What could show that our current view is wrong?”
This requires the lawyer to identify not just the next procedural event, but the foundation on which the estimate actually rests. It might be an expert’s report, a witness examination, a decision on joining a third party under an indemnity claim, a Supreme Court ruling on a recurring issue, or a settlement offer about to expire. Not all procedural events are reserve-moving events; some principally increase the size of the PDF.
There is a further level. If our €800,000 claim concerns a clause appearing in ten thousand policies, the question is no longer whether one reserve should rise from €200,000 to €250,000. It is how many other files the same event might affect.
Sometimes a €100,000 case is worth millions—not because of its own cost, but because of what it teaches us about the portfolio.
This may be where lawyers become truly useful to a claims department: not by purporting to set the reserve in place of those responsible, but by recognising and explaining events that may render it outdated—and, above all, being willing to include among those events the discovery that they themselves were badly mistaken.
An insurer’s business, after all, is to price uncertainty. In litigation, it faces something harder still: valuing today a decision to be taken tomorrow about events that happened yesterday, using incomplete information and assessments made by human beings, all of whom can be wrong.
Including the judge. Including the lawyer.
Perhaps that is precisely what a good reserve should do: not eliminate uncertainty, which is impossible, but change quickly enough when uncertainty begins to show us where we were wrong.
Paolo Fortina · Originally published on LinkedIn on 8 September 2026. Read the original
Comments