Migdal Insurance: How Much Of 2025 CSM Is Truly Clean, And How Much Depends On Markets And Assumptions

What This Follow-Up Is Isolating

The main article argued that a return to profit does not settle the quality question by itself. This follow-up isolates one layer beneath that headline: the CSM, the contractual service margin, which is effectively a stock of future profit not yet recognized. On the surface, that number is very easy to read, so a 10% increase in 2025 can look like straightforward proof that future profitability simply became stronger. In Migdal's case, that reading is too simple.

The key question is not whether CSM grew, but what actually drove it higher. In 2025, total CSM rose from ILS 13.104 billion to ILS 14.411 billion. But new business contributed only ILS 438 million, while financial changes contributed ILS 1.753 billion and interest accretion added another ILS 276 million. At the same time, changes relating to future services reduced CSM by ILS 235 million, and the regular release reduced it by ILS 922 million. That is not a plain story of "more future profit." It is a story in which markets, assumptions, and measurement mechanics sit deep inside the number.

There is another reason not to read the increase too casually. The weakness was not theoretical. In life insurance, the increase in CSM came mainly from improvements in the stochastic model and from variable management fees collected above the risk-free rate, but part of that was offset by higher lapse rates in the new participating book. Those lapse rates were severe enough to reset the CSM balance for most of that book and to create a loss beyond the balance that had already been exhausted. In other words, in the same year that group CSM jumped, part of the book had already moved from future-profit stock into a loss component.

The bottom line of this continuation is straightforward: Migdal's 2025 CSM is real, but it is not all clean. The cleaner layer is accumulated-profit release, new business generation, and a slightly better release profile. The less clean layer depends on market returns, variable management fees in the legacy participating book, and model or assumption improvements that should not be treated as a base earnings run rate.

The 2025 CSM Bridge: A Good Number, A Less Clean Mix

Once the bridge is broken down, the picture becomes much clearer. The largest driver in 2025 was not new business, but financial changes. That line includes financial margin, inflation, yield-curve changes, returns above the risk-free rate, and financial effects on the stochastic component. In life insurance it also includes variable management fees in participating portfolios. That distinction matters because those drivers are not the same thing as clean business growth. They can reflect real economic value, especially when Migdal holds a very large legacy participating book, but they are also tied to market conditions and to measurement.

The ILS 276 million of interest accretion is not a new business engine either. It is mostly the mechanical effect of time passing on an existing CSM balance. So when someone sees a ILS 1.308 billion rise in CSM and translates it almost automatically into a similar improvement in the quality of future earnings, they miss the composition. A meaningful part of the movement came from layers that are not equivalent to better underwriting or stronger new sales.

This is also why Migdal's "additional value creation" metric needs to be read carefully. It reached 237% in 2025, but the numerator includes not only new-business CSM, it also includes the increase in CSM generated by excess returns in participating portfolios above the risk-free rate. That makes it a useful way to understand how strongly a good market year feeds into the book, but it is not a pure measure of clean underwriting quality. It measures market leverage as well as business generation.

The quality of each layer looks different:

That leads to the first conclusion: the 2025 CSM is a stock of future earnings with mixed quality, not a stock built mainly from new sales or pure underwriting improvement.

The Legacy Participating Portfolio: A Powerful Engine, But A Market Engine

To understand why 2025 looks so strong, the analysis has to move into the legacy participating portfolio. In policies issued before 2004, Migdal charges not only fixed management fees but also variable management fees equal to 15% of real returns after fixed fees. In policies issued from 2004 onward, by contrast, there are fixed fees only.

That distinction is critical. The same market does not create the same CSM everywhere in the book. When returns are strong, the legacy book can convert that environment into variable management fees, and those fees are not recognized directly in profit and loss. They are spread forward through life-insurance CSM. In 2025, Migdal recorded ILS 1.416 billion of variable management fees, versus only ILS 135 million in 2024. That is one of the most important numbers in the entire story because it explains how much of the CSM build depended on a very favorable market year.

This chart captures the point well. Ending CSM in the life-insurance segment rose to ILS 8.309 billion from ILS 7.199 billion. That is a material increase. But at the same time, variable management fees expanded by more than ten times. The right way to read 2025, then, is not only as a year of CSM growth, but as a year in which the legacy book monetized very supportive market conditions.

Migdal is also explicitly building on that mechanism going forward. Its own strategic framing says that variable management fees in the participating life-insurance book marketed in 1991 to 2004 are expected to keep growing and to add roughly ILS 400 million per year, on average, to CSM. That does not make the assumption wrong. It does mean management itself is signaling that continued CSM formation still depends in part on the legacy book continuing to generate excess returns.

That leads to the second conclusion: the legacy participating portfolio is a real strength, but it is not "clean" earnings quality in the simplest sense. It gives Migdal powerful leverage to a strong market year, and that is clearly valuable. But it is a different kind of quality. It is less like a predictable underwriting release and more like an embedded option on returns, curve conditions, and policy persistence.

The New Participating Portfolio: This Is Where Quality Breaks

If the legacy book is the layer that provides strength, the new participating book is the layer that shows how fragile CSM can become. Higher lapse rates in that book were enough to reset the contractual service margin balance for most of it and to create a loss beyond the balance that had already been reset. That is not a footnote-level accounting detail. It is an admission that groups of contracts in the newer book did not merely stop generating CSM, they crossed into a loss component.

This point did not appear once and disappear. The same pattern appears in the explanation of the fourth quarter and in the explanation of the full year. Migdal chose to flag it as one of the factors that reduced profit. In plain language, policyholder behavior weakened enough to erase the future profit already booked on part of the portfolio, and in some cases to move beyond zero into a loss position.

This is where two very different kinds of quality need to be separated. The legacy book is market-dependent, but at least it benefits from the variable-fee mechanism. The new book is less exposed to that mechanism because policies from 2004 onward carry fixed fees only. Yet that was exactly the part of the book where 2025 showed a lapse problem strong enough to wipe CSM out and create a loss component. So anyone trying to comfort themselves that the newer book is "cleaner" because it is less dependent on variable fees has to remember that its problem is not the market. Its problem is policyholder behavior.

That is what makes Migdal's CSM impossible to read in one layer. The legacy book is strong but market-linked. The newer book is less exposed to that market tailwind, and in 2025 it showed much higher sensitivity to lapses. Put the two together and the result is a larger CSM balance, but not necessarily a linearly better-quality CSM balance.

Same Markets, Different Economics

In 2025, net real returns in participating policies were strong in both books. In policies issued in 1992 to 2003, net real return was 9.9%, up from 7.9% in 2024. In policies issued from 2004 onward, net real return was 11.3%, up from 7.5% in 2024.

That sharpens the difference between market conditions and earnings quality. Both books enjoyed strong returns, but only the legacy book can translate that strength into variable management fees. In the newer book, the same supportive market backdrop did not prevent the lapse problem. The market helped Migdal, but it did not solve the persistence weakness in the part of the portfolio that moved into a loss component.

This may be the most important analytical point in the entire continuation. When an insurer holds a very large legacy participating book, it is easy to confuse any CSM increase with better business quality. In reality, a meaningful part of the increase comes from Migdal owning a historical book that can monetize a good market year. That is an advantage, but it is a conditional advantage. It does not erase the behavioral risk in the newer portfolio.

Release Pace: More Profit Ahead, But Still Mostly Long-Dated

One of the more attractive 2025 datapoints is the change in the release profile. The share of CSM expected to be released over the next five years rose from 28% at the end of 2024 to 29% at the end of 2025. In cash terms, expected release over the next five years increased from ILS 3.730 billion to ILS 4.202 billion, an improvement of ILS 472 million. That is a real positive, and it should not be dismissed.

But even here, the number needs to be read all the way through. If 29% will be released over the next five years, then 71% will still sit beyond that window. So yes, Migdal ended 2025 with a bigger stock of future profit. But most of that stock is still long-dated. A higher CSM balance improves the future-earnings base, yet it does not immediately turn that base into near-term, certain, or resilient earnings.

That matters even more once the source of the build is remembered. If a large part of the 2025 increase came from financial changes, variable management fees, and stochastic-model improvements, then a longer release profile is not automatically the same thing as a cleaner release profile. It simply means there is more to release. The quality of that release will still depend on persistence, returns, and assumptions remaining stable enough for that stock to actually flow through at the pace the market is tempted to read today.

So How Much Of 2025 CSM Is Actually Clean

The short answer is that a meaningful part of it is real, but a significant part of it still rests on conditions that should not be crowned as "clean" base earnings. The ILS 438 million from new business is the healthier part of the bridge. The ILS 922 million of regular release is accumulated profit being recognized over time. The move from 28% to 29% in the five-year release ratio is also supportive.

But against that, ILS 1.753 billion came from financial changes, ILS 276 million from interest accretion, and life-insurance CSM growth leaned mainly on stochastic-model improvements and variable management fees. At the same time, the new participating book already delivered the opposite signal: higher lapses that erased CSM and created a loss component.

So the right way to read 2025 is not "CSM rose, therefore future earnings became cleaner." It is more precise than that: Migdal exited 2025 with a larger stock of future profit, but the quality of that stock remained split in two. One layer is strong and useful, a large legacy book that can convert excess returns into CSM and extend the profit runway. The other layer is more fragile, a newer book where persistence already deteriorated enough to wipe CSM out and push contracts into a loss component. As long as those two layers coexist, Migdal's CSM remains an important asset, but not a number that should be read without separating its sources.