JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Immunization & Matching

You already know how to measure how hard rates push a value around. Now learn to build a portfolio that barely flinches — Redington's elegant three-condition vaccine for your surplus, the more literal safety of cash-flow matching, and an honest reckoning of what immunization quietly cannot do.

From measuring the wobble to surviving it

Earlier in this rung you learned to *measure* the danger. An insurer holds two piles of dated cash flows — assets it owns and liabilities it has promised — and when rates move, the two piles shift in value. The mismatch between them, boiled down to one number, is the **duration gap**: roughly the duration of the assets minus the duration of the liabilities. A gap of zero means a small rate move nudges both piles by the same percentage, and the surplus between them holds steady. That is the dashboard light of **asset-liability management**.

But measuring the gap is not the same as closing it. This guide is about *doing* something — deliberately arranging the asset pile so that, whichever way rates lurch, the change in asset value cancels the change in liability value. You are not betting on the direction of rates. You are building a position that does not much care which way they go. The grand name for that goal is immunization: vaccinating the balance sheet so a rate shock passes through it without leaving a wound.

Redington's three conditions

In 1952 the British actuary Frank Redington wrote down exactly what it takes to immunize a fund, and his recipe has three ingredients. The first is simply that the present value of the assets equals the present value of the liabilities — you have enough money on the table to begin with. The second is the duration-matching condition you already met: the duration of the assets equals the duration of the liabilities, so a small rate change moves both piles by the same first-order amount and those effects cancel. So far this is just "set the duration gap to zero."

The third condition is the clever one and the reason immunization beats plain duration matching: the convexity (the spread, the bow) of the assets must be *greater* than that of the liabilities. Recall from the duration-and-convexity guide that the value-versus-rate relationship is not a straight line but a gentle bend, and that convexity is your friend — it means assets fall a little less when rates rise and rise a little more when rates fall. If your assets are *more* convex than your liabilities, then after the first-order effects cancel, the leftover second-order wiggle works in your favour for a move in *either* direction. That is the whole magic: a small parallel shift, up or down, leaves you at least as well off as before.

There is a picture that makes the third condition vivid. Suppose you owe a single lump sum in 10 years — your liability has a duration of about 10. You could buy a 10-year zero-coupon bond and match it exactly. Or you could *straddle* the date: hold some 5-year bonds and some 15-year bonds, mixed so the average duration is still 10. The straddle is more spread out — more convex — than the single bullet. When rates fall, the long 15-year leg surges; when rates rise, the short 5-year leg gives you cash to reinvest at the new higher rate. Either way the straddle gains on one side roughly what it loses on the other, and the surplus is protected.

Redington immunization — three conditions

  (1)  PV(assets)         =   PV(liabilities)      enough money
  (2)  Duration(assets)    =   Duration(liabs)      gap = 0, 1st order cancels
  (3)  Convexity(assets)   >   Convexity(liabs)     2nd order in your favour

  Meet all three  ->  small parallel rate shift (up OR down)
                      leaves surplus >= its starting level.
The three conditions stacked up: enough money, matched first-order sensitivity, and a convexity edge that turns the leftover bend into a small bonus either way.

When you want certainty: matching and dedication

Immunization matches a *summary statistic* — duration — and tolerates buying and selling as the years roll by. There is a more literal, more stubborn alternative: **cash-flow matching**. For every payment you must make, line up an asset that pays you exactly the same amount at exactly the same time. Owe 100 next June? Hold something that delivers 100 next June. Do this for the whole liability schedule and the money to pay each bill is already on its way before the bill arrives. There is nothing to forecast and nothing to sell at an awkward moment.

A tiny worked example shows how clean it can be. Suppose you owe 50 in year 1, 50 in year 2, and 1,050 in year 3. Buy a single 3-year bond with a face of 1,000 paying a 5% coupon: it throws off 50, 50, and 50 + 1,000 = 1,050 — a perfect match, date for date. Now interest-rate movements become almost irrelevant. You are not selling anything along the way, so falling bond prices do not hurt you; you are not reinvesting maturing principal, so a low rate cannot bite you. Cash-flow matching neutralizes interest-rate risk and reinvestment risk *at the same time*, which is more than immunization can claim.

When this exact-matching idea is set up as a standing, ring-fenced strategy, it gets its own name: **dedication**. You carve out a specific bundle of bonds, *dedicate* it to one block of liabilities, and leave it to run off — touching nothing else. A dedicated portfolio funding 30 years of pension payments simply holds bonds maturing across all 30 years, so each year's pensions are met by that year's bond income. The word captures the spirit: those assets have a single job and are walled off from the rest of the fund. It is the most reassuring promise an actuary can make to a pensioner — the money for your cheque is already set aside, with your name on it.

The honest limits of immunization

Now the part a careful actuary never skips. Immunization is a beautiful approximation, not a guarantee, and Redington himself would be the first to list its fine print. The first limit is hidden in a single word: *parallel*. Duration assumes the whole yield curve shifts up or down by the same amount everywhere. Real curves do not move so politely — they *twist*. The short end can leap while the long end barely stirs, or the curve can steepen or flatten. A portfolio that is perfectly duration-matched can still spring a leak when the curve changes *shape* rather than *level*, because the straddle that protects you against a parallel move says nothing about a twist.

The second limit is the word *small*. Duration matching cancels the *first-order* effect, and convexity sweetens the leftover *second-order* effect, but both are local approximations to a curved relationship. For a gentle move of a few tenths of a percent the protection is excellent; for a violent jump of several percentage points the higher-order terms wake up and the neat cancellation frays. Immunization is calibrated for everyday weather, not for a once-in-a-generation storm.

The third limit is the word *once*. Immunization holds at the instant you set it up, but durations *drift apart on their own*. As time passes the liabilities age, as rates change the present-value weights inside each duration shift, and bonds roll down the curve — so a position that was matched on Monday is mismatched by the following quarter. The fix is unglamorous but essential: rebalance regularly, recompute the gap, and nudge the portfolio back. And it bears repeating that immunization addresses only *rate* moves; it does nothing about whether a liability turns out bigger than assumed (more claims, longer lives, fewer lapses) — that uncertainty lives on the liability side and is handled by reserving and capital, not by matching durations.

Putting it to work

  1. Value both sides on the same footing: discount the liability cash flows and the candidate asset cash flows off the same yield curve, and check that asset present value at least equals liability present value (Redington's first condition).
  2. Compute the duration of each side and form the duration gap; choose a bond mix that drives the gap to zero (the second condition).
  3. Among gap-zero mixes, prefer the one whose assets are more spread out — more convex — than the liabilities (the third condition), so a move either way leaves you no worse off.
  4. For the certain, fixed core of the book, go further and cash-flow match or dedicate it; reserve immunization for the larger, flexible remainder.
  5. Rebalance on a schedule, because durations drift; and stress-test against non-parallel twists and large shocks, which matching alone does not cover.

Notice what this does and does not promise. Done well, immunization and matching make a fund's surplus remarkably calm under the ordinary breathing of interest rates, and they let an institution keep its promises without having to forecast which way rates will go — a genuine engineering triumph. But they are tools tuned to one risk, calibrated for small and parallel moves, and good only until the next rebalancing. The wider job — twists, big shocks, equities, lapses, longevity, the whole tangled future at once — needs the broader machinery of ALM and the scenario-based modelling that the final guide of this rung turns to.