The results of the Kenya's August 2026 Infrastructure Bond Auction

Our July 2026 research note on Kenya's reopened infrastructure bonds took a deliberately cautious view of pricing. We argued that the recent long-dated Treasury bond auction pointed to a materially higher yield environment and estimated that the three infrastructure bonds could clear at yields between 13.75% and 14.25%, implying prices in the low-to-mid 90s.

The actual auction produced a very different result. The Central Bank of Kenya accepted KSh312.03 billion across IFB1/2019/016, IFB1/2021/018 and IFB1/2021/021, against KSh150 billion initially offered. Investors submitted KSh460.40 billion in bids, equivalent to a 307% performance rate. CBK's market weighted average rates were 12.3465%, 12.8150% and 13.1720%, and the weighted average rates on the bids it actually accepted were lower still, at 12.1960%, 12.6877% and 13.0520%.

In other words, our pre-auction yield scenarios were 108 to 140 basis points above the market average, and 120 to 155 basis points above the rates actually accepted, while the auction prices were substantially higher than the prices implied by our scenarios.

That is the gap as published. Section 7 shows that most of it is not a forecasting error at all, but an artefact of comparing a tax-exempt instrument against taxable ones.

This review is therefore not an attempt to reinterpret the original forecast after the fact. It is a straightforward forecast-versus-outcome assessment: what did we get right, what did we get wrong, and what should change in our approach to future Kenyan bond-auction analysis?

---

1. The original thesis versus the actual auction

Our original analysis was based principally on the most recent comparable long-dated CBK auction. The reopened 20-year and 25-year fixed-coupon bonds had cleared at accepted weighted average rates of 13.9234% and 14.4432%, respectively.

We used those results as the principal market anchors and interpolated across the maturities of the three infrastructure bonds.

That produced the following scenario:

Bond	Tenor	Coupon	Our Scenario Yield	Market WAR	Accepted-Bids WAR	Error vs Market	Error vs Accepted
IFB1/2019/016	9.3 years	11.750%	13.750%	12.3465%	12.1960%	+140.35 bps	+155.40 bps
IFB1/2021/018	12.7 years	12.667%	14.000%	12.8150%	12.6877%	+118.50 bps	+131.23 bps
IFB1/2021/021	16.2 years	12.737%	14.250%	13.1720%	13.0520%	+107.80 bps	+119.80 bps

CBK publishes two averages. The market weighted average rate covers every bid submitted; the weighted average rate of accepted bids covers only those CBK filled, and is 12 to 15 basis points lower on each bond. The original note's scenario is compared against both above. The rest of this review quotes the market rate, which is the series CBK's published price per KSh100 sits alongside.

The direction of the yield curve was correct that the longer-dated bonds will clear at higher yields. The level of the curve was not. Our scenario effectively placed the three bonds on a curve around 1.1-1.4 percentage points above the actual clearing levels.

---

2. The pricing call was further from the outcome

The original analysis translated the scenario yields into implied prices of approximately:

• 91.73 for IFB1/2019/016
• 94.21 for IFB1/2021/018
• 92.59 for IFB1/2021/021

The CBK auction results instead reported prices per KSh100 at the accepted average yields of:

• 101.8778 for IFB1/2019/016
• 104.4961 for IFB1/2021/018
• 103.9946 for IFB1/2021/021

Bond	Our Implied Clean Price	CBK Auction Price	Difference
IFB1/2019/016	91.73	101.8778	+10.15
IFB1/2021/018	94.21	104.4961	+10.29
IFB1/2021/021	92.59	103.9946	+11.40

You can reproduce the yield-to-price arithmetic behind both sets of figures with our Bond calculator, which converts a clearing yield into clean and dirty prices for a given coupon, tenor and settlement date.

The central pricing conclusion of the original note therefore did not materialise. We wrote that investors should budget for entry prices in the low-to-mid 90s. Instead, the reported auction prices were above KSh100 per KSh100 of face value.

Post-auction conclusion: in nominal terms the market paid materially more for these infrastructure bonds, and accepted materially lower yields, than our pre-auction scenario assumed. Section 7 shows that almost all of that gap disappears once the bonds' withholding-tax exemption is taken into account, and that on like-for-like tax terms investors actually demanded more yield than the comparable taxable bonds were paying.

---

3. Demand was not the problem

One of the most important lessons from the auction is that demand was exceptionally strong.

CBK received: KSh460.40 billion in bids against KSh150.00 billion offered. That represents a 3.07× coverage of the original amount offered. CBK ultimately accepted KSh312.03 billion.

The three bonds shared a single KSh150 billion offer, so CBK's per-bond performance rate measures each bond's bids against that combined target rather than against a bond-specific allocation. Its bid-to-cover ratio compares bids received with the amount accepted:

Bond	Bids Received (KSh bn)	Performance Rate	Amount Accepted (KSh bn)	Competitive (KSh bn)	Non-Competitive (KSh bn)	Bid-to-Cover
IFB1/2019/016	166.22	110.81%	112.64	70.99	41.65	1.48×
IFB1/2021/018	154.90	103.27%	105.54	55.15	50.39	1.47×
IFB1/2021/021	139.28	92.85%	93.85	41.91	51.95	1.48×
Total	460.40	307%	312.03	168.04	143.99	1.48×

One line in that table deserves more attention than it usually gets. Non-competitive bids made up 46.1% of everything CBK accepted, and the share rises with maturity: 37.0% on the 9.3-year bond, 47.7% on the 12.7-year, and 55.3% on the 16.2-year. Non-competitive bidders do not name a yield. They take the auction average. So on the longest bond, more than half the accepted amount expressed no view on price at all, which weakens any reading of the clearing yield as a considered market verdict on value.

The key point is that strong demand did translate into lower accepted yields in this auction. That contrasts with the conclusion we drew from the preceding long-bond auction, where strong demand coexisted with relatively firm yields.

The correct lesson is not that oversubscription is irrelevant to pricing. Rather, oversubscription alone is not enough to predict the clearing yield. The price investors are willing to pay depends on the interaction between demand, available supply, bond structure, comparable securities, portfolio requirements and the prevailing market yield curve.

---

4. CBK accepted far more than the original KSh150 billion target

Another feature that deserves attention is the amount ultimately accepted. The initial amount on offer was KSh150 billion, but CBK accepted KSh312.03 billion.

CBK's table applies approximately KSh118.14 billion to redemptions and records the remaining KSh193.89 billion under its own heading of "New Borrowing/Net Repayment". The two figures sum to the KSh312.03 billion accepted.

This matters because the auction should not be interpreted simply as a KSh150 billion fund-raising exercise. CBK was presented with substantially more demand and chose to accept a much larger amount.

That suggests the government had considerable room to raise additional funding without pushing accepted yields to the levels assumed in our original scenario.

It also helps explain why the auction should be read as a strong demand event rather than a marginally oversubscribed issue.

---

5. What happened to the yield curve?

The actual results produced a clean upward-sloping sequence:

Remaining Tenor	Market WAR	Accepted-Bids WAR
9.3 years	12.3465%	12.1960%
12.7 years	12.8150%	12.6877%
16.2 years	13.1720%	13.0520%

The spread between the shortest and longest of the three bonds was approximately 82.55 basis points. This is important because it means our basic interpretation of duration and the yield curve was not wrong. We correctly expected: longer maturity → higher required yield

What we misjudged was the absolute level at which that curve would sit. The post-auction curve can therefore be characterised as:

An upward-sloping curve, but approximately 1 percentage point lower than our scenario curve.

---

6. Why our comparable-bond approach failed

The biggest weakness in the original analysis was the weight placed on the July 2026 fixed-coupon bond auction as the primary pricing anchor.

The comparable bonds, set out in full in the original note, were:

• FXD1/2019/020, with an accepted WAR of 13.9234%
• FXD1/2022/025, with an accepted WAR of 14.4432%

Those were useful reference points, but they were not identical instruments.

The August securities were infrastructure bonds, with their own cash-flow structure and tax treatment. Most importantly, each bond amortises 50% of its face value before final maturity. That makes a simple maturity-to-maturity comparison with bullet fixed-coupon bonds imperfect.

The original analysis recognised the amortisation structure when discussing laddering, but it did not give that structural difference enough weight in the yield forecast.

---

7. The tax exemption explains almost the entire gap

Section 6 noted in passing that the comparables differed in tax treatment. That was not a footnote. It is most of the forecast error.

Interest on Kenyan infrastructure bonds is exempt from withholding tax. Interest on conventional Treasury bonds is not: for bonds issued with an original tenor of ten years or more, the rate is 10%. Both anchors used in the original note, FXD1/2019/020 and FXD1/2022/025, are taxable on that basis.

The original scenario therefore compared a tax-free yield against taxable yields and read the difference as a pricing gap. Putting the infrastructure bonds on a taxable-equivalent basis, by dividing the clearing rate by (1 - 0.10), gives a like-for-like comparison:

Bond	Our Scenario Yield	Market WAR (tax-free)	Taxable-Equivalent	Adjusted Error	On Accepted-Bids WAR
IFB1/2019/016	13.750%	12.3465%	13.7183%	+3.2 bps	+19.9 bps
IFB1/2021/018	14.000%	12.8150%	14.2389%	-23.9 bps	-9.7 bps
IFB1/2021/021	14.250%	13.1720%	14.6356%	-38.6 bps	-25.2 bps

The average absolute error falls from 122 basis points to 22 on the market rate, and from 136 to 18 on the accepted-bids rate. The conclusion holds whichever of CBK's two averages is used. The sign also reverses on the two longer bonds: they cleared at a higher taxable-equivalent yield than we assumed, not a lower one.

Set against the anchors themselves, the point is sharper still. IFB1/2021/018 has 12.7 years to run against FXD1/2019/020's 12.8, so no interpolation is needed at all:

Comparison	Taxable-Equivalent Yield	Taxable Comparable	Difference
IFB1/2021/018 (12.7yr) vs FXD1/2019/020 (12.8yr)	14.2389%	13.9234%	+31.6 bps
IFB1/2021/021 (16.2yr) vs interpolated FXD curve	14.6356%	14.1289%	+50.7 bps

Investors did not pay a premium for these bonds. Once the exemption is priced in, they required the same yield or more than the taxable curve was already paying, and demanded the largest concession on the longest bond. IFB1/2019/016 sits below the range spanned by the two anchors, so no reliable comparison can be drawn for it.

---

8. What the auction says about bond laddering

The original article correctly identified a problem with combining these three bonds into a conventional ladder. The amortisation dates cluster principal repayments around 2030-2031.

After that:

• IFB1/2019/016 matures in 2035.
• IFB1/2021/018 matures in 2039.
• IFB1/2021/021 matures in 2042.

Therefore, an investor holding all three does not receive a smooth annual return of principal.

Instead, the portfolio has: large principal-return events → relatively thin interim cash flows → final maturity events spread into 2042.

This is useful for investors who want a future liquidity event, but less useful for someone whose objective is to minimise reinvestment risk through evenly spaced maturities. This lesson remains valid.

To test how these three bonds behave together, our Bond Ladder calculator maps the combined principal and coupon schedule across the rungs, and the Bond calculator prices each leg individually. Because the amortisation returns half of each bond's face value early, the rate at which those proceeds are reinvested does much of the work in the final return.

---

9. Was IFB1/2019/016 the most attractive after the auction?

No, and the margin is not close. The relative-value question cannot be settled on yield alone, and with secondary-market durations now observable it does not have to be.

Bond	Market WAR	Mod. Duration	Yield Pickup vs Previous Rung	Extra Duration	Yield per Year of Duration
IFB1/2019/016	12.3465%	3.933	-	-	-
IFB1/2021/018	12.8150%	4.076	+46.85 bps	+0.143 yrs	+328 bps
IFB1/2021/021	13.1720%	4.632	+35.70 bps	+0.556 yrs	+64 bps

IFB1/2021/018 is the standout. Stepping up from the shortest bond buys 46.85 basis points of yield for 0.143 years of additional duration, a rate of roughly 328 basis points per year of duration taken on. Yields would have to rise about 328 basis points more than expected within a year before that trade stopped paying. The step from the middle bond to the longest is far less generous at 64 basis points per year of duration, though it still carries a breakeven of about 64 basis points and brings materially higher convexity.

The headline yield pickup of 82.55 basis points from the shortest to the longest rung costs only 0.699 years of duration in total, which works out at 118 basis points per year. Even the full extension is well compensated. What the data do not support is the idea that the long bond is the obvious pick simply because it yields most, or that it carries dramatically more interest-rate risk.

This overturns the original note's recommendation to overweight the shortest and longest rungs. On the duration-adjusted evidence the middle bond was the one to own, and the barbell we suggested was the least efficient of the three shapes available.

---

10. Where the bonds have traded since

The auction cleared on 17 August 2026. Secondary quotes nine days later, on 25 and 26 August, give the first independent test of whether it cleared at a sensible level.

Bond	ISIN	Auction Market WAR	Secondary Yield	Move	Clean Price	Mod. Duration	Convexity
IFB1/2019/016	KE6000005543	12.3465%	12.34%	-0.7 bps	98.977	3.933	27.236
IFB1/2021/018	KE7000003546	12.8150%	12.83%	+1.5 bps	100.977	4.076	34.819
IFB1/2021/021	KE7000005327	13.1720%	12.96%	-21.2 bps	100.901	4.632	45.973

Two of the three are trading within two basis points of where they cleared. The market has ratified the auction. Whatever our original note got wrong, the clearing levels were not an aberration that the secondary market has since unwound.

The exception is the informative one. IFB1/2021/021, the bond Section 7 identified as clearing at the widest concession to the taxable curve, is the only one to have rallied, tightening about 21 basis points in nine days. On the same tax-equivalent basis its premium over the interpolated FXD curve has narrowed from +50.7 bps at auction to +27.1 bps. The market appears to have reached the same conclusion Section 7 does, that the longest bond was the cheapest of the three, and has been closing the gap since. IFB1/2021/018 has not moved on that measure, sitting at +32.2 bps at auction and +33.8 bps now.

---

11. What we got right

A fair post-review should also recognise the parts of the original analysis that held up.

Correct: the yield curve should be upward sloping

The actual results confirmed that longer maturities required higher yields.

Correct: the amortisation structure matters

The three bonds create a distinctive cash-flow ladder, with a significant principal-return cluster around 2030-2031.

Correct: coupon is not the same as yield

The auction once again demonstrates that investors should not assess a bond simply by comparing its coupon with other coupons.

Partly correct: duration matters, but far less than maturity suggests

The 16.2-year bond is more exposed to changes in market yields than the 9.3-year bond, but not dramatically so. Modified durations are 3.933, 4.076 and 4.632, a spread of just 0.699 years across bonds maturing 9.3, 12.7 and 16.2 years out. Macaulay durations tell the same story at 4.415, 4.599 and 5.232.

The mid-life amortisation compresses all three into a much narrower band of interest-rate risk than their maturities imply. That is the strongest vindication of the original note's emphasis on the amortisation structure, and at the same time a caution against the shorthand we used ourselves: on these bonds, maturity is a poor proxy for risk.

---

12. The bigger market signal

The August auction provides an important signal about Kenya's domestic government securities market.

There was simultaneously:

• strong investor demand,
• substantial acceptance,
• nominal yields well below our scenario, though not below it once the tax exemption is priced in,
• and an upward-sloping yield curve.

Demand for government paper was unquestionably strong. What the auction does not show, once Section 7's adjustment is applied, is a market willing to fund the government more cheaply in real terms than comparable taxable bonds already implied. On like-for-like tax terms the government paid a premium of roughly 30 to 50 basis points over its own taxable curve to place the two longer bonds. The cheapness was in the tax code, not in the bid.

That the exemption is a transfer rather than a saving is easy to miss. Every basis point of yield the government avoids paying is a basis point of tax revenue it forgoes. The auction is better read as evidence of how much that exemption is worth than as evidence of improving funding conditions.

---

13. Final assessment

Our original analysis was close to right on required yield and badly wrong on price, and both follow from a single mistake: we benchmarked a tax-exempt instrument against taxable comparables without adjusting for the exemption.

On the numbers as published, the forecast missed by 108 to 140 basis points. On a taxable-equivalent basis it missed by 3 to 39, and in the wrong direction for the two longer bonds. The yield judgement, the shape of the curve, the importance of duration and the laddering analysis all survive intact. The pricing conclusion does not. We told investors to budget for entry prices in the low-to-mid 90s, and the bonds cleared between 101.88 and 104.50, because a tax-free coupon stream is worth more per shilling than the taxable stream we measured it against.

The arithmetic that would have caught this is a single step. Converting our own scenario yields into tax-free terms, by multiplying by 0.90, gives 12.375%, 12.600% and 12.825% against actual market averages of 12.3465%, 12.8150% and 13.1720%. That is a near miss on the first bond and a modest undershoot on the other two, and it would have put our implied prices at or slightly above par, which is precisely where the auction cleared.

The secondary market has since settled the question of whether the auction itself was mispriced. It was not. Nine days on, two of the three bonds trade within two basis points of their clearing yields, and the third has rallied towards the level Section 7 implies it should have cleared at. The error was ours, not the market's.

The most important lesson is therefore not that auction precedent is an unreliable guide. It is narrower, and more useful:

Comparable-bond analysis is only valid between instruments taxed the same way. Adjust for the tax basis first, then interpolate.

For Kenya's infrastructure bonds specifically, future analysis should gross taxable comparables into tax-free terms before any interpolation, and should treat amortisation, secondary-market liquidity and the tax status of the marginal bidder as pricing inputs rather than as closing commentary. On the relative-value call we would also now weight duration over maturity: the three bonds span 6.9 years of maturity but only 0.7 years of modified duration, and reading the first as a proxy for the second is what produced our barbell recommendation.

---

Sources. Auction figures are from the Central Bank of Kenya results for the re-opened Treasury bonds IFB1/2019/016, IFB1/2021/018 and IFB1/2021/021 dated 17 August 2026. Secondary-market yields, clean prices, durations and convexities are as quoted on 25 and 26 August 2026 for KE6000005543, KE7000003546 and KE7000005327.

Disclaimer: This article is a post-auction research review and should not be interpreted as investment advice. Auction outcomes reflect market conditions at the time of sale and do not guarantee future yields or prices.