One number I'm accountable for: the probability of a containerboard down-cycle over the next three years. The model turns that into a value. I update the number only when evidence lands, I say what the evidence was, and I write the entry before I look at the stock price that day. Public information only. Not advice; I may hold or trade IP, PKG, or SW.
2026-10-21v2 · the 10-K passmodel: IP_valuation_model.xlsx v2 · montecarlo.py seed 2026 · 21 checks
The pensions and leases came back clean. The footnotes didn't: $1.81 a share. The 30% stands.
price$32.00
my P(down-cycle)30%
mid-case value$38
P(value > price)62%
10th / 90th$16 / $54
odds the price implies44%
What happened. I pulled IP's FY2025 10-K to look for the debt that net debt leaves out. Pensions: the U.S. qualified plan holds $8,482M against $8,116M owed, and all plans together are $144M overfunded. I count the $342M owed by plans in deficit and $133M of retiree medical, after tax, and give no credit for the surplus. Leases: the $707M of operating leases are already inside EBITDA, and the $71M of finance leases are already in debt. The footnotes: a $487M deferred tax on the Temple-Inland timber notes, due when the notes mature in 2027, against $222M of equity left in the structure; $270M of environmental reserves for old mill sites and $103M of asbestos claims, after insurance; $48M booked for the Brazil tax guarantee to Sylvamo. Together, $957M after tax: $1.81 a share.
What I expected. A pension hole; paper companies usually have one. This one doesn't. The surprises were the timber tax and how much a century of mills leaves behind in reserves.
What changed in the model. A debt-like line on Inputs, built from the 10-K lines and taken off the equity value, not the WACC weights, since none of it is financing. Mid-case $40.25 → $38.44; P(value > price) 65% → 62%; the odds the price implies 48% → 44%. Two new tie-out checks, 21 in all, all pass. The post's numbers are now generated from the model, so they moved with it.
My P(down). Unchanged at 30%. The 10-K is about the balance sheet, not the cycle, and this log's rule is that the number moves only on evidence about the cycle. The edge narrows from 18 points to 14 points because the company is worth a little less in every scenario.
Not counted, on purpose. $344M of unrecognized tax benefits (timing unknown) and the unbooked $226M of IP's potential share of the Brazil assessment. They'd be another $0.65 and $0.43 a share if they landed in full.
Revised the same day: a first pass, written from the parts of the 10-K I had read by then, put the total at $1.18 a share. Reading Notes 10–21 added the environmental and asbestos reserves and the booked Brazil guarantee ($0.63). Entry sources: IP Form 10-K for 2025 (filed February 2026): MD&A pension table and contractual obligations, the balance sheet, Notes 10 (leases), 14 (contingencies), 15 (timber entities), 19 (retirement plans), Risk Factors.
2026-10-17v1 · thesismodel: IP_valuation_model.xlsx v1 · montecarlo.py seed 2026
At $32 the market is pricing a coin flip on the cycle. I think it's 30%.
price$32.00
my P(down-cycle)30%
mid-case value$40
P(value > price)65%
10th / 90th$17 / $56
breakeven P(down)≈50%
What I believe and why, in one paragraph: the 2025 closures (3.9 Mt, ~10% of North American capacity) were cost economics — the low performers went — so they're permanent, which supports price, but they weren't a demand squeeze, so the tightness is thinner than the headline. Of the $170/ton the index has recognized this year, about half sticks at the box level, with a lag of months; that's $590M of 2027 EBITDA net of what's already in guidance. EMEA is losing money on the paper-to-box lag and is for sale or spin. The dividend isn't covered by cash flow. Mid-cycle NA margin 17%, between SW and PKG. All of that nets to $40 a share on three methods, and a 65% chance the shares are worth more than the price — if the down-cycle probability is 30%. If it's 50%, there's no edge.
Position: none yet. Tuition-sized if any, after the Q3 call. Full reasoning in the valuation.
| What I'm watching | Why it matters | What would move my 30% |
|---|
| Fibre Box Association Q4 shipments | Demand is the whole bear case | Two consecutive down quarters → 40%+. Flat-to-up through Q1 → 25%. |
| Fastmarkets index, Oct and Nov | Does the September third round hold | Any give-back by December → "stick" drops to 35% and the mid case falls ~$4. |
| IP Q3 2026 call (late Oct) | Price realization vs the $175M in guidance; Pine Hill; H2 guide | Realized > $175M with the full-year guide held → stick to 60%. Another guide cut → 40% down-cycle. |
| DS Smith capital markets day, Nov 3 | The SOTP and any sale price | A bid above 6.5× EMEA EBITDA raises the SOTP toward $45; a spin with no bid leaves the market's ~1× implied value in place for a while. |
| OCC price | The cost side of the margin | Above $120/ton sustained → inflation assumption from $250M to $350M. |
| Dividend decision post-spin | The 5.8% is either real or a warning | A cut is already in the bear case; a hold with FCF < $500M would be a yellow flag on capital allocation, not a reason to change the value. |
Entry sources: IP Q2 2026 release; Packaging Dive (Apr 30, Aug 2025, Sept 18, 2026); MarketBeat Q4 2025 call; Investing.com Oct 15, 2026; Multiples.vc; Benzinga.
next entryafter the Q3 call
Template: date · model version · what happened (with the number) · what I expected · what changed in the model and why · the new P(down) · position, if any.