Bill Miller

Amazon and Countrywide tested the same expectations-driven process, with opposite results

The Same Move Produced a Winner and a Failure

Bill Miller did not make technology stocks into value stocks by changing their label. At Legg Mason Value Trust, he asked what future a price already assumed and whether the market was too optimistic or too pessimistic about it. That approach could justify owning a company with few current earnings. It could also justify buying more of a declining company when he believed fear had pushed the quote below a defensible appraisal.

Amazon and Countrywide Financial expose both sides of that process. Amazon required Miller to value reinvestment and uncertain future economics that conventional multiples obscured. Countrywide required him to decide whether a collapsing share price represented panic or new information about credit losses and funding. He treated both as expectations problems. One strengthened Value Trust's return while the other helped produce a severe loss.

Value Meant the Expectations Inside the Price

In a 2005 letter to Value Trust shareholders, Miller separated business quality from investment merit. A strong company can be a poor investment if its valuation already assumes exceptional results. A weak company can be attractive if the price assumes an outcome worse than the one that occurs. The relevant comparison is therefore not growth stock against value stock. It is market expectations against a reasoned range of possible business outcomes.

This still requires valuation. Miller's 2006 commentary acknowledged that discounted cash-flow estimates become less reliable as the forecast horizon lengthens. His team used multiple methods rather than treating one long-range forecast as fact. The models exposed which assumptions about growth, margins, capital needs, and duration had to be true. Outperformance required those assumptions to be different from the market's and eventually more accurate.

The 2005 letter also described how the judgment entered a portfolio. The fund mixed companies Miller considered cyclically mispriced with those he considered secularly mispriced. Turnover had averaged roughly 15% to 20%, consistent with holding periods beyond five years. When a holding fell, the team generally bought more to lower its average cost. This made a correct appraisal more valuable, but it also made errors larger before the distinction was settled.

Amazon Required a Forward Operating Model

Miller's 1999 essay "Amazon and the Ethics of Belief" shows the problem before Amazon's later success was known. Current losses did not by themselves establish overvaluation because the company was spending to acquire customers, expand distribution, and enter new categories. The valuation depended on the scale Amazon might reach, the margins it might eventually earn, the capital required to get there, and the competitive position it might preserve.

Amazon's own 1997 shareholder letter confirms the operating policy behind that uncertainty: management said it would prioritize long-term market leadership and the present value of future cash flows over short-term accounting appearance. That statement does not prove Miller's investment was sound. It establishes what the company was trying to build and why a valuation based only on reported earnings would miss the central claim.

The fund record supplies a bounded outcome. In Value Trust's SEC-filed report for the fiscal year ended March 31, 2008, Amazon gained 79.2% and added about 280 basis points to the fund's return. That is not Amazon's lifetime return for Miller, and it does not isolate every purchase or sale. It does show a defined period in which the forward-economics case materially helped the actual mutual fund.

Countrywide Turned a Discount Into a Thesis Error

Countrywide documents the failure before hindsight could simplify it. In his third-quarter 2007 commentary, Miller wrote that the company was worth in the $40s while its shares traded around $14 to $15. He interpreted the decline as fear and compared the opportunity with earlier purchases of Amazon and AES. The comparison shows that buying Countrywide was not an unrelated departure. It followed the same belief that the market was overweighting visible distress.

The business condition was different. Mortgage delinquencies, credit losses, securitization markets, and short-term funding could alter Countrywide's cash generation and its ability to continue operating. A lower share price did not merely increase the implied return to an unchanged value. The information arriving with the decline changed the range of outcomes that needed to be valued.

The March 2008 fund report records the consequence. Countrywide and Bear Stearns together reduced Value Trust's fiscal-year return by about 480 basis points, while Sprint Nextel reduced it by another 290 basis points. Primary Class lost 23.86%, compared with a 5.08% loss for the S&P 500. Amazon's contribution partly offset the damage but could not make the portfolio immune to several related errors.

The Famous Streak Was Calendar Arithmetic

Value Trust beat the S&P 500 in each of the fifteen calendar years from 1991 through 2005. The claim is genuine, but Miller warned shareholders against giving it the wrong meaning. In both a January 2000 letter and the 2005 letter, he called the sequence an artifact of the calendar. By 2005, December-to-December was the only rolling twelve-month frame in which the fund had always led the index. His stated objective was performance over three to five years, not victory in every annual interval.

The distinction matters because a streak is a sequence of comparisons, not a causal explanation. Technology exposure helped in the late 1990s. Sector weights, portfolio scale, team decisions, shareholder flows, security selection, and chance in the chosen endpoint all affected the path. The streak does not prove that every holding was undervalued or that averaging down was safe.

Filed results also show how quickly the framing changed. At March 31, 2008, Primary Class had returned 3.91% annualized over ten years, 11.35% over fifteen years, and 13.94% since its April 1982 inception. The matching S&P 500 figures were 3.50%, 9.45%, and 12.81%. By December 31, 2008, Primary Class had lost 55.05% over one year; its five- and ten-year annualized returns were negative 12.13% and negative 4.21%. Its fifteen-year and since-inception figures remained positive at 6.60% and 11.09%.

Those December figures include fund expenses and reinvested distributions. They show a collapse severe enough to turn the ten-year record negative, but they do not support the simpler claim that every earlier gain was erased. Share class, start date, endpoint, fees, and benchmark determine what a performance number means.

Scale and Vehicle Belonged to the Method

Value Trust was a registered, daily-priced mutual fund, not Miller's personal account. It operated with public performance comparisons, multiple share classes, expenses, shareholder subscriptions and redemptions, and a research organization. Miller wrote in 1999 that the fund would not casually trade out of a billion-dollar position merely to improve the next twelve months. At that scale, price impact and liquidity narrowed the available response when a thesis weakened.

Portfolio risk also depended on shared causes, not only the number of names. Financial and housing-sensitive holdings could all deteriorate when credit availability, collateral values, and counterparties changed together. The fund could research each company separately and still accumulate one macroeconomic failure mode. The 2008 experience therefore tests portfolio construction as much as stock valuation.

The institutional boundary extends to attribution. Miller was the identified manager, but Value Trust's published return was produced by a team inside a mutual-fund vehicle. Public reports do not reveal a complete trade ledger, daily drawdowns, tax outcomes for each shareholder, or a clean decomposition of manager skill, factor exposure, and luck. Those omissions do not invalidate the record; they limit the claims that can be made from it.

What Amazon and Countrywide Establish

Miller's durable contribution was to make the expectations embedded in price the unit of analysis across conventional style categories. Amazon shows why backward-looking multiples can fail when a company is deliberately reinvesting for a different future. Countrywide shows why a lower price cannot be treated as proof that the expected return has improved. Financing and solvency can change the future itself.

The two cases also bound what transfers. Investors can make growth, margins, capital needs, and duration explicit; compare several valuation methods; and examine whether positions share one failure mode. They cannot reproduce Value Trust's research access, team judgment, scale, liquidity obligations, or ability to hold billion-dollar positions by copying the instruction to average down.

Miller Value Partners' current team page now presents the elder Bill Miller as the firm's legacy and identifies Bill Miller IV as chairman and chief investment officer from 2023. Keeping that current role separate from the historical Value Trust record preserves the central conclusion: a broad definition of value expanded the opportunity set, but only portfolio design and evidence-sensitive updating could control what happened when the market's pessimism was justified.

Inside CompanyGraph

The deep-value configuration is observable: companies priced well below book value while current assets exceed current liabilities by a wide margin and the equity ratio sits high for the industry.