Net worth grew by $152.6 million in 1984, or $133 a share. Most managers would have that figure framed. Buffett files it under mediocre.

Book value per share has compounded at 22.1% a year for twenty years, from $19.46 to $1,108.77. The 1984 gain was 13.6%. "This sounds pretty good but actually it's mediocre. Economic gains must be evaluated by comparison with the capital that produces them." Intrinsic value, he judges, grew at about the same rate, so the harder number offers no consolation either.

Two letters ago he warned that geometric progressions forge their own anchors. The anchor is now in plain view: the historical 22% rate, he writes, is just that, history. To earn even 15% a year over the next decade, Berkshire would need profits adding up to about $3.9 billion. That will take a few big ideas, and "small ones just won't do." Exhibit 1 shows why the ideas have to keep getting bigger.

The three holdings Buffett calls Berkshire's largest (GEICO, The Washington Post, and General Foods) have all bought back their own shares when the price sat far below the value. When an outstanding business in a comfortable financial position is that cheap, he says, nothing it can do benefits its owners as surely.

The first benefit is arithmetic, drawn in Exhibit 2. A repurchase well below per-share intrinsic value raises that value for everyone who stays: "When companies purchase their own stock, they often find it easy to get $2 of present value for $1." Acquisition programs almost never do as well and often fail to get anything close to $1 of value for each $1 spent. GEICO's retirement of over a third of its shares, the 1980 letter's example, was this arithmetic at full strength.

The second benefit is subtler and can be fully as important. A manager who repurchases when the stock is cheap demonstrates that he acts to build the owners' wealth rather than his own domain, and shareholders rationally pay more for a business in such hands. "The key word is 'demonstrated'." A manager who consistently turns his back on repurchases that are clearly in the owners' interest reveals more than he knows, however eloquently he mouths this season's favorite phrase, "maximizing shareholder wealth": "His heart is not listening to his mouth (and, after a while, neither will the market)." The endorsement stops at greenmail, "a practice we find odious and repugnant," drawn as the mugging it is in Exhibit 3.

Berkshire sat on the selling side of two such buybacks, GEICO's 1983 tender and General Foods' open-market purchases through 1984. It sold each company shares in exactly the proportion it bought from everyone else, so its stake never moved: 8.75% of General Foods before and after, with $21.8 million of cash in between. To Buffett, and to the tax code, that is a dividend. Late in 1984 the auditors' New York office decided it was a sale of stock, disagreeing with their Omaha office and Chicago reviewing partner, and rather than take a qualified opinion Berkshire adopted the New York view and restated 1983. "None of this, however, has any effect on intrinsic business value." Only the label moved, which in accounting counts as an event.

Here is the basket, bought through the insurers. It cost $585 million and was worth $1.27 billion at year-end, $2.17 for every dollar paid, with GEICO still the largest holding and, for once, standing still.

Exxon, American Broadcasting, and Northwest Industries are new to the list. Time is far larger than a year ago, and General Foods a little smaller after the shares sold back to it. Finding stocks that pass both his tests, business quality and price against value, has not been this hard in over ten years, "although we find doing nothing the most difficult task of all." A postscript added after the letter went to the typographer: Berkshire has agreed to buy three million shares of Capital Cities Communications at $172.50 each, contingent on Capital Cities completing its purchase of American Broadcasting Companies.

Last year's letter introduced Mrs. B. This year Buffett says he understated the case, and uses two numbers from a rival's own filing to show why one store in Omaha outsells every home furnishings store in America.

Sales rose $14.3 million to $115 million, all from the one store. "The business achieves this success because it deserves this success." The proof comes from Levitz, the country's largest independent specialty retailer of home furnishings, whose customers on average pay $100 for merchandise that cost it $55.60: a gross margin of 44.4%. The Mart's margin is not much more than half of that, and Exhibit 5 shows how it can afford to be: running the store takes about 16.5 cents of every sales dollar at the Mart, against 35.6 at Levitz, which Buffett calls well managed. The customers keep at least $30 million a year of the difference, and the Mart still earns excellent returns.

Asked what secrets the Blumkins bring, Buffett lists four habits he calls not very esoteric (Exhibit 6). Mrs. B's version is shorter: "sell cheap and tell the truth." The truth-telling was tested at the purchase. The Mart had never had an audit and Berkshire did not ask for one, took no inventory, verified no receivables, and checked no titles. "We gave Mrs. B a check for $55 million and she gave us her word. That made for an even exchange." She is 91, at the store seven days a week, and since May an honorary doctor of commercial science, courtesy of New York University and no prior day in a schoolroom.

Across the group, See's shows the other half of deserving: the product. It sold 24.8 million pounds, with 214 shops open at year-end, and earned $13.4 million after tax, a shade below 1983, with prices up only 1.4% and same-store pounds down 1.1%. The calendar in Exhibit 7 is the business: four weeks before Christmas carry 40% of the year's candy and about three-quarters of its profit. Most larger competitors, facing somewhat milder peaks, smooth them with preservatives or by freezing the finished product. See's refuses, "opting, in effect, for production headaches rather than product modification." Its record, he writes, reflects "an exceptional product and an exceptional manager, Chuck Huggins."

The News earned considerably more than Buffett expected in 1984. He uses the result to state an uncomfortable fact about his own industry: a dominant newspaper prospers whether it is good or bad, so its quality is a choice its management makes.

The economics of a dominant newspaper are excellent, "among the very best in the business world." Owners like to believe they earn that by turning out a wonderful product. "That comfortable theory wilts before an uncomfortable fact." First-class dominant papers make excellent profits, and third-rate dominant papers make as good or better. Quality may have been crucial to becoming dominant, and at the News he believes it was, thanks in very large part to people such as Alfred Kirchhofer, who preceded Berkshire. But once dominant, the paper itself decides how good it will be. "Good or bad, it will prosper."

Why the News's readers are worth so much is in Exhibit 8. Its circulation sits, to an unusual degree, where advertisers want it. A subscriber several hundred miles away is no prospect for the puppy in your classified ad, nor for the grocer with stores only in town, so advertisers call such copies "wasted": a paper's expenses are determined largely by gross circulation while its advertising revenue, usually 70% to 80% of the total, responds only to the useful kind. Among the hundred largest papers the News ranks first in weekday penetration and second on Sundays, up from third a year ago.

What the market will not demand, management must. "Since high standards are not imposed by the marketplace, management must impose its own." The News gave 50.9% of its space to news in 1984, against a typical 35% to 40%, held newsroom employment level while other departments cut hours by about 2.8%, and let newsroom costs rise 9.1% in a year when total costs rose 4.9% (Exhibit 9). The policy costs money, and there is room for it: news costs at comparable papers differ by perhaps three percentage points, while pre-tax margins are often ten times that.

Berkshire's own insurers paid out about $1.34 for every premium dollar in 1984, the third year running far worse than the industry. Then Buffett reveals something worse: the figures he reported for the previous three years were themselves too kind.

Phil Graham, as publisher of The Washington Post, called the daily paper "a first rough draft of history." An insurer's statements, Buffett says, are at best a first rough draft of earnings. The trouble is costs: most are claims, and many claims that belong to this year's revenue are hard to estimate. "Sometimes the extent of these losses, or even their existence, is not known for decades." Exhibit 10 draws his own example, a $10,000 estimate that became a $100,000 settlement five years later, and the correction landed on the wrong year.

Exhibit 11 shows the drafts side by side. For each year since 1981, the figure recomputed a year later was worse than the one first reported: 1983's $33.2 million loss became $51.0 million. His own verdict: "my errors in reporting to you have been substantial and recently have always presented a better underwriting picture than was truly the case." The causes: casualty and reinsurance lines, where a retiree's disease from decades-old work reveals its cost far more slowly than a burned building, juries set on making the "deep pocket" pay, and reinsurance customers whose reserving mistakes became Berkshire's. He tells of a man who asks his sister to arrange their father's funeral and send the bill, pays it, then gets a $15 bill the next month, and the next, and the next: "We buried Dad in a rented suit." Berkshire has tried to count all its rented-suit liabilities, "but our record of past error should make us humble, and you suspicious."

Not every reserving error is innocent. Some insurers that would be out of business if they counted honestly simply take an optimistic view of what they owe, and auditors cannot effectively police it. Insurance is different: "you can be broke but flush," because the cash arrives when the policy is sold and the claims are paid much later (Exhibit 12). The "walking dead" keep writing at almost any price to keep the cash coming, and Berkshire pays twice: in the price war, then through state guarantee funds when the insolvency is finally admitted.

The industry had its worst year on Best's table, paying an estimated $1.18 or so per premium dollar, because premiums since 1979 grew 30% while the old rule needed 61%. Berkshire's own ratio was "a humbling 134," though Mike Goldberg has corrected many of Buffett's mistakes, and in the second half prices in certain important lines were raised significantly with little loss of business. The day he has long predicted, when Berkshire's financial strength makes a real difference to its competitive position, "may have arrived": "The buyer of insurance receives only a promise in exchange for his cash. The value of that promise should be appraised against the possibility of adversity, not prosperity."

Between October 1983 and June 1984 Berkshire's insurers bought, at a cost of $139 million, bonds of a power authority that had just defaulted on $2.2 billion of other bonds. Buffett explains the purchase the only way he knows: as if he were buying a business.

The bonds belong to Projects 1, 2, and 3 of the Washington Public Power Supply System, not the abandoned Projects 4 and 5 whose bonds defaulted in July 1983, though the default casts a major cloud over the other three. Buffett buys stocks by the standards he would use to buy a whole business, and he extends them to these bonds. The $139 million "business" contractually earns $22.7 million a year after tax, paid in cash, which is 16.3% on unleveraged capital. A real business earning that might command $250 million to $300 million, or sometimes far more (Exhibit 13).

The minuses are real: what he views as a very slight risk that the business is worth nothing within a year or two, a risk that interest stops for a considerable period, and a hard ceiling, since the most the bonds can ever be worth is about their $205 million face value, 48% above cost. A ceiling is an important minus, though most operating businesses have one too, because most businesses cannot lift their return on equity, so their earnings grow only as new capital goes in. In 1946, twenty-year AAA tax-exempt bonds yielded slightly under 1%: seen as a business, the buyer paid 100 cents on the dollar for something earning 1% on book that could never earn a dime more, when businesses with excellent prospects sold at or near book while earning 10%, 12%, or 15% after tax. Exhibit 14 draws the bargain the bond buyers walked past.

He is candid about the risk of looking foolish. Over fifty such judgments he and Charlie Munger would expect their record to prove reasonably satisfactory, but they get nowhere near fifty a year, so "in any given year we run a risk that we will look extraordinarily foolish." Most managers avoid the intelligent decision that might look idiotic, since a conventional failure brings no pink slip. Owning 47% of Berkshire between them, the two behave with its money as with their own: they concentrate the insurers' investments, which only exceptional financial strength makes prudent, and they say nothing about what they are buying, since a couple of coat-tail riders could easily have cost $5 million. Long bonds in general must still clear a special hurdle, because he expects substantial inflation ahead, with a small but not insignificant chance of a runaway.

Companies announce dividend policies and seldom explain them. Buffett sets out the one test that should decide whether a company keeps your share of the profit or mails it to you, then applies it to Berkshire.

Not all earnings are created equal. In businesses that need a lot of assets per dollar of profit, inflation makes part of the reported profit "ersatz": these "restricted" earnings must be plowed back just to hold unit volume, competitive position, and financial strength. They are "conscripted by the business, no matter how poor its economic potential." Consolidated Edison showed the extreme a decade ago: with its stock as low as a quarter of book value, every retained dollar became 25 cents of market value, and the company reinvested most of its earnings anyway, under the slogan "Dig We Must."

Unrestricted earnings should be kept only when there is a reasonable prospect that each dollar retained will create at least a dollar of market value, which happens only when the retained capital earns at least what investors could get elsewhere. His illustration is a risk-free perpetual 10% bond whose owner may take each coupon in cash or in more 10% bonds (Exhibit 15). When market rates are 5%, take the bonds. When rates are 15%, take the cash. A company's reinvestment rate is a guess rather than a contract, but once the guess is made, the rest is simple.

Many managers reason this way about their subsidiaries and forget it at the top (Exhibit 16). A chief executive orders a division earning 5% on new money to send its cash to one earning 15%, since "The CEO's business school oath will allow no lesser behavior." Then, with his own record at 5% and market rates at 10%, he is likely to set the parent's dividend by habit or industry pattern and keep the rest. A marvelous core business, which needs little new capital, can hide disastrous acquisitions elsewhere the way a professional's best-ball score carries a team of duffers at a Pro-Am. His verdict: "If earnings have been unwisely retained, it is likely that managers, too, have been unwisely retained."

Berkshire passes its own test, so far: each kept dollar has become more than a dollar of market value. Significant payouts in the early years might have been disastrous. The three companies he and Munger started with lived on textiles, department stores, and trading stamps, which have respectively survived but earned almost nothing, shriveled with large losses, and shrunk to about 5% of their size (Exhibit 17). Only moving the money into much better businesses overcame those origins, "like overcoming a misspent youth." The policy from here: keep every dollar while the prospective return clears the dollar-for-dollar bar, judging the record on a five-year rolling average and the prospects alongside it, and pay out the unrestricted earnings if it ever does not. For now the dollars go into the insurers' capital, because most competitors are in weakened financial condition while industry premium gains of well over $15 billion look imminent in 1985.