随笔

巴菲特股东信读后：十点洞察（1977—2024）

2026-07-20

本文并非逐年通读巴菲特48年间致股东的全部信件，而是深入抽取了几封关键节点的原始信件——1977年首封信、1989年"错误"总结信、1996年"股东手册"、2008年金融危机信、2014年"五十周年"回顾（含巴菲特本人与芒格各自独立撰写的两篇文章）、以及2024年信——从中提炼出十条有据可查的洞察，而非泛泛而谈的语录汇编。

一、ROE（净资产收益率），而非每股收益增长，才是真正的记分牌。（1977年信）巴菲特否定了"每股收益增长"作为衡量指标的意义——一旦净资产基数在扩大，这个数字就会失真；真正诚实的衡量标准是净资产收益率。他用这把尺子去衡量伯克希尔当年19%的净资产收益率，既对照自身历史，也对照全行业水平。

二、错失的代价，远大于犯错的代价——而且两位合伙人各自独立得出了同样的结论。在2014年的信中，巴菲特与芒格分别撰写了各自的文章，且都未读过对方的初稿。芒格写道："几乎所有巨大的失误，都是没有出手买进造成的，包括没有买入沃尔玛的股票……如果当年抓住了那几次机会，伯克希尔现在的净资产至少要高出500亿美元。"巴菲特在2008年的信里表达了同样的意思："我犯了一些不作为的错误——面对新出现的事实，我本该重新审视自己的判断并及时行动，却选择了迟疑。"

三、巴菲特自己一生中最大的一笔错误交易，就是买下伯克希尔本身。（2014年信）1965年接手伯克希尔，用他自己的话说，是"一个愚蠢至极的决定"——起因不过是西伯利·斯坦顿在握手谈好的价格上硬生生扣掉八分之一点，让他一时意气用事，而不是理性分析的结果。代价更大的一笔错误，是他把国民保险公司（National Indemnity）买进了伯克希尔（当时61%由外部股东持有），而不是买进他自己完全掌控的合伙企业BPL——他估算，这一决定把大约"1000亿美元"的财富，从他的合伙人手里转移给了一群与他毫无关系的伯克希尔旧股东。"这个问题我想了48年，至今也没想出一个像样的答案。"

四、芒格的一句话，重新设计了整套打法。"忘掉你懂得的那套'用便宜的价格买一般的公司'，转而'用合理的价格买一家真正卓越的公司'。"这句话促成了1972年对喜诗糖果（See's Candy）的收购——花费2500万美元买下，此后至今已累计贡献了19亿美元的税前利润，而追加投入的资本只有4000万美元。

五、浮存金，是那台安静运转的复利引擎。保险浮存金（持有但不属于自己的资金）从1970年的3900万美元，增长到2024年的1710亿美元——这主要得益于伯克希尔多年来大部分年份都能实现承保利润，也就是说，不但免费使用这笔钱，有时甚至因为持有它而获得额外报酬。

六、看人重品行，而非学历——这一点被明确写了出来。（2024年信）"我从不看候选人毕业于哪所学校。从不！"他举了皮特·利格尔（Pete Liegl）的例子——凭一次握手就把森林河公司（Forest River）卖给了伯克希尔，自己开口定薪10万美元；还有本·罗斯纳（Ben Rosner）——"从没上过六年级以上的学"。巴菲特的结论是："商业才能很大程度上是天生的，天赋压过了后天培养。"

七、危机中的纪律，是提前准备好的，不是临场发挥的。（2008年信）"投资时，悲观是你的朋友，狂热才是敌人。"2008年恐慌期间，伯克希尔在三周内向美国企业投放了156亿美元——之所以能做到，是因为公司始终保留至少200亿美元的国库券。巴菲特把现金比作"氧气——存在时无人在意，一旦缺失便成了唯一在意的事"。

八、对"用公式取代判断"始终保持警惕。（2008年信，针对布莱克-舒尔斯期权定价公式被错误套用到百年期期权上）"提防那些手持公式的书呆子。"（2014年信）"波动性远远不等同于风险……把两者划等号的流行公式，正在把学生、投资者和CEO们引入歧途。"

九、对失败的坦诚，是可以量化的习惯。（2024年信）巴菲特提到，仅2019至2023这五年间，他在信中使用"错误"或"失误"这两个词就多达16次，而许多同等规模的大公司"从未使用过这两个词中的任何一个"。他对接班人格雷格·阿贝尔的告诫是："一旦你开始欺骗股东，用不了多久你自己也会开始相信那些谎话，最终连自己也被自己蒙骗了。"

十、企业集团的结构，本质上是一台"避税＋减少摩擦成本"的机器，而不只是简单的多元化。（2014年信）伯克希尔可以"不产生税负、也几乎不产生其他成本"地在各项业务之间调拨资金——这是任何单一投资者或单一行业公司都不具备的优势，让喜诗糖果这样成熟的现金牛业务产生的利润，可以直接拿去买新业务，而不必像独立公司那样，先把利润分给股东、被课一道税，股东再自行决定如何再投资。

参考资料：原始信件——1977年、1989年、1996年、2008年、2014年、2024年（均来自伯克希尔哈撒韦官网 berkshirehathaway.com/letters/letters.html）。

This isn't a mechanical read-through of all 48 years of Buffett's letters, but a deep sampling of several pivotal ones — the original 1977 letter, the 1989 "mistakes" letter, the 1996 "Owner's Manual," the 2008 crisis letter, the full 2014 50th-anniversary retrospective (including separate essays by both Buffett and Munger, each written without reading the other's), and the 2024 letter — distilled into ten sourced insights rather than a generic quote compilation.

ROE, not EPS growth, is the real scorecard. (1977) Buffett rejected earnings-per-share growth as a meaningful metric once the equity base is expanding — the honest measure is return on equity capital, and he judged Berkshire's 19% ROE that year against both its own history and industry aggregates.

  1. Errors of omission dwarf errors of commission — and both partners said so, independently. In the 2014 letter, Buffett and Munger each wrote their own essay without reading the other's first. Munger's line: "almost all huge errors were in not making a purchase, including not purchasing Walmart stock... Berkshire's net worth would now be at least $50 billion higher." Buffett makes the same point in 2008 about "errors of omission, sucking my thumb when new facts came in."

  1. Buffett's own worst mistake was buying Berkshire itself. (2014) Taking control in 1965 was, in his words, "a monumentally stupid decision" — born of pique at Seabury Stanton shaving an eighth of a point off a handshake price, not analysis. His even costlier mistake: buying National Indemnity inside Berkshire (61%-owned by outsiders) rather than through his own partnership BPL, which he calculates diverted "$100 billion or so" from his partners to unrelated shareholders. "I've had 48 years to think about that question, and I've yet to come up with a good answer."

  1. One sentence from Munger rewired the whole strategy. "Forget what you know about buying fair businesses at wonderful prices; instead, buy wonderful businesses at fair prices." This turned the 1972 See's Candy purchase ($25M) into the template — $1.9 billion pre-tax earned since, on just $40 million of added investment.

  1. Float is the quiet compounding engine. Insurance float (money held but not owned) grew from $39M in 1970 to $171B in 2024, largely because Berkshire runs an underwriting profit in most years — meaning it gets paid to hold and invest other people's money.

  1. Character over credentials, explicitly. (2024) "I never look at where a candidate has gone to school. Never!" — illustrated with Pete Liegl (sold Forest River on a handshake, set his own $100K salary) and Ben Rosner ("never went past 6th grade"). Buffett's conclusion: "a very large portion of business talent is innate, with nature swamping nurture."

  1. Crisis discipline is pre-loaded, not improvised. (2008) "When investing, pessimism is your friend, euphoria the enemy." Berkshire deployed $15.6 billion in three weeks during the panic precisely because it always keeps $20B+ in Treasury bills — cash described as "oxygen... never thought about when it is present, the only thing in mind when it is absent."

  1. Distrust of formulas that replace judgment. "Beware of geeks bearing formulas" (2008, on Black-Scholes misapplied to 100-year options) and "volatility is far from synonymous with risk... popular formulas that equate the two terms lead students, investors and CEOs astray" (2014).

  1. Radical, quantified honesty about failure. (2024) Buffett notes he used "mistake" or "error" 16 times in letters from 2019-23 alone, versus peer companies that "have never used either word." His warning to his successor Greg Abel: "if you start fooling your shareholders, you will soon believe your own baloney and be fooling yourself as well."

  1. The conglomerate structure is a tax-and-friction-avoidance machine, not just diversification. (2014) Berkshire can move capital between businesses "without incurring taxes or much in the way of other costs" — an advantage no individual investor or single-industry company has, letting cash from a mature cash-cow (like See's) fund new purchases without the double taxation an independent company would face distributing it.

Sources: the 1977, 1989, 1996, 2008, 2014, and 2024 letters (all from Berkshire Hathaway's own site, berkshirehathaway.com/letters/letters.html).