返回列表
💬 讨论题 · 🧠 阿头学

赖希税率帖不是分析，而是口径混搭的情绪动员

这篇文章最有力的判断是：赖希把三个不可比的税率数字并排展示，构造了愤怒感，但作者自己也明显带着政治站位，纠错有价值，定性成“骗子”则证据过猛。
打开原文 ↗

2026-05-22 原文链接 ↗
阅读简报
双语对照
完整翻译
原文
讨论归档

核心观点

  • 三组数字确实不能直接比较 Bezos 的 0.98% 来自“已缴税款/财富增长”，Amazon 的 1.4% 偏向“单年现金缴税切片”，普通美国人的 14.5% 则是“全体报税人的平均联邦所得税率”，这三者分母、时间窗、统计对象都不同，硬并排本身就是误导。
  • 对 14.5% 的反驳最站得住脚 如果原帖把 14.5% 说成“普通美国人的典型税率”，那就是统计口径错误，因为平均值不等于典型值，更不等于中位数纳税人；这一点属于硬伤，不是立场分歧。
  • 对 Bezos 0.98% 的批判有力，但作者把争议说死了 说它不是标准“有效所得税率”是对的，但直接说“从定义上就是假的”并不稳，因为 ProPublica 讨论的是“财富暴涨却可长期低实现收入”的制度现象，不是会计教科书定义；作者把“非标准口径”上升为“造假”，判断过头。
  • 对 Amazon 的辩护有事实基础，但混入了转移战场 加速折旧导致当年现金税负偏低、并不等于永远免税，这点成立；但作者随后搬出就业、投资、小企业拉动效应来为税负设计辩护，这并不能直接回答“当期联邦税负是否偏低”，属于把税务问题扩展成产业政策宣传。
  • 本文本身也是强立场传播，不是中立审计 作者反复使用“骗子”“煽动术”等高强度词汇，且文末直接导流订阅，再叠加其智库背景，说明这不是纯事实核查，而是一次针对赖希叙事的政治反击；可信之处在口径拆解，不可信之处在动机判决。

跟我们的关联

  • 对 ATou 意味着什么、下一步怎么用 这篇文章提醒 ATou，看到爆款观点时不能先争立场，必须先拆分母、时间窗、样本定义；下一步可以把“定义—口径—基准”做成固定阅读模板，先验真再表态。
  • 对 Neta 意味着什么、下一步怎么用 对 Neta 来说，这是一堂典型的“高传播内容如何靠不可比数字制造判断”的课；下一步适合把它沉淀成信息卫生规则，用于筛选媒体、研究报告和政策讨论里的口径污染。
  • 对 Uota 意味着什么、下一步怎么用 对 Uota 来说，关键不是税务知识，而是识别“同源省略”——同一来源里是否藏着更不利于作者结论的数据；下一步可以训练自己每次都追到原始材料，而不是停留在截图和二手转述。
  • 对投资判断意味着什么、下一步怎么用 对投资者来说，这篇材料说明企业税率争议常被政治化，单看某年现金税负很容易误判公司质量；下一步应把递延税、资本开支激励、长期现金流与会计税率放在一起看，避免被单点数字带节奏。

讨论引子

1. 如果一个指标不是税法上的“标准有效税率”，但能揭示超富人通过未实现增值延后纳税的现实，它到底算洞察还是操纵？ 2. Amazon 当期低现金税负若主要来自政策性递延，这应该被视为合理激励，还是应被继续质疑其分配后果？ 3. 在公共讨论里，故意省略同一来源中不利于自己结论的数据，应该被定性为“误导”还是“欺骗”？

有一种特别的虚伪，只有有头衔的人才使得出来。普通骗子只能指望听众不去核查。有头衔的骗子则相信听众不会去核查，毕竟，一个内阁部长、一个伯克利教授、一部关于不平等的 Netflix 纪录片主角，总不至于说出明显错误的话。罗伯特·赖希懂这一套。他的职业生涯，就是靠这个搭起来的。

看看他在 2026 年 5 月发在 𝕏 上的帖子。这类东西总是会像往常一样传开，最后被几百万读者看到。原文是这样写的。“Jeff Bezos 在 2014 年到 2018 年支付的有效税率：0.98%。Amazon 在 2025 年支付的联邦有效税率：1.4%。普通美国人支付的典型税率：14.5%。只是觉得我应该把这件事指出来。”

三个数字，并排摆出，被呈现得仿佛可以直接比较，仿佛足以定罪。可这三个数字，每一个都以不同的方式错了。0.98% 从定义上就是假的。1.4% 在结构上具有误导性。14.5% 在统计上被夸大了。把它们并置在一起，不是分析。这在修辞上更像魔术师的障眼法，目的就是让观众盯着一只手，好让另一只手完成真正的动作。

先说 Bezos。0.98% 这个数字出自 ProPublica 在 2021 年的一篇报道，依据的是非法泄露的 IRS 数据。这个数字无论按哪种意义，都不是会计师、税务律师或联邦法官会认可的有效税率。ProPublica 发明了一个新指标，并把它命名为 “true tax rate”，它的算法是，用已缴联邦所得税除以《福布斯》估算的某位亿万富翁同期净资产增长额。按照 ProPublica 自己的数据，2014 年到 2018 年，Bezos 申报了 42.2 亿美元收入，缴纳了 9.73 亿美元联邦税。用 9.73 亿美元除以 42.2 亿美元，得到的是 23%。这才是 Bezos 在那些年里，针对其实际申报收入所支付的真实联邦所得税有效税率。对高收入者来说，这是一个完全正常的税率。大致相当于曼哈顿一家律师事务所高级合伙人的税率。它就埋在 ProPublica 那篇文章里，只要有人愿意读过标题往下看，就能看到。

要得出 0.98%，ProPublica 必须换掉分母。它不再用已缴税款除以已获得收入，而是改用已缴税款除以 Bezos 所持 Amazon 股票的升值额。问题在于，这部分升值他并没有卖出，没有收到手里，没有落袋，也根本不能花。他的股票只是在账面上上涨了。ProPublica 却决定，为了配合它的算法，账面升值也应被算作收入，尽管地球上没有任何一个司法辖区、没有《美国法典》中的任何一条法规、也没有《美国宪法第十六修正案》中的任何一项条文，把这类东西视为收入。National Taxpayers Union Foundation 把这种构造称为“往轻了说是政策分析上的失职，往重了说就是故意不诚实”。American Enterprise Institute 也指出了显而易见的问题：“但这根本不是什么真实税率。首先，这暗示他们缴纳的唯一税种是所得税，可实际上他们还缴纳了一大堆别的税：资本利得税、财产税、销售税等等。其次，财富不是收入。”

赖希知道这一点。他不是外行。他是受过耶鲁训练的律师，曾在三届政府中任职，如今还是伯克利 Goldman School of Public Policy 的院长。他知道什么是有效税率。他也知道 ProPublica 为了造出这个数字做了什么。他还是选择了 0.98% 这个数字。而且他在转述时去掉了 “true tax rate” 这个限定语，去掉了算法说明，也去掉了原始来源里只隔着一段文字的那个真实 23% 税率。

再看看 ProPublica 所选时间窗口结束之后发生了什么。2020 年到 2021 年之间，Bezos 卖出了大约 157 亿美元的 Amazon 股票，并按长期联邦资本利得税 20% 再加上 3.8% 的净投资收益税缴税。2024 年，他又卖出了额外 136 亿美元的股票。按 Forbes 的分析，仅这一笔交易，他估计就缴纳了 27 亿美元联邦税。从 2017 年《减税与就业法案》通过之后到 2025 年，他累计出售的 Amazon 股票总额大约达到 367 亿美元。即便按照偏左的 Institute on Taxation and Economic Policy 的口径来算，这也带来了数十亿美元的联邦税款，而这个机构显然没有任何动机去替他粉饰。被赖希描述成只缴 0.98% 税率的这个人，在赖希刻意截取的时间窗口之后，恰恰是美国纳税额最高的个人之一。推文里对这些只字未提。

再看 Amazon。公司在 2025 年的 1.4% 联邦税率，从技术上说，确实可以被算出来，但前提是你得特意挑选某个分子和某个分母，而且这种挑法没有哪个严肃的企业税务分析师会替它辩护。Amazon 在 2025 财年的联邦所得税支出，按公司披露，是在 973 亿美元税前收入基础上计提了 191 亿美元，按会计口径算出来的有效税率大约是 19.6%。这个数字就写在公司的 10-K 年报里，这份年报已向 Securities and Exchange Commission 提交，经过独立会计师事务所审计，并由首席财务官在承担刑事责任的前提下签字确认。

那赖希是怎么得出 1.4% 的呢。方法是把单一年度里当年实际支付的联邦所得税现金额单独拎出来，忽略公司账上已经确认的递延所得税负债，然后再拿这一个切片去除以总营收或美国税前利润，好让这个比率看起来足够小。2025 年出现低现金缴税额的真正原因，是 One Big Beautiful Bill Act 中关于加速折旧的规定。这个规定允许企业把资本投资在发生当年一次性抵扣，而不是分摊到未来数十年。税并没有被免掉。它只是被递延了。等折旧红利用完，未来年份还是要缴。

这不是什么漏洞。这是税法中有意设计出来的一项功能，而且是两党共同推动通过的，目的就是鼓励 Amazon 实际去做的那种投资。根据 Progressive Policy Institute 的数据，2025 年这家公司在美国投资了 3400 亿美元，连续第七年超过所有其他美国企业。这些钱投向了数据中心、履约中心、AI 基础设施和工资。Amazon 目前支撑着大约 200 万个美国就业岗位，其中 100 万是直接岗位，100 万是间接岗位。Oxford Economics 的研究发现，一座 Amazon 设施在某个县开设后的五年内，平均会多出 6000 家小企业，失业人口减少 12000 人，劳动力人口增加 10000 人。赖希的推文把促成这些投资的税收安排说成是企业犯罪的证据。他是在攻击一家公司使用了国会专门设计出来的激励机制。

再看第三个数字。它最少受到审视，因为它听上去最温和。所谓普通美国人的 14.5%，出自 Tax Foundation 对 IRS Statistics of Income 2022 税年数据的分析。它代表的是所有报税人的联邦个人所得税平均税率。它并不是任何某个具体人的税率，更不可能是税负中位数纳税人的税率。按照同一组 IRS 数据，纳税人中处于底部 50% 的人，也就是 2022 年收入低于 50,339 美元的那一半人，平均有效联邦所得税税率是 3.7%。再往上的那个十分位区间，税率大约在 7% 到 10% 之间。14.5% 这个平均值之所以高，是因为赖希声称几乎不缴税的那些超级富裕家庭把平均数抬上去了。收入最高的 1% 平均有效税率是 26.1%。收入最高的 10% 是 21.5%。换句话说，赖希实际上是在利用富人缴纳的高税率，去构造那个所谓普通人的基准线，然后再拿这个基准线去指控富人缴得太少。

美国的税制，是发达国家里累进性最强的税制。Tax Foundation 对 2022 年 IRS 数据的分析显示，收入最高的 1% 拿走了全部收入的 22.4%，却缴纳了全部联邦所得税的 40.4%；最高的 10% 缴纳了 72%；最高的 50% 缴纳了 97.1%；而底部 50% 只缴纳了 2.9%。2022 年，收入最高的 1% 缴纳的联邦所得税，比底部 95% 合起来还多。1980 年，最高边际税率是 70%，当时最高 1% 缴纳了 19% 的所得税。到了 2022 年，最高边际税率降到 37%，最高 1% 却缴纳了 40.4%。赖希曾把《减税与就业法案》说成是送给富人的大礼包，可这项法案却把最高 1% 承担的税收份额推高到 2021 年 46% 的峰值。Heritage Foundation 已经记录了这半个世纪的变化轨迹。赖希在这同样半个世纪里，一直坚持宣称事实恰好相反。

赖希在这条 𝕏 帖子里干的事，有一个准确的说法。它不是分析，也不是倡议，因为这两者都可以是诚实的。它是一种被包装过的煽动术，具体做法是把数字从原本的定义中剥离出来，按情绪效果并置，再借一位很清楚自己删改了什么的学院派权威之口发布出去。他所依赖的来源里，23% 这个税率就摆在那里。他选择不提。Amazon 的 10-K 里，191 亿美元这个数字就摆在那里。他选择不提。Tax Foundation 那份给出 14.5% 平均值的同一份报告里，底部一半人 3.7% 的税率也写得明明白白。他还是选择不提。每一项单独看，也许都像是一时疏漏。合在一起看，这就是一种方法。

判断公共论辩里一个人是否 intellectually honest，最简单的标准，就是看他在引用数据时，是否呈现了那些数据中最有力、最完整的版本。赖希没有这样做。他摆出了三个数字。把它们并排放在一起，看上去像一桩丑闻。他赌的是，读者不会顺着线头往下扯。多数人确实没有。现在，有些人开始扯了。

如果你喜欢我的作品，请订阅 https://x.com/amuse/creator-subscriptions/subscribe

https://x.com/amuse/creator-subscriptions/subscribe

Alexander Muse 是 John Milton Freedom Foundation 的研究员，并在 amuseonx.com 每日发布政治分析。本文引用的主要来源均已在正文中以内联方式链接；竞选财务数据取自 FEC 申报文件，民调数据取自公开发布的交叉表，法律主张取自已提交的诉状。更正会发布在原始 URL 页面上，并附有带日期的变更记录。欢迎发现错误的读者直接联系作者。

There is a particular sort of dishonesty available only to a man with credentials. The ordinary liar must hope his audience does not check. The credentialed liar trusts that his audience will not check, because surely a Cabinet secretary, a Berkeley professor, the subject of a Netflix documentary on inequality, would not say something obviously false. Robert Reich understands this. He has built a career on it.

有一种特别的虚伪，只有有头衔的人才使得出来。普通骗子只能指望听众不去核查。有头衔的骗子则相信听众不会去核查，毕竟，一个内阁部长、一个伯克利教授、一部关于不平等的 Netflix 纪录片主角，总不至于说出明显错误的话。罗伯特·赖希懂这一套。他的职业生涯，就是靠这个搭起来的。

Consider the post he published on 𝕏 in May 2026, which travelled, as such things do, to several million readers. "Effective tax rate paid by Jeff Bezos from 2014 to 2018: 0.98%. Effective 2025 federal tax rate paid by Amazon: 1.4%. Typical tax rate paid by the average American: 14.5%. Just thought I should point that out."

看看他在 2026 年 5 月发在 𝕏 上的帖子。这类东西总是会像往常一样传开，最后被几百万读者看到。原文是这样写的。“Jeff Bezos 在 2014 年到 2018 年支付的有效税率：0.98%。Amazon 在 2025 年支付的联邦有效税率：1.4%。普通美国人支付的典型税率：14.5%。只是觉得我应该把这件事指出来。”

Three numbers, presented in parallel, presented as comparable, presented as damning. And every one of them is wrong in a different way. The 0.98% is definitionally false. The 1.4% is structurally misleading. The 14.5% is statistically inflated. The juxtaposition is not analysis. It is the rhetorical equivalent of a magician's patter, designed to keep the audience watching one hand while the other does the work.

三个数字，并排摆出，被呈现得仿佛可以直接比较，仿佛足以定罪。可这三个数字，每一个都以不同的方式错了。0.98% 从定义上就是假的。1.4% 在结构上具有误导性。14.5% 在统计上被夸大了。把它们并置在一起，不是分析。这在修辞上更像魔术师的障眼法，目的就是让观众盯着一只手，好让另一只手完成真正的动作。

Begin with Bezos. The 0.98% number comes from a 2021 ProPublica report based on illegally leaked IRS data. The figure is not an effective tax rate in any sense that an accountant, a tax attorney, or a federal judge would recognize. ProPublica invented a new metric, which it labelled the "true tax rate," and which it constructed by dividing federal income taxes paid by the Forbes estimate of how much a billionaire's net worth grew during the same period. From 2014 to 2018, ProPublica's own figures show that Bezos reported $4.22 billion in income and paid $973 million in federal taxes. Divide $973 million by $4.22 billion and you get 23%. That is the actual effective federal income tax rate Bezos paid on his actual reported income during those years. It is a perfectly normal rate for a high earner. It is roughly what a senior partner at a Manhattan law firm pays. It is buried in the ProPublica article, available to anyone willing to read past the headline.

先说 Bezos。0.98% 这个数字出自 ProPublica 在 2021 年的一篇报道，依据的是非法泄露的 IRS 数据。这个数字无论按哪种意义，都不是会计师、税务律师或联邦法官会认可的有效税率。ProPublica 发明了一个新指标，并把它命名为 “true tax rate”，它的算法是，用已缴联邦所得税除以《福布斯》估算的某位亿万富翁同期净资产增长额。按照 ProPublica 自己的数据，2014 年到 2018 年，Bezos 申报了 42.2 亿美元收入，缴纳了 9.73 亿美元联邦税。用 9.73 亿美元除以 42.2 亿美元，得到的是 23%。这才是 Bezos 在那些年里，针对其实际申报收入所支付的真实联邦所得税有效税率。对高收入者来说，这是一个完全正常的税率。大致相当于曼哈顿一家律师事务所高级合伙人的税率。它就埋在 ProPublica 那篇文章里，只要有人愿意读过标题往下看，就能看到。

To arrive at 0.98%, ProPublica had to substitute a different denominator. Instead of dividing taxes paid by income earned, it divided taxes paid by the appreciation in Bezos's Amazon shares, an appreciation he had not sold, had not received, had not pocketed, and could not spend. His shares went up on paper. ProPublica decided that paper appreciation should count as income for the purpose of the numerator, even though no jurisdiction on earth, no statute in the United States Code, and no provision of the Sixteenth Amendment treats it as income. The National Taxpayers Union Foundation called this construction "at best policy analysis malpractice, and at worst deliberately dishonest." The American Enterprise Institute pointed out the obvious: "Except it's not a true tax rate. First, this suggests that the only taxes they've paid are income taxes, when in reality they've paid a slew of other taxes: capital gains, property, sales, etc. Second, wealth is not income."

要得出 0.98%，ProPublica 必须换掉分母。它不再用已缴税款除以已获得收入，而是改用已缴税款除以 Bezos 所持 Amazon 股票的升值额。问题在于，这部分升值他并没有卖出，没有收到手里，没有落袋，也根本不能花。他的股票只是在账面上上涨了。ProPublica 却决定，为了配合它的算法，账面升值也应被算作收入，尽管地球上没有任何一个司法辖区、没有《美国法典》中的任何一条法规、也没有《美国宪法第十六修正案》中的任何一项条文，把这类东西视为收入。National Taxpayers Union Foundation 把这种构造称为“往轻了说是政策分析上的失职，往重了说就是故意不诚实”。American Enterprise Institute 也指出了显而易见的问题：“但这根本不是什么真实税率。首先，这暗示他们缴纳的唯一税种是所得税，可实际上他们还缴纳了一大堆别的税：资本利得税、财产税、销售税等等。其次，财富不是收入。”

Reich knows this. He is not a layman. He is a Yale-trained lawyer who served in three administrations and chairs the Goldman School of Public Policy at Berkeley. He understands what an effective tax rate is. He understands what ProPublica did to construct its number. He selected the 0.98% figure anyway. And he presented it without the qualifier "true tax rate," without the methodology, and without the actual 23% rate sitting one paragraph away in the original source.

赖希知道这一点。他不是外行。他是受过耶鲁训练的律师，曾在三届政府中任职，如今还是伯克利 Goldman School of Public Policy 的院长。他知道什么是有效税率。他也知道 ProPublica 为了造出这个数字做了什么。他还是选择了 0.98% 这个数字。而且他在转述时去掉了 “true tax rate” 这个限定语，去掉了算法说明，也去掉了原始来源里只隔着一段文字的那个真实 23% 税率。

Consider also what has happened since the ProPublica window closed. Between 2020 and 2021, Bezos sold approximately $15.7 billion in Amazon stock, paying federal long-term capital gains tax of 20% plus the 3.8% net investment income tax on the gain. In 2024 he sold an additional $13.6 billion, paying an estimated $2.7 billion in federal taxes on the transaction according to a Forbes analysis. Between the 2017 Tax Cuts and Jobs Act and 2025, his cumulative Amazon stock sales totalled roughly $36.7 billion, producing federal tax payments in the billions even by the accounting of the Institute on Taxation and Economic Policy, a left-leaning shop that has no incentive to flatter him. The man Reich presents as paying 0.98% is, in the years following Reich's chosen window, one of the largest individual taxpayers in the United States. None of this appears in the tweet.

再看看 ProPublica 所选时间窗口结束之后发生了什么。2020 年到 2021 年之间，Bezos 卖出了大约 157 亿美元的 Amazon 股票，并按长期联邦资本利得税 20% 再加上 3.8% 的净投资收益税缴税。2024 年，他又卖出了额外 136 亿美元的股票。按 Forbes 的分析，仅这一笔交易，他估计就缴纳了 27 亿美元联邦税。从 2017 年《减税与就业法案》通过之后到 2025 年，他累计出售的 Amazon 股票总额大约达到 367 亿美元。即便按照偏左的 Institute on Taxation and Economic Policy 的口径来算，这也带来了数十亿美元的联邦税款，而这个机构显然没有任何动机去替他粉饰。被赖希描述成只缴 0.98% 税率的这个人，在赖希刻意截取的时间窗口之后，恰恰是美国纳税额最高的个人之一。推文里对这些只字未提。

Turn to Amazon. The 1.4% figure for the company's 2025 federal tax rate is technically a thing one can calculate, but it requires choosing a particular numerator and a particular denominator in a way that no serious corporate tax analyst would defend. Amazon's actual reported federal income tax provision for fiscal 2025 was $19.1 billion on $97.3 billion in pre-tax income, an accounting effective tax rate of roughly 19.6%. That figure appears in the company's 10-K, filed with the Securities and Exchange Commission, audited by an independent accounting firm, and signed under criminal penalty by the chief financial officer.

再看 Amazon。公司在 2025 年的 1.4% 联邦税率，从技术上说，确实可以被算出来，但前提是你得特意挑选某个分子和某个分母，而且这种挑法没有哪个严肃的企业税务分析师会替它辩护。Amazon 在 2025 财年的联邦所得税支出，按公司披露，是在 973 亿美元税前收入基础上计提了 191 亿美元，按会计口径算出来的有效税率大约是 19.6%。这个数字就写在公司的 10-K 年报里，这份年报已向 Securities and Exchange Commission 提交，经过独立会计师事务所审计，并由首席财务官在承担刑事责任的前提下签字确认。

How does Reich get to 1.4%? By isolating current-year cash federal income taxes paid in a single year, ignoring the deferred tax liability the company has booked, and then dividing that one slice by total revenue or pre-tax US income in a way that makes the ratio look small. The actual mechanism producing the low cash payment in 2025 was the One Big Beautiful Bill Act's accelerated depreciation provisions, which allow companies to write off capital investment in the year it is made rather than over multiple decades. The tax is not avoided. It is deferred. It will be paid in future years when the depreciation runs out.

那赖希是怎么得出 1.4% 的呢。方法是把单一年度里当年实际支付的联邦所得税现金额单独拎出来，忽略公司账上已经确认的递延所得税负债，然后再拿这一个切片去除以总营收或美国税前利润，好让这个比率看起来足够小。2025 年出现低现金缴税额的真正原因，是 One Big Beautiful Bill Act 中关于加速折旧的规定。这个规定允许企业把资本投资在发生当年一次性抵扣，而不是分摊到未来数十年。税并没有被免掉。它只是被递延了。等折旧红利用完，未来年份还是要缴。

This is not a loophole. It is a deliberate feature of the tax code, enacted on a bipartisan basis, designed to encourage exactly the activity Amazon engaged in. In 2025 the company invested $340 billion in the United States, more than any other American firm for the seventh consecutive year according to the Progressive Policy Institute. That money went into data centers, fulfillment centers, AI infrastructure, and wages. Amazon currently supports approximately 2 million US jobs, 1 million direct and 1 million indirect. Oxford Economics found that within five years of an Amazon facility opening in a county, on average 6,000 more small businesses emerge, 12,000 fewer people are unemployed, and 10,000 more workers join the labor force. Reich's tweet treats the tax provision that enabled this investment as evidence of corporate criminality. He attacks the company for using the incentive Congress designed.

这不是什么漏洞。这是税法中有意设计出来的一项功能，而且是两党共同推动通过的，目的就是鼓励 Amazon 实际去做的那种投资。根据 Progressive Policy Institute 的数据，2025 年这家公司在美国投资了 3400 亿美元，连续第七年超过所有其他美国企业。这些钱投向了数据中心、履约中心、AI 基础设施和工资。Amazon 目前支撑着大约 200 万个美国就业岗位，其中 100 万是直接岗位，100 万是间接岗位。Oxford Economics 的研究发现，一座 Amazon 设施在某个县开设后的五年内，平均会多出 6000 家小企业，失业人口减少 12000 人，劳动力人口增加 10000 人。赖希的推文把促成这些投资的税收安排说成是企业犯罪的证据。他是在攻击一家公司使用了国会专门设计出来的激励机制。

Now turn to the third number, the one that goes least examined because it sounds the most modest. The 14.5% figure for the "typical American" comes from Tax Foundation analysis of IRS Statistics of Income data for tax year 2022. It is the average federal individual income tax rate across all filers. It is not the rate paid by anyone in particular, and it is certainly not the rate paid by the median taxpayer. According to the same IRS data, the bottom 50% of taxpayers, those earning under $50,339 in 2022, paid an average effective federal income tax rate of 3.7%. The next decile paid roughly 7% to 10%. The 14.5% average is inflated by the very wealthy households Reich claims pay nothing. The top 1% paid an average effective rate of 26.1%. The top 10% paid 21.5%. Reich is, in effect, using the high tax rates paid by the rich to construct the "typical" baseline that he then uses to indict the rich for paying too little.

再看第三个数字。它最少受到审视，因为它听上去最温和。所谓普通美国人的 14.5%，出自 Tax Foundation 对 IRS Statistics of Income 2022 税年数据的分析。它代表的是所有报税人的联邦个人所得税平均税率。它并不是任何某个具体人的税率，更不可能是税负中位数纳税人的税率。按照同一组 IRS 数据，纳税人中处于底部 50% 的人，也就是 2022 年收入低于 50,339 美元的那一半人，平均有效联邦所得税税率是 3.7%。再往上的那个十分位区间，税率大约在 7% 到 10% 之间。14.5% 这个平均值之所以高，是因为赖希声称几乎不缴税的那些超级富裕家庭把平均数抬上去了。收入最高的 1% 平均有效税率是 26.1%。收入最高的 10% 是 21.5%。换句话说，赖希实际上是在利用富人缴纳的高税率，去构造那个所谓普通人的基准线，然后再拿这个基准线去指控富人缴得太少。

The American tax system is the most progressive in the developed world. The Tax Foundation's analysis of IRS data for 2022 shows the top 1% paid 40.4% of all federal income taxes on 22.4% of all income, the top 10% paid 72%, the top 50% paid 97.1%, and the bottom 50% paid 2.9%. The top 1% paid more federal income tax in 2022 than the bottom 95% combined. In 1980, when the top marginal rate was 70%, the top 1% paid 19% of income taxes. In 2022, with a top marginal rate of 37%, the top 1% paid 40.4%. The Tax Cuts and Jobs Act, which Reich called a giveaway to the rich, increased the share of taxes paid by the top 1% to a peak of 46% in 2021. The Heritage Foundation has documented this trajectory across half a century. Reich has spent that same half century insisting that the opposite is true.

美国的税制，是发达国家里累进性最强的税制。Tax Foundation 对 2022 年 IRS 数据的分析显示，收入最高的 1% 拿走了全部收入的 22.4%，却缴纳了全部联邦所得税的 40.4%；最高的 10% 缴纳了 72%；最高的 50% 缴纳了 97.1%；而底部 50% 只缴纳了 2.9%。2022 年，收入最高的 1% 缴纳的联邦所得税，比底部 95% 合起来还多。1980 年，最高边际税率是 70%，当时最高 1% 缴纳了 19% 的所得税。到了 2022 年，最高边际税率降到 37%，最高 1% 却缴纳了 40.4%。赖希曾把《减税与就业法案》说成是送给富人的大礼包，可这项法案却把最高 1% 承担的税收份额推高到 2021 年 46% 的峰值。Heritage Foundation 已经记录了这半个世纪的变化轨迹。赖希在这同样半个世纪里，一直坚持宣称事实恰好相反。

There is a phrase for what Reich does in the 𝕏 post. It is not analysis, and it is not advocacy, both of which can be honest. It is a particular kind of laundered demagoguery in which numbers are stripped of their definitions, juxtaposed for emotional effect, and presented under the authority of a credentialed academic who knows precisely what corners he is cutting. The 23% rate is in the source he relies on. He chose not to mention it. The $19.1 billion is in Amazon's 10-K. He chose not to mention it. The 3.7% bottom-half rate is in the same Tax Foundation report that produced his 14.5% average. He chose not to mention it. Each omission, considered individually, might be a slip. Considered together, they constitute a method.

赖希在这条 𝕏 帖子里干的事，有一个准确的说法。它不是分析，也不是倡议，因为这两者都可以是诚实的。它是一种被包装过的煽动术，具体做法是把数字从原本的定义中剥离出来，按情绪效果并置，再借一位很清楚自己删改了什么的学院派权威之口发布出去。他所依赖的来源里，23% 这个税率就摆在那里。他选择不提。Amazon 的 10-K 里，191 亿美元这个数字就摆在那里。他选择不提。Tax Foundation 那份给出 14.5% 平均值的同一份报告里，底部一半人 3.7% 的税率也写得明明白白。他还是选择不提。每一项单独看，也许都像是一时疏漏。合在一起看，这就是一种方法。

The simplest test of intellectual honesty in public argument is whether a writer presents the strongest version of the data he is invoking. Reich did not. He presented three numbers that, placed side by side, looked like a scandal, and he trusted that his readers would not pull the thread. Most did not. Some are pulling it now.

判断公共论辩里一个人是否 intellectually honest，最简单的标准，就是看他在引用数据时，是否呈现了那些数据中最有力、最完整的版本。赖希没有这样做。他摆出了三个数字。把它们并排放在一起，看上去像一桩丑闻。他赌的是，读者不会顺着线头往下扯。多数人确实没有。现在，有些人开始扯了。

If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe

如果你喜欢我的作品，请订阅 https://x.com/amuse/creator-subscriptions/subscribe

https://x.com/amuse/creator-subscriptions/subscribe

https://x.com/amuse/creator-subscriptions/subscribe

Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly.

Alexander Muse 是 John Milton Freedom Foundation 的研究员，并在 amuseonx.com 每日发布政治分析。本文引用的主要来源均已在正文中以内联方式链接；竞选财务数据取自 FEC 申报文件，民调数据取自公开发布的交叉表，法律主张取自已提交的诉状。更正会发布在原始 URL 页面上，并附有带日期的变更记录。欢迎发现错误的读者直接联系作者。

There is a particular sort of dishonesty available only to a man with credentials. The ordinary liar must hope his audience does not check. The credentialed liar trusts that his audience will not check, because surely a Cabinet secretary, a Berkeley professor, the subject of a Netflix documentary on inequality, would not say something obviously false. Robert Reich understands this. He has built a career on it.

Consider the post he published on 𝕏 in May 2026, which travelled, as such things do, to several million readers. "Effective tax rate paid by Jeff Bezos from 2014 to 2018: 0.98%. Effective 2025 federal tax rate paid by Amazon: 1.4%. Typical tax rate paid by the average American: 14.5%. Just thought I should point that out."

Three numbers, presented in parallel, presented as comparable, presented as damning. And every one of them is wrong in a different way. The 0.98% is definitionally false. The 1.4% is structurally misleading. The 14.5% is statistically inflated. The juxtaposition is not analysis. It is the rhetorical equivalent of a magician's patter, designed to keep the audience watching one hand while the other does the work.

Begin with Bezos. The 0.98% number comes from a 2021 ProPublica report based on illegally leaked IRS data. The figure is not an effective tax rate in any sense that an accountant, a tax attorney, or a federal judge would recognize. ProPublica invented a new metric, which it labelled the "true tax rate," and which it constructed by dividing federal income taxes paid by the Forbes estimate of how much a billionaire's net worth grew during the same period. From 2014 to 2018, ProPublica's own figures show that Bezos reported $4.22 billion in income and paid $973 million in federal taxes. Divide $973 million by $4.22 billion and you get 23%. That is the actual effective federal income tax rate Bezos paid on his actual reported income during those years. It is a perfectly normal rate for a high earner. It is roughly what a senior partner at a Manhattan law firm pays. It is buried in the ProPublica article, available to anyone willing to read past the headline.

To arrive at 0.98%, ProPublica had to substitute a different denominator. Instead of dividing taxes paid by income earned, it divided taxes paid by the appreciation in Bezos's Amazon shares, an appreciation he had not sold, had not received, had not pocketed, and could not spend. His shares went up on paper. ProPublica decided that paper appreciation should count as income for the purpose of the numerator, even though no jurisdiction on earth, no statute in the United States Code, and no provision of the Sixteenth Amendment treats it as income. The National Taxpayers Union Foundation called this construction "at best policy analysis malpractice, and at worst deliberately dishonest." The American Enterprise Institute pointed out the obvious: "Except it's not a true tax rate. First, this suggests that the only taxes they've paid are income taxes, when in reality they've paid a slew of other taxes: capital gains, property, sales, etc. Second, wealth is not income."

Reich knows this. He is not a layman. He is a Yale-trained lawyer who served in three administrations and chairs the Goldman School of Public Policy at Berkeley. He understands what an effective tax rate is. He understands what ProPublica did to construct its number. He selected the 0.98% figure anyway. And he presented it without the qualifier "true tax rate," without the methodology, and without the actual 23% rate sitting one paragraph away in the original source.

Consider also what has happened since the ProPublica window closed. Between 2020 and 2021, Bezos sold approximately $15.7 billion in Amazon stock, paying federal long-term capital gains tax of 20% plus the 3.8% net investment income tax on the gain. In 2024 he sold an additional $13.6 billion, paying an estimated $2.7 billion in federal taxes on the transaction according to a Forbes analysis. Between the 2017 Tax Cuts and Jobs Act and 2025, his cumulative Amazon stock sales totalled roughly $36.7 billion, producing federal tax payments in the billions even by the accounting of the Institute on Taxation and Economic Policy, a left-leaning shop that has no incentive to flatter him. The man Reich presents as paying 0.98% is, in the years following Reich's chosen window, one of the largest individual taxpayers in the United States. None of this appears in the tweet.

Turn to Amazon. The 1.4% figure for the company's 2025 federal tax rate is technically a thing one can calculate, but it requires choosing a particular numerator and a particular denominator in a way that no serious corporate tax analyst would defend. Amazon's actual reported federal income tax provision for fiscal 2025 was $19.1 billion on $97.3 billion in pre-tax income, an accounting effective tax rate of roughly 19.6%. That figure appears in the company's 10-K, filed with the Securities and Exchange Commission, audited by an independent accounting firm, and signed under criminal penalty by the chief financial officer.

How does Reich get to 1.4%? By isolating current-year cash federal income taxes paid in a single year, ignoring the deferred tax liability the company has booked, and then dividing that one slice by total revenue or pre-tax US income in a way that makes the ratio look small. The actual mechanism producing the low cash payment in 2025 was the One Big Beautiful Bill Act's accelerated depreciation provisions, which allow companies to write off capital investment in the year it is made rather than over multiple decades. The tax is not avoided. It is deferred. It will be paid in future years when the depreciation runs out.

This is not a loophole. It is a deliberate feature of the tax code, enacted on a bipartisan basis, designed to encourage exactly the activity Amazon engaged in. In 2025 the company invested $340 billion in the United States, more than any other American firm for the seventh consecutive year according to the Progressive Policy Institute. That money went into data centers, fulfillment centers, AI infrastructure, and wages. Amazon currently supports approximately 2 million US jobs, 1 million direct and 1 million indirect. Oxford Economics found that within five years of an Amazon facility opening in a county, on average 6,000 more small businesses emerge, 12,000 fewer people are unemployed, and 10,000 more workers join the labor force. Reich's tweet treats the tax provision that enabled this investment as evidence of corporate criminality. He attacks the company for using the incentive Congress designed.

Now turn to the third number, the one that goes least examined because it sounds the most modest. The 14.5% figure for the "typical American" comes from Tax Foundation analysis of IRS Statistics of Income data for tax year 2022. It is the average federal individual income tax rate across all filers. It is not the rate paid by anyone in particular, and it is certainly not the rate paid by the median taxpayer. According to the same IRS data, the bottom 50% of taxpayers, those earning under $50,339 in 2022, paid an average effective federal income tax rate of 3.7%. The next decile paid roughly 7% to 10%. The 14.5% average is inflated by the very wealthy households Reich claims pay nothing. The top 1% paid an average effective rate of 26.1%. The top 10% paid 21.5%. Reich is, in effect, using the high tax rates paid by the rich to construct the "typical" baseline that he then uses to indict the rich for paying too little.

The American tax system is the most progressive in the developed world. The Tax Foundation's analysis of IRS data for 2022 shows the top 1% paid 40.4% of all federal income taxes on 22.4% of all income, the top 10% paid 72%, the top 50% paid 97.1%, and the bottom 50% paid 2.9%. The top 1% paid more federal income tax in 2022 than the bottom 95% combined. In 1980, when the top marginal rate was 70%, the top 1% paid 19% of income taxes. In 2022, with a top marginal rate of 37%, the top 1% paid 40.4%. The Tax Cuts and Jobs Act, which Reich called a giveaway to the rich, increased the share of taxes paid by the top 1% to a peak of 46% in 2021. The Heritage Foundation has documented this trajectory across half a century. Reich has spent that same half century insisting that the opposite is true.

There is a phrase for what Reich does in the 𝕏 post. It is not analysis, and it is not advocacy, both of which can be honest. It is a particular kind of laundered demagoguery in which numbers are stripped of their definitions, juxtaposed for emotional effect, and presented under the authority of a credentialed academic who knows precisely what corners he is cutting. The 23% rate is in the source he relies on. He chose not to mention it. The $19.1 billion is in Amazon's 10-K. He chose not to mention it. The 3.7% bottom-half rate is in the same Tax Foundation report that produced his 14.5% average. He chose not to mention it. Each omission, considered individually, might be a slip. Considered together, they constitute a method.

The simplest test of intellectual honesty in public argument is whether a writer presents the strongest version of the data he is invoking. Reich did not. He presented three numbers that, placed side by side, looked like a scandal, and he trusted that his readers would not pull the thread. Most did not. Some are pulling it now.

If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe

https://x.com/amuse/creator-subscriptions/subscribe

Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly.

📋 讨论归档

讨论进行中…