低估值研究：先确定分母，再谈谁更便宜
“最低估值”没有唯一排名。 PE 比较价格与利润，PB 比较价格与账面权益；剔除非经常性损益、商誉或无形资产后，回答的问题变了，排序也可能变化。下方六家公司只用于说明计算口径，不是全市场筛选，也没有检验未来收益。
证据阶段：已核对的当前快照计算，观察日为 2026 年 10 月 18 日；市值取自 10 月 16 日。财务记录实际抓取于 UTC 10 月 17 日，即北京时间 10 月 18 日，其中海康威视包含 10 月 17 日披露的数据。不能把这组输入说成 10 月 16 日收盘时已经拿到的信号。 本次没有计算历史策略收益。
口径定义表
以下 M 是以元计的全部普通股权益市值；P 是最近十二个月归母净利润；D 是同口径扣非归母净利润；E 是归属于普通股股东的权益；G、I 分别是对应权益范围的商誉和其他已入账无形资产。
| 指标 | 目标定义 | 比较前必须固定的事项 |
|---|---|---|
| PE TTM | M ÷ P | 合并报表中的归母利润，不能换成集团净利润；最近十二个月，不能直接把半年利润乘二 |
| 扣非 PE TTM | M ÷ D | 同一归母范围、同一滚动区间；扣非不等于剔除全部投资收益 |
| PB | M ÷ E | 扣除少数股东权益，适用时还须扣除优先权益；用一个资产负债表时点，不是四季加总 |
| 扣商誉 PB | M ÷（E − G） | 商誉归属范围、减值版本和权益范围必须一致 |
| 扣商誉及无形资产 PB | M ÷（E − G − I） | 明确无形资产范围、少数股东分配，以及是否另扣资本化开发支出 |
中期财报的 TTM＝本年累计值＋上年全年值−上年同期累计值；年报直接使用全年值。缺少任何一项就保留缺失。利润、权益及扣除项要使用相同合并范围，不能把母公司单体报表与合并归母数据拼在一起。
市值也必须对应同一权益。本例有意选择六家仅有 A 股的公司，没有由此推导 A/H、B 股、优先股和可转债的统一处理方法。A 股价格乘所有类别股数，不一定等于各类股份的经济价值；流通市值也不能直接除以全公司的利润或权益。
数据字段与本次计算的限制
脚本把 daily_basic.total_mv × 10,000 换算为元；以 income.n_income_attr_p 对应归母利润，以 fina_indicator.profit_dedt 作为供应商扣非利润代理，以 balancesheet.total_hldr_eqy_exc_min_int 对应归母权益。供应商对 profit_dedt 的字段名称不能独立证明每家公司的归母／普通股范围；本次尚未逐份核对原始财报及优先权益，因此表中称“扣非 PE 代理”和“归母 PB”。
goodwill、intan_assets 是合并口径资产，归母权益则不含少数股东权益。直接从归母权益全额扣除这两项，属于未分配少数股东份额的扣资产代理，不能称为已认证的归母有形净资产。没有擅自按持股比例分配。两项来源字段分别扣除；本例没有另扣资本化开发支出，没有计算相关递延税影响，也不包含表外无形资产或清算回收。账面价值不等于清算价值，无形资产较多也不意味着公司经济价值更差。
商誉或无形资产缺失时保持缺失，不能默认为零。利润或权益分母小于等于零时，不进入正倍数“便宜程度”排名；亏损、负权益需要单独研究，并不代表公司没有价值。贵州茅台的商誉字段在本快照中缺失，因此不补零，也不进入五种口径的共同排名。
六家公司计算样本
这是有意选择、覆盖几类经营模式的小样本，不代表市场分布。每家公司使用利润表、财务指标和资产负债表都有数据的最新共同报告期；不同公司报告期可能不同。只在展示时保留两位小数。窄屏可横向滑动数值表，代码和倍数保持完整。
| 代码 | 公司 | 报告期 | PE TTM | 扣非 PE 代理 | 归母 PB | 扣商誉 PB 代理 | 扣商誉及无形资产 PB 代理 |
|---|---|---|---|---|---|---|---|
| 000651.SZ | 格力电器 | 20260630 | 7.94 | 8.36 | 1.49 | 1.50 | 1.61 |
| 002027.SZ | 分众传媒 | 20260630 | 21.86 | 29.28 | 5.33 | 5.39 | 5.43 |
| 002415.SZ | 海康威视 | 20260930 | 17.34 | 17.70 | 3.66 | 3.67 | 3.75 |
| 300059.SZ | 东方财富 | 20260630 | 19.65 | 20.28 | 2.91 | 3.00 | 3.00 |
| 600519.SH | 贵州茅台 | 20260630 | 20.40 | 20.42 | 6.61 | — | — |
| 600900.SH | 长江电力 | 20260630 | 19.86 | 20.81 | 3.33 | 3.34 | 3.77 |
分众传媒的 PE 从 21.86 倍变为扣非代理口径的 29.28 倍。在五家公司共同样本内，海康威视与长江电力的 PB 相对顺序，在扣除商誉和已入账无形资产后发生对调。这些变化说明口径会影响比较，不能据此推断后续收益更高。
同一样本的排序
以下五列严格使用五种分母均为正且非缺失的同一批五家公司。1 表示本小样本内倍数最低；并列采用最小名次。第六家公司可计算的数值仍显示在上表，但不在任何共同样本列里单独排名。跨行业排名在这里只说明计算差异。
| 代码 | PE | 扣非 PE 代理 | 归母 PB | 扣商誉代理 | 扣商誉及无形资产代理 |
|---|---|---|---|---|---|
| 000651.SZ | 1 | 1 | 1 | 1 | 1 |
| 002027.SZ | 5 | 5 | 5 | 5 | 5 |
| 002415.SZ | 2 | 2 | 4 | 4 | 3 |
| 300059.SZ | 3 | 3 | 2 | 2 | 2 |
| 600519.SH | — | — | — | — | — |
| 600900.SH | 4 | 4 | 3 | 3 | 4 |
可得时间、修订与复现
读取前，脚本核对数据 owner 的归档封存记录、标准化子清单哈希及数据文件哈希。只读取 report_type=1 的累计合并报表；同一证券、报告期选择当前已披露且已实际观测的最新版本。只有日期、没有时刻的公告，必须早于当前本地自然日才纳入。无法区分的同版本重复记录会报错。来源归档沿用 owner 配置的 TuShare 链路，本次无需新请求供应商，也没有切换生产数据别名。
完整性校验证明留存文件与哈希一致，不能证明过去每份财报在原公告时刻的版本，也不能证明供应商修订记录完整或经济口径正确。当前 TTM 的历史组成项可能已被后来修订。这是周末看到的财务快照与上一交易日价格的组合，不是周五收盘时的历史信号。
复现脚本为 scripts/analyze_valuation_definitions.py：--vintage 指向 owner 封存快照，--daily-basic 指向一个已留存的每日指标分区，--output 必须是仓库外的新目录。公开的派生 JSON 保存输入哈希、计算日期、报告期及来源披露／观测日期，不包含供应商原始财务金额或价格日线。脚本固定本例六家公司，缺失扣除项仍不可用。换一个数据版本重跑就是新观察，不能当作同一快照。
哪些已经展示，哪些还没完成
factor-observatory 的 PB/ROE 专题 已展示在 PB 上加入盈利能力的历史比较；market-research 对应说明 保留数值、研究背景及限制。它没有比较本文全部五种分母。本次口径表与当前计算样本发布在 market-research；将来完成标准化因子研究后，再按已有发布契约进入 observatory。
后续须先冻结历史观测账本和全市场资格规则，分别报告共同样本、各指标可计算样本及按行业／年份的缺失率。金融与非金融应分开比较，并控制规模、行业和盈利能力，保留后来退市的公司。固定形成、执行、基准、修订和成本口径后，才能做配对收益检验。本次没有发布全市场“最低估值名单”，也没有验证哪种估值更赚钱。
一手字段来源
- TuShare 每日指标：市值单位及字段。
- TuShare 利润表：归母利润及报表类型。
- TuShare 资产负债表：权益、商誉及无形资产字段。
- TuShare 财务指标：供应商扣非利润字段。
Low valuation: define the denominator before ranking
There is no single “lowest valuation” ranking. PE compares price with earnings; PB compares it with recorded equity. Excluding nonrecurring earnings or recorded intangible assets changes the question and can change the order. The six-name illustration below demonstrates the definitions, not future returns or a market-wide stock screen.
Evidence stage: reviewed current-snapshot arithmetic, observed on 2026-10-18. Prices are from 2026-10-16. Financial records were retrieved on October 17 UTC (October 18 in China), including an October 17 disclosure for 002415.SZ. These inputs cannot be represented as available at the October 16 close. No historical strategy return has been calculated.
Definition table
Let M be total common-equity market value in yuan; P the parent-attributable profit over the last twelve months; D the parent-attributable profit excluding nonrecurring items; E the equity attributable to common shareholders; G and I the corresponding attributable goodwill and other recorded intangible assets.
| Measure | Intended definition | What must be fixed before comparison |
|---|---|---|
| PE TTM | M / P | Consolidated parent-attributable earnings, not total group profit; trailing year, not annualized half-year |
| PE excluding nonrecurring items | M / D | Same parent scope and trailing periods; removing nonrecurring items is not the same as removing all investment income |
| PB | M / E | Parent equity, excluding minority and any senior equity claims where applicable; a single balance-sheet date, not four-quarter sum |
| PB excluding goodwill | M / (E − G) | Ownership scope, goodwill definition and impairment version |
| PB excluding goodwill and intangibles | M / (E − G − I) | Recorded intangible scope, ownership allocation, and whether capitalized development costs are additionally excluded |
TTM at an interim date is current cumulative YTD + previous full year − matching previous YTD. An annual report uses its full-year value directly. Missing legs leave the result unavailable. Earnings, equity and deductions must use the same consolidated reporting scope; do not mix parent-company-only statements with consolidated parent-attributable data.
M must cover the same equity claim as the denominator. This illustration deliberately chooses six A-only issuers, without inferring a general treatment for A/H, B shares, preference shares or convertible instruments. An A-share price multiplied by all classes’ shares is not automatically the economic value of every class. Circulating market value is also unsuitable against full-company earnings/equity.
Data mapping and important qualifications
The script maps daily_basic.total_mv × 10,000 to yuan, income.n_income_attr_p to P, fina_indicator.profit_dedt to a supplier-defined nonrecurring-profit proxy, and balancesheet.total_hldr_eqy_exc_min_int to parent equity. The supplier’s profit_dedt label does not independently certify its parent/common-share scope for every issuer. Source reports and preference-share allocations have not been independently checked; therefore the sample’s PB is labelled parent-equity PB and its nonrecurring PE remains provisional.
goodwill and intan_assets are group-consolidated asset balances, while the equity field excludes minority interests. Their full deduction from parent equity is an unallocated asset-deduction proxy, not certified parent tangible equity. There is no assumed pro-rata minority allocation. The two source fields are treated separately; capitalized development costs, deferred tax consequences, off-balance-sheet intangibles and future liquidation recoveries are not included. A recorded asset’s book value is not its liquidation value. The sample does not claim intangible-intensive firms are economically worse.
Missing goodwill or intangible values remain missing; they are never silently replaced by zero. A nonpositive denominator is unavailable for positive-multiple cheapness rankings, with loss/negative-equity cases requiring separate analysis. This convention does not mean such companies have no economic value. In this sample Moutai has an unavailable goodwill field and is excluded from the common ranking rather than assigned zero goodwill.
Six-name illustration
Purposively chosen issuers span several business models and are not a representative cross-section. Multiples below share each issuer’s latest period available across income, indicator and balance-sheet sources. Report periods may differ between issuers. Figures are rounded only for display. On narrow screens, scroll the numeric tables horizontally.
| Code | Issuer | Report period | PE TTM | Nonrecurring PE proxy | Parent PB | Ex-goodwill PB proxy | Ex-goodwill/intangibles PB proxy |
|---|---|---|---|---|---|---|---|
| 000651.SZ | Gree | 20260630 | 7.94 | 8.36 | 1.49 | 1.50 | 1.61 |
| 002027.SZ | Focus Media | 20260630 | 21.86 | 29.28 | 5.33 | 5.39 | 5.43 |
| 002415.SZ | Hikvision | 20260930 | 17.34 | 17.70 | 3.66 | 3.67 | 3.75 |
| 300059.SZ | East Money | 20260630 | 19.65 | 20.28 | 2.91 | 3.00 | 3.00 |
| 600519.SH | Kweichow Moutai | 20260630 | 20.40 | 20.42 | 6.61 | — | — |
| 600900.SH | China Yangtze Power | 20260630 | 19.86 | 20.81 | 3.33 | 3.34 | 3.77 |
For Focus Media, removing the supplier-defined nonrecurring component changes PE from 21.86 to 29.28. In the five-name common sample, Hikvision and Yangtze Power reverse their relative PB order after goodwill and recorded intangibles are deducted. Neither observation establishes better subsequent returns.
Same-sample ranks
All five columns below use exactly the same five issuers with positive finite denominators in every measure. Rank 1 is the lowest multiple in this small sample; ties use the minimum rank. The sixth issuer’s available PE and PB values are still shown above, but it receives no rank in any common-sample column. Cross-industry ranks here illustrate arithmetic only.
| Code | PE | Nonrecurring PE proxy | Parent PB | Ex-goodwill proxy | Ex-goodwill/intangibles proxy |
|---|---|---|---|---|---|
| 000651.SZ | 1 | 1 | 1 | 1 | 1 |
| 002027.SZ | 5 | 5 | 5 | 5 | 5 |
| 002415.SZ | 2 | 2 | 4 | 4 | 3 |
| 300059.SZ | 3 | 3 | 2 | 2 | 2 |
| 600519.SH | — | — | — | — | — |
| 600900.SH | 4 | 4 | 3 | 3 | 4 |
Availability, revisions and reproduction
The analysis checks the owner archive seal, normalized child-manifest hashes and data hashes before reading selected issuers. It admits only report type 1 cumulative consolidated statements, selects the latest disclosed/observed version for each issuer-period, and requires date-only disclosures to precede the current local calendar day. Ambiguous same-version duplicates fail. The dated input archive was obtained through the owner’s configured TuShare route; no new supplier request or production alias change was needed.
Those integrity checks prove the retained files match their recorded hashes. They do not prove that older report values were available in their original historical form, that supplier restatements are complete, or that the supplier’s economic definitions are correct. Old legs in a current TTM calculation may reflect subsequently revised comparatives. The result is a weekend observed snapshot using the previous session’s prices; it is not a Friday-close historical signal.
Reproduction uses scripts/analyze_valuation_definitions.py with --vintage pointing to an owner sealed vintage, --daily-basic to one retained daily partition, and --output to a new directory outside the repository. The reviewed derived JSON includes hashes, calculation dates, report periods and source observation/disclosure dates, without raw supplier financial amounts or price bars. The script checks the six fixed sample names and leaves missing deductions unavailable. Reproduction on another vintage is a new observation, not the same snapshot.
What was already published, and what remains
The PB/ROE study already compares profitability added to PB; the market-research companion retains its historical results and qualification. It does not compare all five denominators defined here. This definitions note and current arithmetic illustration belong to market-research; a standardized factor study may later be admitted to the observatory through its existing publication contract.
Next, freeze a historical observation ledger and full-universe eligibility rules, compare measures on common and measure-specific samples, and report missingness by sector and year. Separate financial and nonfinancial sectors, control size/industry/profitability, and retain later-delisted issuers. Only after fixing formation, execution, benchmark, revisions and costs should a paired return study estimate whether any denominator adds value. There is no all-market “cheapest stock” list or performance claim in this release.
Primary field references
- TuShare daily indicators: market-value unit and field mapping.
- TuShare income statements: parent-attributable earnings and report types.
- TuShare balance sheets: equity, goodwill and recorded intangible fields.
- TuShare financial indicators: supplier nonrecurring-profit field.