Accretion Dilution

Accretion / Dilution Model

先载入 xlsx-author 技能，再动手建表。 出处载体（Class A 的 Source: 批注 / Class B 的 来源 表加 来源编号 列）、四色含义、「填机构不填接口名」、URL 的两种诚实 写法、交付时的 ::zcode-file-citation，都在那份 SKILL.md 里。只照路径调它的 recalc.py 不等于读过它——2026-08-24 那批里有一份工作簿正是这样，五项全漏。

The core banking calculation: what happens to the acquirer's earnings per share the year after this deal closes, and why. The deliverable is an .xlsx workbook built with the xlsx-author conventions, recalculated before delivery, and audited before it is handed over.

This skill models one named transaction. It does not value the target as an LBO candidate — that is lbo-model in the write-research plugin — and it does not say whether the deal should be done.

Language of the deliverable

Authoring prose here is English. The workbook's labels, the delivery message, and the coverage block follow the user's language: a Chinese request produces a Chinese workbook, an English request an English one. Provenance tags follow the deliverable — the Chinese set [披露] [测算] [预期] [推断] [媒体] in a Chinese workbook, the English aliases [Reported] [Est.] [Consensus] [Inferred] [Media] in an English one. One style per document, never mixed.

Reference prices: filter by volume first. get_stock_performance does not skip suspended-trading days — it fills all four OHLC values with the previous close and nulls only the volume. Taking a "prior trading day close" therefore yields a carried-over price: the number is right, the base date is wrong, and nothing looks odd. A row with null volume is not a trading day. See ../deal-comps/references/stock-price-traps.md.

Non-negotiables

Environment. Build with Python/openpyxl following xlsx-author. Write formula strings (ws["D20"] = "=D12/D18"), never values computed in Python. Then run python3 ../xlsx-author/scripts/recalc.py "<the workbook you just wrote>" 30 (path relative to this skill's directory) before delivery.

Formulas over hardcodes. The only hardcoded numbers permitted are: retrieved historicals and market data, the announced or assumed deal terms, and the assumption drivers. Every EPS, every share count roll-forward, every synergy phase-in, every bridge line, and every breakeven cell is a live formula. If you catch yourself computing pro-forma EPS in Python and writing the result — stop. The model exists so the deal team can move the exchange ratio and watch the answer move.

Cell comments are this skill's citation vehicle (cell_comments). Every hardcoded input cell carries, written as the cell is created:

Source: <System or Document>, <Date>, <Reference>, <URL if applicable>

e.g. Source: 同花顺 iFinD 上市公司财务数据, 检索于 2026-07-25, [证券代码] FY2025 归母净利润 or Source: 发行股份购买资产暨关联交易预案, 2026-06-18, 第三节 「交易对价及支付方式」, https://….

A calculated cell carries a formula instead of a source comment. A source comment sitting on a calculated cell means a hardcode is hiding in it — find it and replace it with the formula.

Every synergy is [测算]/[测算]. It goes in the assumptions block with its basis and its phasing, it is shown as its own line in the bridge, and it is never an unexplained hardcode buried in a consideration or cost line. A reader must be able to zero the synergy block and see the deal without it.

Deal terms carry their class. A term from an announcement is [披露]/[披露] and its cell comment names the announcement. A term from a news report is [媒体]/[媒体]. A term we chose is [测算]/[测算] and lives in the assumptions block. Put the tag in the row label or the cell comment; never leave three classes sitting in one unlabelled column.

Step 1 — Resolve both sides and the mandate

Confirm before retrieving anything:

• Acquirer and target, each resolved to a code via get_stock_info (group vs listed vehicle vs subsidiary is a real ambiguity in A-share deals).
• Which side you are modelling for. The same deal reads differently from the buyer's and the seller's chair.
• Whether the deal is announced. If it is, the terms are retrievable and the model documents a real structure. If it is not, every term is [测算] and the workbook is explicitly a scenario, labelled as one in the header block.
• The fiscal year the pro-forma is struck for, and whether it is a full-year or stub-period pro-forma.

Step 2 — Retrieve standalone earnings

Call get_stock_financials for the acquirer and again for the target (one call per name), and take, for the periods you will show:

Do not average 归母 and 全口径, and do not carry them in one column. Where the source publishes a ratio you could also compute, show both and explain the definitional gap.

Step 3 — Retrieve share counts, and pick the right one

Call get_stock_shareholders for each side. It returns 总股本 and 流通股本 with 限售股, 流通A股 and 自由流通股 broken out, plus the top-ten holders and the 实际控制人.

This is where accretion models go wrong. 总股本, 流通股本 and 自由流通股 produce three different EPS numbers off identical earnings, and a model that silently uses the free-float count computes an EPS that no filing will ever agree with.

• Use 总股本 for EPS unless the user asks for something else. That is the count the reported EPS is struck on, and a pro-forma EPS that cannot be compared with a reported EPS is useless.
• Say which count you used, in the row label (Shares outstanding — 总股本), in the cell comment, and in the coverage block.
• State 基本 vs 摊薄. If the acquirer has convertibles, options, or unvested awards, either model them into the diluted count or state explicitly that the model is on a basic count and that dilutive instruments are not reflected.
• Use the same convention on both sides. Acquirer 总股本 against target 流通股本 is not a model, it is a mistake with a spreadsheet around it.
• 限售解禁 timing does not change the count today, but it changes the float, and it belongs in the notes when the consideration is stock.

Step 4 — Build the consideration and financing structure

Every term here is either retrieved from wind-docs.get_company_announcements ([披露]) or assumed ([测算], in the assumptions block):

• Offer value — per-share offer price × target shares, or a lump-sum equity value.
• Consideration mix — cash / stock / debt / rollover, in percentages that sum to 100% (assert the sum on the sheet).
• Exchange ratio — the acquirer shares issued per target share. Model it both as a direct input and as a derived value (offer price per target share / acquirer share price), and show which one drives the model.
• New shares issued = target shares × exchange ratio, on the stated share count.
• Financing cost — the interest rate on new acquisition debt, and the foregone interest on cash used. Foregone interest on the acquirer's own cash is the single most commonly omitted line in an accretion model; include it explicitly, even when the yield is small.
• Fees — advisory, financing, and their tax treatment. Financing fees are usually amortised, advisory fees usually expensed; whichever convention you use, state it.

Step 4.5 — 业绩承诺与补偿安排, where the target is unlisted equity

An A-share 重大资产重组 that buys unlisted equity almost always carries a 业绩承诺 with a compensation mechanism. Skipping it does not make the model conservative — it makes the dilution answer wrong in a knowable direction, because share compensation returns shares to the acquirer and therefore reduces dilution.

Retrieve the terms from wind-docs.get_company_announcements (they live in the 交易预案 / 重组草案 / 问询函回复) and, where the deal is already on the record, wind-stock.get_stock_events (并购指标, 备考财务数据). Record:

• 承诺期与承诺数 — which years, and the committed 净利润 (state 归母 vs 全口径) per year.
• 补偿形式 — 股份补偿 / 现金补偿 / 两者混合, and the 补偿触发与计算公式 as disclosed.
• 股份补偿的回购注销安排 — how many shares return, at what price, and when.
• 减值测试补偿 — a separate obligation from the annual shortfall compensation; do not merge them.
• 承诺方 — who owes the compensation, and whether it is joint and several.

Model it as two columns, never one.

• The compensation column is a scenario, [测算], and its trigger assumption must be stated (「假设承诺期首年未达成，按公式补偿 X 股」). It is not a forecast that the commitment will be missed, and the deliverable says so.
• A 业绩承诺 is contingent consideration in substance. Say that plainly: the headline 对价 overstates the economics if compensation is likely, and understates the acquirer's downside if the 承诺方 cannot pay.
• Where the announcements were searched and carry no 承诺, record 检索范围内未发现 — never assume a deal has no commitment because none was found, and never invent terms. Where the target is a listed company or the deal is an asset purchase with no 承诺 mechanism, mark it 不适用 and say why.

Step 5 — Phase in the synergies

• Split revenue synergies from cost synergies, and both from cost to achieve. Never net them into one line.
• Phase over the years shown — a typical structure is a run-rate figure with a realisation percentage per year — and put the run-rate and the percentages in the assumptions block as separate inputs so the phasing can be flexed.
• Tax the synergies at the stated effective rate.
• Every one of these is [测算]/[测算] with its basis in the cell comment. "管理层 指引" is a basis and gets cited; "行业惯例" without a document is not.

Step 6 — Pro-forma earnings and the bridge

The bridge is the deliverable. Build it as an explicit vertical walk, each line a formula, so a reader can see exactly which effect drives the answer:

Acquirer standalone net income (归母)
(+) Target standalone net income (归母)
(+) Synergies, after tax                        [测算]
(−) Cost to achieve, after tax                  [测算]
(−) Incremental interest on new debt, after tax
(−) Foregone interest on cash used, after tax
(+/−) Financing fee amortisation, after tax
(+/−) Other pro-forma adjustments (PPA D&A, etc.), after tax
= Pro-forma net income

Acquirer standalone shares (总股本)
(+) New shares issued as consideration
= Pro-forma shares

Standalone EPS      = acquirer net income / acquirer shares
Pro-forma EPS       = pro-forma net income / pro-forma shares
Accretion/(dilution), absolute = pro-forma EPS − standalone EPS
Accretion/(dilution), %        = pro-forma EPS / standalone EPS − 1

Purchase price allocation belongs here only if you have enough disclosure to do it: if the announcement gives the identifiable-intangibles split, model the step-up amortisation; if it does not, write the line as n.d. with an open item rather than inventing an allocation. A guessed PPA silently moves the accretion answer.

Step 7 — Breakeven synergy and breakeven exchange ratio

These are what the client actually asks for. Neither is optional.

• Breakeven synergy — the after-tax synergy at which pro-forma EPS exactly equals standalone EPS, holding the consideration structure fixed. Build it as a live formula off the bridge (solve the bridge for the synergy line), not as a value you found by nudging an input. Present it next to the modelled synergy so the reader sees the cushion: "the deal breaks even at X of after-tax synergy; the model assumes Y".
• Breakeven exchange ratio — the exchange ratio at which pro-forma EPS equals standalone EPS, holding synergies fixed. Present it against the modelled ratio, and state what it implies for the offer price per target share at the current acquirer price.

Both are [测算]/[测算], both state the assumption set they hold fixed, and neither is a recommendation about what to pay.

Step 8 — Sensitivity

At minimum one grid, odd dimensions (5×5 or 7×7) so there is a true centre cell, with the axes symmetric around the base case ([base − 2Δ, base − Δ, base, base + Δ, base + 2Δ]):

• Exchange ratio (or offer price) × synergy run-rate → accretion/(dilution) %
• Optionally: cash/stock mix × financing rate → accretion/(dilution) %

The centre cell must equal the model's actual accretion figure — that is the proof the grid is wired correctly. Highlight it (#BDD7EE fill, bold). Every cell is a formula; no linear approximations, no placeholders, no "use Excel's Data Table feature".

Workbook structure

Five sheets, in dependency order:

The Checks sheet asserts, at minimum:

• Consideration mix percentages sum to 100%.
• Cash + stock + debt + rollover consideration = total offer value.
• Standalone net income + every bridge adjustment = pro-forma net income (difference = 0).
• Acquirer shares + new shares issued = pro-forma shares.
• Pro-forma EPS × pro-forma shares = pro-forma net income.
• Breakeven synergy, substituted back into the bridge, produces zero accretion.

Formatting

Per the house formatting policy, which arbitrates:

• Font colours — four, not three. Blue #0000FF hardcoded input · black formula · green #008000 link to another sheet · purple #800080 link on the same sheet with no calculation (=B9). The purple case is the one models drop, and audits check it.
• Fills — 3 blues and a grey. #1F4E79 header with white bold text · #BDD7EE grouping band and key outputs · #D9E1F2 input/assumption block · #F2F2F2 neutral. No greens, yellows, or oranges as decoration.
• Borders are mandatory and carry meaning: 1.5pt above section headers, 1.0pt under subtotals, 0.5pt interior grid, double rule under grand totals, thin vertical rule between the last historical and first pro-forma column.
• Numbers right-aligned, text left-aligned. Currency #,##0;(#,##0);"-", percentages 0.0%, EPS 0.00, multiples 0.0"x", exchange ratio 0.0000.
• State the unit and currency in the column or row header, never only in prose, and never mix 万元 and 亿元 in one column.

Common errors in this model specifically

Verification before delivery

1. Recalc. python3 ../xlsx-author/scripts/recalc.py "<the workbook you just wrote>" 30. Exit 0 clean · 2 errors found, fix every location and re-run · 3 recalc_unavailable — LibreOffice is missing, no formula was evaluated, and this is NOT a pass; run xlsx-author's substitute protocol (reference check on every formula, independent recompute in Python, assert the identities above) and record the limitation in the coverage block · 1 hard failure.
2. Audit. Run the model-scope audit conventions from the audit-xls skill (it ships with the write-research plugin; if it is not loaded in this session, work its checklist by hand). The merger-specific bugs it lists are exactly this model's failure modes: accretion/dilution using the wrong share count (pre- vs post-deal), synergies not phased in, purchase price allocation that does not balance, foregone interest on cash omitted, and transaction fees missing from sources and uses.
3. Colour and comment sweep. Every blue cell has a Source: comment; no calculated cell has one; same-sheet pass-throughs are purple; cross-sheet links are green.
4. Identity sweep. Every row on Checks passes.
5. Sanity. Signs correct, magnitudes plausible, no #REF! / #DIV/0! / #VALUE! / #NAME? anywhere.

Delivery — the coverage block

Close the delivery message with this block, verbatim heading, even when everything passed. Use ## 覆盖范围与局限 for a Chinese deliverable and ## Coverage and Limitations for an English one.

## 覆盖范围与局限
检索于 / Retrieved: <timestamp>

- Financials retrieved: <which items, which periods, for which entity, from which
  tool — e.g. acquirer and target FY2025 归母净利润, 有息负债, cash via hexin-stock
  get_stock_financials; share counts (总股本) via get_stock_shareholders>
- Share count basis: <总股本 / 流通股本 / 自由流通股; 基本 or 摊薄; whether dilutive
  instruments are modelled>
- Deal terms — announced: <terms taken from an announcement, each with its document
  and date> [披露]/[披露]
- Deal terms — assumed: <exchange ratio, mix, financing rate, fees, synergy run-rate
  and phasing, tax rate — each also [测算]/[测算] in its cell comment and in the
  assumptions block> [测算]/[测算]
- Deal terms — media only, uncorroborated: <or 检索范围内未发现> [媒体]/[媒体]
- Formula evaluation: <one of> recalc.py evaluated all N formulas via LibreOffice,
  zero errors (exit 0) / recalc.py could NOT evaluate the formulas (exit 3,
  LibreOffice unavailable) — only a static lint plus the openpyxl
  reference/recompute/identity checks ran; the model is NOT verified and the user
  should confirm the numbers on open
- 本次未能覆盖 / Not covered: <sources that failed or were out of scope, and what they
  would have covered — e.g. PPA detail not disclosed, so no step-up amortisation is
  modelled>
- 数据滞后性 / Known lag: <disclosure lag on the financials and the share register>

Where a check could not run, say 源不可用 and name the tool that failed. 检索范围内未发现 means the source was queried and returned nothing — it is never rendered as "there is no such term". recalc_unavailable is not a pass, and a workbook delivered without formula evaluation is never described as verified.

This model stages arithmetic for the deal team. It does not say whether the deal should be done, does not opine on fairness, and does not set a price.