← Quantery Blog

How to test whether retained earnings create value

October 16, 2026 · 9 min readthesis-buildingcapital-allocationretained-earnings

Retained earnings are profits kept inside a company instead of distributed as dividends. Keeping them creates value only when management can turn that capital into more operating earning power without leaning on fresh debt or share issuance. A workable thesis therefore measures the capital kept, the improvement that followed, and any outside financing that helped produce it.

This isn't a one-ratio problem. The retained-earnings balance is a cumulative accounting record, and share price can move for reasons that have little to do with reinvestment. Build the test from flows reported over the same window. Then make management earn credit for the result.

What are retained earnings?

Net income closes into shareholders' equity after each reporting period. Dividends reduce what remains for owners inside the company. The accumulated balance is called retained earnings. It sits in the equity section of the balance sheet, but the movement begins on the income statement and passes through the statement of shareholders' equity.

The balance isn't a cash account. A manufacturer can retain profits and spend the cash on equipment. A software company can use it for payroll or an acquisition. Losses reduce the account even if the company still has cash from an earlier financing. How the three financial statements fit together covers this handoff in detail.

For a testing window, the useful economic estimate is:

capital retained = net income - common dividends

Use flows over the same period. Don't subtract today's retained-earnings balance from an old one and call the difference reinvestment. Prior-period adjustments, accounting changes, and transactions inside equity can move the reported balance. Net income less dividends states the claim more cleanly: this is profit that wasn't paid out during the window.

What does good reinvestment look like?

Start with a narrow claim:

Companies that retain a meaningful share of profits and convert it into higher operating income, without major dilution or a surge in debt, show better capital allocation.

Every clause has a job.

Retention says management had internally generated capital available. A retention ratio divides net income less dividends by net income. It works only when cumulative net income is positive. A loss-maker can produce a bizarre ratio, so return null and withhold the score.

Operating progress asks what the business produced afterward. Use operating income because financing and tax choices sit below it. Compare the newest fiscal quarter with the same fiscal quarter from a year earlier, or compare trailing windows when the history allows. This doesn't prove that the retained dollar caused the improvement. It does require an observable result in the operation.

Debt change checks whether creditors funded the same progress. More debt isn't automatically poor allocation. A refinancing or acquisition can be sensible. But a thesis about internally funded reinvestment can't give management full credit while borrowings rise sharply.

Diluted-share change catches capital raised from shareholders and dilution from compensation. Repurchases can offset grants, so a flat count shows the net effect rather than gross issuance.

Return on capital asks how much operating profit the current capital base produces. This thesis asks whether capital kept during a window led to progress. Read return on capital and its denominator choices before treating a high current return on capital as proof of good incremental investment.

Berkshire supplied a demanding test

Berkshire Hathaway's 1983 chairman's letter says retained earnings should be judged on a rolling five-year basis by whether each dollar kept created at least a dollar of market value for owners. The 1984 letter repeats the standard and explains why a multi-year average matters for a company whose annual results fluctuate.

That's a useful owner test. It is also a difficult first-stage screen. Market value absorbs interest rates, changes in valuation multiples, sector fashion, and expectations about years beyond the measurement window. A rising share price can make weak reinvestment look clever. A falling multiple can hide genuine operating progress.

Use the market-value test after the operating test. First ask whether retained profit was followed by better earning power per share and whether debt supplied part of the improvement. Then ask whether the market eventually recognized it. Those are separate results. Keeping them separate stops price appreciation from serving as both evidence and conclusion.

Berkshire's 2007 shareholder letter shows the business version through See's Candies. See's had $8 million of net tangible assets when Blue Chip Stamps bought it in 1972. By 2007, the business required $40 million of capital, while cumulative pretax earnings had reached $1.35 billion. The operation needed only $32 million of that for reinvestment. Everything else went to Berkshire (or, in the early years, to Blue Chip) for deployment elsewhere.

That example sharpens the question. Good allocation doesn't require every subsidiary to retain everything. A business with few attractive uses for capital can distribute cash to the parent or shareholders. The return available on the next dollar determines whether retention is justified. The quality of the company that earned the last dollar doesn't settle it.

How can you express the thesis in Quantery?

A direct long-window incremental-return calculation needs careful period alignment. The current DSL can sum recent flows and compare quarterly fields with lag(), but it shouldn't pretend a balance-sheet shortcut measures where every retained dollar went. A first pass can test the financial footprint: positive earnings, meaningful retention, operating improvement, and restrained outside financing.

The shape below is illustrative. The bundled templates are the reference for exact fields, and the Quantery thesis documentation covers the DSL.

# "Retained profit followed by operating progress" (illustrative)
params:
  min_history: 5
  retention_floor: 0.50
  operating_growth_floor: 0.05
  debt_change_ceiling: 0.05
  dilution_ceiling: 0.02

quality:
  flags:
    - { when: "history_periods < $min_history", flag: "short_history:{history_periods}" }

features:
  income_ttm: ttm(net_income)
  dividends_ttm: sum(dividends_paid, 4)
  retained_ttm: income_ttm - coalesce(dividends_ttm, 0)
  retention_ratio: if(income_ttm > 0, retained_ttm / income_ttm, null)

  op_now: newest(operating_income)
  op_prior: lag(operating_income, 4)
  op_growth: if(op_prior > 0, op_now / op_prior - 1, null)

  debt_now: newest(total_debt)
  debt_prior: lag(total_debt, 4)
  debt_change: if(income_ttm > 0,
    (debt_now - debt_prior) / income_ttm, null)

  shares_now: newest(shares_diluted)
  shares_prior: lag(shares_diluted, 4)
  share_growth: if(shares_prior > 0,
    shares_now / shares_prior - 1, null)

criteria:
  earnings_retained:
    rules:
      - { when: "is_null(retention_ratio)", score: 0, flag: no_retention_read }
      - { when: "retention_ratio >= $retention_floor", score: 1 }
      - { else: 0 }
  operating_result:
    rules:
      - { when: "is_null(op_growth)", score: 0, flag: no_operating_comparison }
      - { when: "op_growth >= $operating_growth_floor", score: 1 }
      - { else: 0 }
  outside_financing:
    rules:
      - { when: "is_null(debt_change) or is_null(share_growth)", score: 0, flag: no_financing_comparison }
      - { when: "debt_change <= $debt_change_ceiling and share_growth <= $dilution_ceiling", score: 1 }
      - { else: 0 }

gate:
  mode: strict

The thresholds are proposed research parameters. retention_floor asks for a meaningful amount of profit to stay in the company. operating_growth_floor demands some subsequent operating progress. The debt change is scaled by trailing net income, so its ceiling asks whether borrowing was small beside current earning power. None of those cutoffs defines good allocation for every business.

Retention and operating growth overlap in this compact version, so it can't establish a causal sequence. For the stronger test, evaluate retention during an earlier window and operating progress in a later one. If the DSL can't express those windows exactly, export the run record or maintain two frozen thesis versions. Don't disguise an overlapping comparison as “capital retained first, result later.”

Where will the rule mislead you?

Acquisitions can manufacture the shape. A company can retain earnings, buy another business, and report more operating income. The rule sees progress. It doesn't know whether the purchase earned an adequate return. Inspect acquisition cash flow, goodwill, and the purchase-price note. Goodwill and impairment charges show where an acquisition premium lands and how late failure can become visible.

Cyclical recovery can take the credit. Operating income may rebound because selling prices recovered or demand returned. A year-over-year increase after a trough says little about management's use of retained capital. Run the test across several filing windows and separate industries whose profits follow commodity prices.

Inflation can lift the numerator. Nominal operating income may grow while unit economics stand still. Margins and return on capital help distinguish higher prices from better economics, though neither settles the question alone.

Dividend data can be incomplete. Some filers don't populate a clean common-dividend field in every quarter. coalesce() prevents missing data from breaking the arithmetic, but it can also turn an uncollected payment into zero. Check the statement of cash flows and the statement of equity before accepting a high retention score.

Share count can hide gross issuance. Repurchases may offset employee grants. The diluted count records the net effect on each share, which is the right first pass for owners, but it doesn't grade the price paid for buybacks. Review issuance and repurchase cash flows separately.

One year is short. Investment can precede its payoff by several years. A factory under construction can depress current cash flow before it contributes to operating income. That is why Berkshire's standard used a longer window. Treat the compact rule as a filter for filing review. It can't produce a final capital-allocation grade.

How should you backtest retained-earnings quality?

Start with a profitable-company baseline. Add the retention rule, then operating progress, then the financing restraint check. Keep the universe, benchmark, reporting lag, rebalance cadence, and holding policy fixed. Testing each criterion family incrementally shows whether the allocation rules add information or just shrink the sample.

Use filing dates. A simulated decision can't see a quarter before its report became public. Keep the rules frozen and run them walk-forward on point-in-time fundamentals. The resulting returns are hypothetical and exclude costs.

Then inspect composition. High retention can cluster in younger companies, while dividends cluster in mature industries. Operating growth can favor businesses emerging from a weak comparison period. Compare sectors and size buckets before naming capital allocation as the mechanism.

Run a distribution policy control too. Replace the retention hurdle with a low-retention group while keeping profitability and valuation comparable. If both groups behave alike, retention may not be doing useful work. If the retained group differs, open the filings and ask whether the difference came from organic investment, acquisitions, debt, or dilution.

Reinvestment takes time, and one fiscal comparison is a noisy start. Freeze the definition before looking at results, preserve the weak runs, and test nearby round thresholds. The point isn't to certify management from a ratio. It is to make the capital-allocation claim explicit enough to fail.

Retained earnings give management another decision. A thesis should award credit only after the chosen use produces an observable result. Keep retention, operating progress, and outside financing in separate rows. When the proxy fails on a real filing, change it and run the test again. The rules are yours, including the rule that decides whether management earned the right to keep the next dollar.

Want to try this on your own rules? Quantery is free for 14 days: the full app, no card required.

← All articles