PB and ROE: historical comparison and evidence boundary
PB (price-to-book) compares a stock price with its accounting book value. ROE (return on equity) compares profit with book equity. This reviewed summary through 2026-08-31 checks whether adding ROE changed the historical result for the same candidate stocks.
The consolidated factor-research entry point is the Quant Factor Observatory ↗. This page preserves the original public review and market-research context.
In this historical comparison, adding static ROE lowered the result versus ranking the same stocks by PB alone.
From 2010-01-29 to 2026-08-31, the 50-stock PB ranking returned an annualized 12.05%, compared with 9.06% after adding static ROE, a gap of 2.99 percentage points. Maximum drawdowns were 46.74% and 45.40%. In this sample, the small reduction in drawdown came with a lower return. Both rankings still required usable ROE data, so the result applies to the shared eligible pool.
For research today, this is a useful baseline for testing whether quality information adds to value ranking. It does not settle the effect of ROE in other universes. Source research records were updated on 2026-09-18. Narrative and published figures were checked on 2026-10-30 against the linked method and evidence notes. The data still end on 2026-08-31. No new backtest was run, and current tradability remains unverified because filing availability, financial revisions and actual fills are not fully audited.
Five rankings, four metrics
The chart compares five rankings on the same candidate universe of 50 stocks, using the same fees and daily execution approximation. Bar length represents annualized portfolio return; annualized excess return is relative to the CSI All Share Total Return Index, and drawdown is the historical maximum drawdown of portfolio NAV. Values are labeled directly; bar length is not a forecast interval.
“PB only” refers only to the ranking signal. Eligibility still requires computable ROE, so this is not a strict pure-PB portfolio.
Missing ROE and quality conditions change the sample
Removing the non-null ROE eligibility rule increases eligible stock-month records from 611,508 to 658,311: 46,803 additional records, or 7.11% of the expanded sample. Stock-month records are not counts of unique securities or portfolio holdings.
A separate experiment in the expanded universe jointly toggles three quality conditions: goodwill, leverage, and revenue. Its returns and drawdowns cannot be directly differenced against the five within-universe rankings above or attributed to any one condition.
With the conditions enabled, return is higher but maximum drawdown deepens from 44.60% to 48.26%. All three conditions change together, so this is not a single-variable causal estimate.
Results have been reviewed; the full experiment still needs independent replication
Data run through 2026-08-31. The public summary was checked against result CSVs, run manifests, and audit records in a private source project. Source code, raw data, and complete artifacts are not public, and this site has not independently rerun the full history. Financial-report availability dates have not been independently checked filing by filing. The historical period informed the research and is not a strict out-of-sample test. Close-price execution approximations do not include full market impact or replay of actual fills.
These results describe historical research only. They do not support replacing a production strategy or trading on the results.