Residual Income Valuation and Stock Returns: Evidence from a Value-to-Price Investment Strategy

Ahmad Haboub; Aris Kartsaklas; Vasilis Sarafidis

REVIEW 2 major objections 6 minor 10 references

Residual Income Valuation and Stock Returns: Evidence from a Value-to-Price Investment Strategy

T0 review · 2 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that portfolio sorts on the residual-income value-to-price (V/P) ratio generate excess returns over one to three years, and that the Fama–French five-factor model cannot explain all of these returns, especially for…

desk verdict Replicates the V/P anomaly carefully, but the new years-2/3 factor alphas rely on overlapping-return OLS that needs HAC correction before the central claim is trustworthy. read the letter →

arxiv 2506.00206 v1 pith:KRJK6YVO submitted 2025-05-30 econ.EM

Ahmad Haboub , Aris Kartsaklas , Vasilis Sarafidis This is my paper

classification econ.EM

keywords residualincomevaluationvalue-to-priceratioV/Panomalyvalueinvestingassetpricingfactorsmispricinglong-horizonreturnsFama-Frenchfive-factormodel

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tests a simple accounting-based value signal: the ratio of residual-income-model intrinsic value to market price (V/P). It finds that high-V/P portfolios beat low-V/P portfolios by 5.3%, 13.7%, and 27.8% in raw buy-and-hold returns over 12, 24, and 36 months. After adjusting for the Fama–French five factors, the top V/P quintile still earns monthly alphas of about 0.004 to 0.005 in overlapping, second-year, and third-year returns. The paper argues this persistent unexplained return favors a mispricing interpretation over a pure risk explanation, and that standard factor models leave a meaningful part of the value-to-price anomaly unexplained.

What carries the argument

The key machinery is the residual income valuation model of Ohlson (1995) and Feltham and Ohlson (1995), implemented as a system of forecasting and valuation equations (book value, abnormal earnings, other information, and market value) estimated jointly by seemingly unrelated regression. Intrinsic value V is computed out-of-sample for each firm-year-industry using a jackknife procedure, and V/P is then used to form quintile portfolios. The paper's long-horizon evidence comes from an overlapping portfolio construction that holds current, prior-year, and two-year-old cohorts in each month, and from regressions of these portfolio excess returns on the CAPM, Fama–French three-factor, and five-factor models, with the Gibbons–Ross–Shanken statistic used to test whether all alphas are jointly zero.

What would settle it

Re-estimate the five-factor alphas for the highest-V/P portfolio in years two and three using Newey–West or two-way clustered standard errors on the overlapping monthly return series; if the t-statistics fall below conventional thresholds, the paper's claim that factors cannot explain all variation loses its statistical support.

Watch

Extended reading notes

Core claim

The central discovery is that the V/P effect is not a one-year phenomenon: high-V/P stocks keep earning risk-adjusted excess returns in the second and third years after portfolio formation. In the paper's numbers, the highest-V/P quintile produces a five-factor alpha of roughly 0.004 to 0.005 per month (t-values about 2.34 to 2.66) in overlapping, year-two, and year-three returns, and these alphas rise when dividends are included. The paper also shows that V/P predicts 36-month buy-and-hold returns at the firm level after controlling for beta, size, idiosyncratic volatility, leverage, distress, profitability volatility, and book-to-market. Because the highest-V/P stocks load positively on the investment factor in year two and on the profitability factor in year three yet still earn positive alpha, the paper concludes these stocks are significantly mispriced relative to their future equity growth and profitability persistence.

Load-bearing premise

The significance of the long-horizon alphas relies on treating monthly returns from overlapping three-year holding periods as statistically independent, even though those returns are mechanically correlated because the same stock cohorts appear in many consecutive months.

Editorial extensions

If this is right

If the V/P anomaly persists into years two and three, a simple accounting-based value signal can be used to build multi-year value strategies that outperform book-to-market sorting.
The five-factor model's failure to price high-V/P stocks in later years implies either an omitted risk dimension or a slow correction of mispricing, both of which invite further asset-pricing work.
Profitability and investment factors matter for middle-V/P portfolios but not for the extreme high-V/P portfolio, suggesting the high-V/P premium is not just the standard profitability or investment premium.
Firm-level regressions showing V/P predicts 36-month returns after controlling for common risk proxies strengthen the case that the V/P effect is a distinct anomaly, not a restatement of size or value effects.
Dividend-adjusted returns show even larger long-horizon alphas, so the strategy's apparent outperformance is not an artifact of ignoring dividends.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The reported t-statistics for year-two and year-three alphas come from overlapping portfolios, which mechanically induces serial correlation in monthly returns; correcting standard errors for this overlap could weaken the significance of the long-horizon alphas.
The paper's own characteristics of high-V/P stocks (small size, high idiosyncratic volatility, low analyst coverage) align with limits-to-arbitrage explanations, suggesting the mispricing story could be tested further by conditioning on arbitrage frictions.
A natural extension is to test whether the V/P alpha survives the q-factor model or a model that includes a separate V/P mimicking factor, and whether similar long-horizon alphas appear in international markets.
Because high-V/P stocks load on investment and profitability factors in ways that vary by year, decomposing the anomaly around earnings announcements could reveal whether the post-year-one returns reflect slow fundamental information diffusion.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Two things to know. First, this paper delivers a clean replication of the V/P anomaly: portfolio spreads over one, two, and three years are large, monotonic, and survive size adjustment, and the authors are careful to apply Newey–West corrections in the buy-and-hold comparisons. Second, the genuinely new part—Feltham-Ohlson SUR valuation with conservatism, year-two/year-three factor analysis, and dividend-adjusted returns—is worth taking seriously, but the claim that the five-factor model cannot explain years two and three depends on OLS t-statistics for overlapping monthly portfolios, and that inference is not yet credible.

What the paper does well: the valuation procedure is out-of-sample (jackknife), so V/P is not fit to the returns it predicts. The replication of Frankel and Lee is done with a longer sample, and the comparison to B/M is useful. The GRS tests and factor-loading patterns are informative. The paper makes an honest attempt to distinguish mispricing from risk using firm-level regressions, and the literature coverage is thorough.

Where it is soft: the main hole is the one the stress-test note identifies. The overlapping portfolios in Tables 5–8 are formed from cohorts rebalanced annually, so each month's return shares stocks with the previous year's portfolio. That mechanically induces serial correlation at lags up to 23 months. The paper applies Newey–West in Table 1 but not in the factor regressions. The t-stats on Q5 alphas around 2.3–2.7 would likely shrink under HAC or cohort-clustered errors; whether they remain significant is an open question. The firm-level V/P coefficient in the full model has t=1.9, which is marginal. These are fixable with a few lines of code, but they need to be fixed before the “cannot explain all variation” conclusion can stand. Minor issues include some typos (“weekly” for “weakly,” “corelated”) and a repetitive literature review.

Bottom line: this paper is a solid empirical exercise that mostly replicates known results. The incremental contribution—years two and three and dividend-adjusted alphas—is plausible but not yet established because of the inference problem. It deserves a serious referee; I would send it out with a requirement that the authors re-run the factor regressions with serial-correlation-robust or cluster-robust standard errors and report whether the Q5 alphas survive. If they do, this is a nice paper. If they don't, it's a replication with a weaker extension, but still cite-worthy as a careful out-of-sample test.

Referee Report

2 major / 6 minor

Summary. This paper constructs a residual-income value-to-price (V/P) measure using the Feltham-Ohlson/Ohlson linear information dynamics estimated by out-of-sample industry-year SUR, forms US quintile portfolios, and tests whether V/P-sorted returns are explained by the CAPM, the Fama-French three-factor model, and the Fama-French five-factor model over one-, two-, and three-year horizons. The main empirical findings are: (i) high-V/P portfolios outperform low-V/P portfolios, with spreads increasing from 12 to 36 months; (ii) at the firm level, V/P retains a positive coefficient after controlling for risk proxies, though only marginally in the full model; and (iii) five-factor alphas for the highest V/P portfolio are significant in overlapping returns and in years 2 and 3, leading the authors to conclude that standard factor models cannot fully explain the V/P effect. The paper also studies dividend-adjusted returns and the characteristics of high-V/P firms.

Significance. If the long-horizon results hold, the paper makes a useful contribution by extending the V/P anomaly literature beyond the usual first-year returns and by showing that profitability and investment factors do not absorb high-V/P returns in years 2 and 3. The V/P measure is constructed from accounting data and analyst forecasts via an out-of-sample SUR procedure that does not use the returns used for testing, so circularity is not a concern; the factor models are external benchmarks. The paper's explicit use of Newey-West corrections in Table 1 for overlapping buy-and-hold returns is a positive feature, and the literature review is extensive. However, the central long-horizon claim depends on alphas whose reported t-statistics come from OLS regressions on overlapping annual cohorts, and this inference issue must be resolved before the contribution can be assessed.

major comments (2)

[§5.3.1–5.3.3, Tables 5–8] The Q5 five-factor alphas that carry the paper's central claim—overlapping returns of 0.004 (t=2.34) in Table 5 Panel C, and second- and third-year returns of 0.004 (t=2.37) and 0.005 (t=2.66) in Table 6 Panel C—are based on monthly OLS regressions of annually rebalanced portfolio return series. Because each annual cohort is held for 12 months or longer, the residuals have a 12-month block structure, and the effective number of independent formation cohorts is about 23 rather than 276. The paper applies Newey-West corrections for overlapping buy-and-hold returns in Table 1 but not to these factor regressions. Please report standard errors clustered by formation year (or Newey-West with at least 12 monthly lags) for all alphas and factor loadings in Tables 5–8, and adjust the GRS F-statistics accordingly. If the Q5 alphas in years 2 and 3 are no longer significant after this correction, the conclusion that 'these factors cannot explain all variation in excess returns especially for years two and three' loses its empirical support.
[§5.2, Table 4, Eq. (4)] In the full model of Eq. (4), the V/P coefficient is 0.013 with t=1.9 and is displayed with two stars, but with 11,693 observations the two-sided p-value is approximately 0.057, so the coefficient is not significant at the 5% level. The text's statement that 'the coefficient on V/P remains significant and positive' and that 'omission of risk factors is not a likely explanation of the V/P effect' is therefore overstated. Additionally, Ret36 for adjacent firm-years overlap by up to 24 months, so the OLS standard errors in Table 4 may understate uncertainty even beyond the marginal t-statistic. Please report exact p-values, use one-sided tests only with explicit justification, and consider clustering by firm or formation year.

minor comments (6)

[§2.3] The reference 'Frankel and Lee (1988)' appears to be a typo for 'Frankel and Lee (1998)' and should be corrected.
[Tables 7–8 footnotes] The definition of overlapping returns in the footnotes of Tables 7 and 8 says the portfolio consists of first-year returns from year t, second-year returns from year t-2, and third-year returns from year t-3, but Tables 5 and 6 describe the cohorts as t, t-1, and t-2; this inconsistency should be fixed.
[Table A1] The table title says the period is 1987–2000, but the rows extend to 2014; the title should be updated to reflect the full sample period.
[Table 3] The row 'Years' reports 1993–2014 for Model I and 1992–2014 for Model II; the discrepancy should be explained or aligned.
[§5.3.1] The text refers to 'Fame and French (2015)' in one instance; this is a typo for 'Fama and French (2015)'.
[Table 2 notes] The Altman Z-score formula in the Table 2 notes uses coefficients 0.012, 0.014, 0.033, 0.006, and 0.999, while the appendix uses the standard 1.2, 1.4, 3.3, 0.6, and 1.0; the table note should be corrected for consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: V/P is built from out-of-sample residual-income estimates, and the factor regressions use external benchmarks.

full rationale

The paper's V/P measure is not fitted to the returns it later predicts. The intrinsic value V is estimated from a system of residual-income forecasting equations and a valuation equation, with parameters estimated by SUR under a jackknife that excludes the target firm's own data for the prediction year. Although the valuation equation uses market value as the dependent variable, for any given firm-year the fitted parameters come from other firms, so V is a genuine out-of-sample function of accounting fundamentals and analyst forecasts, not of the target firm's future return or of the target firm's price in the estimation period. The subsequent tests regress V/P-sorted portfolio returns on external Fama-French factors downloaded from Ken French's website; no parameter in those regressions is used to construct V/P. The only self-citation is Sarafidis and Wansbeek (2012, 2021) as methodological support for the SUR estimator, which is not a load-bearing premise for the empirical findings. The paper also acknowledges and corrects overlap in buy-and-hold returns using Newey-West in Table 1, while the monthly factor regressions use plain OLS; that is a potential inference robustness concern, not circularity. No equation reduces to an identity, no fitted input is renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work. The empirical alphas are outputs of the data and factor models, with no definitional equivalence to the V/P construction.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the residual income valuation model, whose parameters are estimated from accounting data (listed above), and on the choice of a 12% discount rate. No new entities are introduced. The main domain assumptions are the validity of clean-surplus accounting, linear information dynamics, analysts' forecasts as expectations, and the use of FF5 factors as the risk benchmark.

free parameters (6)

Discount rate r = 12% (main analysis); 8-16% and CAPM/FF3 rates in robustness
Chosen by hand for the residual income valuation; affects the level of V and hence the V/P ratio. Robustness checks are reported.
Residual income persistence omega_11 = 0.732
Estimated from accounting data via SUR; appears in the valuation equation (1d) and the other-information calculation (2a).
Conservatism parameter omega_12 = 0.017
Estimated from accounting data via SUR; enters equation (1a) and the valuation mapping.
Other information persistence omega_13 = 0.426
Estimated from accounting data via SUR; enters the valuation equation (1d).
Book value growth parameter omega_22 = 1.05
Estimated from book value dynamics in equation (1b); affects the valuation mapping.
Other information autoregressive parameter omega_33 = 0.56
Estimated from the other-information process in equation (1c); enters the valuation equation.

assumptions (6)

standard math Clean surplus accounting and linear information dynamics of Ohlson (1995) and Feltham and Ohlson (1995)
Equations (1a)-(1d) and the residual income valuation formula (A1) rely on these assumptions.
domain assumption Analysts' consensus forecasts are unbiased expectations of future earnings
Used to construct the other information variable v in equations (2a) and (2b).
domain assumption Common risk proxies (beta, size, book-to-market, leverage, analyst coverage, Altman's Z, ROA volatility) adequately capture cross-sectional risk
Equations (3) and (4) use these as risk controls when evaluating the V/P effect.
domain assumption The Fama-French five-factor model is the correct risk benchmark for evaluating abnormal returns
Positive alphas from the FF5 regressions are interpreted as evidence that standard factors do not explain V/P returns.
domain assumption Sample filters (price > $1, positive book value, positive forecasts, non-financial firms, December fiscal year end) do not introduce selection bias that drives the anomaly
Section 4 applies these filters; they could affect the generality of the results.
ad hoc to paper A fixed 12% discount rate approximates the cost of equity capital for valuation
Chosen for simplicity and consistency with prior studies; the paper reports robustness to alternative rates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Residual Income Valuation and Stock Returns: Evidence from a Value-to-Price Investment Strategy." pith.science (2026). https://pith.science/paper/KRJK6YVO

@misc{pith2026250600206,
  author       = {Pith},
  title        = {Pith review of: Residual Income Valuation and Stock Returns: Evidence from a Value-to-Price Investment Strategy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KRJK6YVO}},
  note         = {Machine review of arXiv:2506.00206}
}

read the original abstract

We hypothesize that portfolio sorts based on the V/P ratio generate excess returns and consist of companies that are undervalued for prolonged periods. Results, for the US market show that high V/P portfolios outperform low V/P portfolios across horizons extending from one to three years. The V/P ratio is positively correlated to future stock returns after controlling for firm characteristics, which are well known risk proxies. Findings also indicate that profitability and investment add explanatory power to the Fama and French three factor model and for stocks with V/P ratio close to 1. However, these factors cannot explain all variation in excess returns especially for years two and three and for stocks with high V/P ratio. Finally, portfolios with the highest V/P stocks select companies that are significantly mispriced relative to their equity (investment) and profitability growth persistence in the future.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages

[1]

- Consumer Non-Durables - Food, Tobacco, Textiles, Apparel, Leather, Toys (SIC code: 0100–0999, 2000–2399, 2700–2749, 2770–2799, 3100–3199, 3940–3989)

Non-Dur. - Consumer Non-Durables - Food, Tobacco, Textiles, Apparel, Leather, Toys (SIC code: 0100–0999, 2000–2399, 2700–2749, 2770–2799, 3100–3199, 3940–3989)

work page 2000
[2]

- Consumer Durables - Cars, TVs, Furniture, Household Appliances (SIC code: 2500–2519, 2590–2599, 3630–3659, 3710–3711, 3714, 3716, 3750–3751, 3792, 3900–3939, 3990–3999)

Dur. - Consumer Durables - Cars, TVs, Furniture, Household Appliances (SIC code: 2500–2519, 2590–2599, 3630–3659, 3710–3711, 3714, 3716, 3750–3751, 3792, 3900–3939, 3990–3999)

work page
[3]

Manufac. - Manufacturing - Machinery, Trucks, Planes, Office Furniture, Paper, Com Printing (SIC code: 2520– 2589, 2600–2699, 2750–2769, 3000–3099, 3200–3569, 3580–3629, 3700–3709, 3712–3713, 3715, 3717–3749, 3752–3791, 3793–3799, 3830–3839, 3860–3899)

work page
[4]

Energy - Oil, Gas and Coal Extraction and Products (SIC code: 1200–1399, 2900–2999)

work page
[5]

Chemical - Chemicals and Allied Products (SIC code: 2800–2829, 2840–2899)

work page
[6]

- Business Equipment, Computers, Software, Electronic Equipment (SIC code: 3570–3579, 3660–3692, 3694–3699, 3810–3829, 7370–7379)

Equip. - Business Equipment, Computers, Software, Electronic Equipment (SIC code: 3570–3579, 3660–3692, 3694–3699, 3810–3829, 7370–7379)

work page
[7]

Telecom- Telephone and Television Transmission (SIC code: 4800–4899)

work page
[8]

Laundries, Repair Shops) (SIC code: 5000–5999, 7200–7299, 7600–7699)

Utility - Utilities (SIC code: 4900–4949) 9)Retail - Wholesale, Retail, Services (e.g. Laundries, Repair Shops) (SIC code: 5000–5999, 7200–7299, 7600–7699)

work page

Show all 10 references

[10]

Health - Healthcare, Medical Equipment, Drugs (SIC code: 2830–2839, 3693, 3840–3859, 8000–8099)
[11]

Other - Other Mines, Construction, Materials, Transport, Hotels, Services, Entertainment. 44 45 Table 1 Characteristics of quantile-portfolio formed by ME, B/P and V/P ratios Panel A- Market equity portfolios (In sample size quintiles) Q1 Low ME Q2 Q3 Q4 Q5 High ME All Firms Q...

1987

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

- Consumer Non-Durables - Food, Tobacco, Textiles, Apparel, Leather, Toys (SIC code: 0100–0999, 2000–2399, 2700–2749, 2770–2799, 3100–3199, 3940–3989)

Non-Dur. - Consumer Non-Durables - Food, Tobacco, Textiles, Apparel, Leather, Toys (SIC code: 0100–0999, 2000–2399, 2700–2749, 2770–2799, 3100–3199, 3940–3989)

work page 2000

[2] [2]

- Consumer Durables - Cars, TVs, Furniture, Household Appliances (SIC code: 2500–2519, 2590–2599, 3630–3659, 3710–3711, 3714, 3716, 3750–3751, 3792, 3900–3939, 3990–3999)

Dur. - Consumer Durables - Cars, TVs, Furniture, Household Appliances (SIC code: 2500–2519, 2590–2599, 3630–3659, 3710–3711, 3714, 3716, 3750–3751, 3792, 3900–3939, 3990–3999)

work page

[3] [3]

Manufac. - Manufacturing - Machinery, Trucks, Planes, Office Furniture, Paper, Com Printing (SIC code: 2520– 2589, 2600–2699, 2750–2769, 3000–3099, 3200–3569, 3580–3629, 3700–3709, 3712–3713, 3715, 3717–3749, 3752–3791, 3793–3799, 3830–3839, 3860–3899)

work page

[4] [4]

Energy - Oil, Gas and Coal Extraction and Products (SIC code: 1200–1399, 2900–2999)

work page

[5] [5]

Chemical - Chemicals and Allied Products (SIC code: 2800–2829, 2840–2899)

work page

[6] [6]

- Business Equipment, Computers, Software, Electronic Equipment (SIC code: 3570–3579, 3660–3692, 3694–3699, 3810–3829, 7370–7379)

Equip. - Business Equipment, Computers, Software, Electronic Equipment (SIC code: 3570–3579, 3660–3692, 3694–3699, 3810–3829, 7370–7379)

work page

[7] [7]

Telecom- Telephone and Television Transmission (SIC code: 4800–4899)

work page

[8] [8]

Laundries, Repair Shops) (SIC code: 5000–5999, 7200–7299, 7600–7699)

Utility - Utilities (SIC code: 4900–4949) 9)Retail - Wholesale, Retail, Services (e.g. Laundries, Repair Shops) (SIC code: 5000–5999, 7200–7299, 7600–7699)

work page