Skip to main content
SpringerLink
Log in

Earnings prediction with DuPont components and calibration by life cycle

  • Published:
Review of Accounting Studies Aims and scope Submit manuscript

Abstract

(Soliman, The Accounting Review 83:823–853, 2008) finds that separating return on net operating assets (RNOA) into DuPont components—profit margin and asset turnover—improves prediction of future RNOA. (Dickinson, The Accounting Review 86:1969–1994, 2011) finds that a firm’s life-cycle stage explains changes in future RNOA. (Vorst and Yohn, The Accounting Review 93:357–381, 2018) find that life-cycle calibration improves prediction more than industry grouping in prediction models that do not include the DuPont components. We unite and extend the above studies by using data updated since the early 2000s and performing out-of-sample tests. We show that the DuPont components continue to improve prediction of one-year-ahead RNOA. Industry grouping and life-cycle calibration using the components improve prediction further. The improvement by life-cycle calibration is stronger for mature companies, more R&D-intensive companies, less capital-intensive companies, and companies in less concentrated industries. Sell-side equity analysts and investors appear to initially rely more on basic prediction models than the expanded models that include DuPont components, industry grouping, and life-cycle calibration. While there is some evidence of investor surprise associated with the expanded models, hedge portfolios formed based on the expanded model predictions do not produce abnormal returns.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1

Similar content being viewed by others

Data Availability

All data used and analyzed in this study are publicly available.

Notes

  1. https://www.bbc.com/news/business-53257933

  2. https://www.businessinsider.in/thelife/news/tesla-will-be-the-most-profitable-player-in-electric-vehicles-for-years-to-come-ubs-analysts-say/articleshow/81317654.cms

  3. The CFRA report is available upon request.

  4. RNOA (return on net operating assets) is preferred over ROA (return on assets) in measuring a firm’s fundamental performance, because RNOA eliminates non-operating factors, such as changes in financing costs.

  5. Vorst and Yohn (2018) compare life-cycle calibration and industry grouping for a basic model without the DuPont components and find that life-cycle calibration has more predictive ability than industry grouping. We perform a similar comparison using the DuPont components and find that life-cycle calibration adds more predictive power than industry grouping. However, we report sequential tests to clearly present how the predictive ability of the model with DuPont components progressively improves with industry grouping and life-cycle calibration.

  6. Low industry concentration means there are more firms competing in an industry.

  7. Vorst and Yohn (2018) suggest that heterogeneity across industries and homogeneity within industries are necessary for industry grouping to improve prediction of profitability. Vorst and Yohn do not find evidence that these conditions hold for their analysis of RNOA without DuPont components. These conditions are likely to hold for PM and ATO.

  8. The 10-year rolling windows are used to populate the cross-sectional tests with sufficient observations. This choice is based on the premise that the predictive properties of earnings components are stable over the rolling 10-year periods. In untabulated tests, we perform pooled regressions using year fixed effects and find that the reported results do not change qualitatively, suggesting that this assumption holds.

  9. In contrast to Fairfield and Yohn (2001), Soliman (2008) and Dickinson (2011) document a negative coefficient on ∆RNOAt.

  10. Soliman (2008) uses levels and changes of PM and ATO in separate regressions. Dickinson (2011) only uses changes in PM and ATO. In contrast, we use both the levels and changes of PM and ATO in the same model, Eq. (2).

  11. We sort firms into life-cycle stages each year but do not require that they stay in the same life-cycle stage over time. That is, annual regressions for a life-cycle stage in year t use all firm-year observations that are in that life-cycle stage during a given year between years t-9 and t.

  12. Because there is common information across the model predictions, analyst reliance on one model does not mean that the coefficients on the predictions by the other models would be zero.

  13. The dependent variable for month m + 12, Cumulative Returns[m, m+12], represents returns cumulated from one day before the earnings announcement for year t to two days before the earnings announcement for year t + 1 (which occurs in month m + 12). In cases of irregular earnings announcements, i.e., when the calendar month for the earnings announcements for year t and t + 1 differ, month m + 12 either replaces month m + 11 or is extended by up to 30 days. Observations where the earnings announcement month for year t + 1 is within 10 months of the earnings announcement month for year t are dropped, as are observations where the month m + 12 period goes beyond 60 days.

  14. We use equal-weighted returns, instead of value-weighted returns, to avoid dominance by a few large firms and to maintain consistency with the regression-based tests of Eqs. (4a) to (4d).

  15. We perform our analyses with the four-digit and six-digit GICS codes, and the results are consistent with our main results.

  16. We modify the shake-out classification by excluding firm-year observations that have cash flow patterns of (+ , + , +) and (-, -, -). Dickinson (2011) classifies these patterns as shake-out. These patterns are rare and transitory. We thus reason that they do not capture the true shake-out stage.

  17. Soliman (2008) found that the coefficient estimates on PMt and ∆PMt are insignificant. Similarly, Dickinson (2011) found that the coefficient estimate on ∆PMt is insignificant.

  18. The highest variation inflation factor (VIF) in Table 2 is 2.35, which mitigates multicollinearity concerns.

  19. We perform a sensitivity test regarding the classification of intangible capital. Following Dickinson (2011), we reclassify life-cycle stages by considering research and development (R&D) as investing cash flows, instead of operating cash flows. Furthermore, we adjust RNOA and the DuPont components, PM and ATO, by assuming R&D capitalization and subsequent amortization (Brown and Kimbrough 2011; Kothari, Laguerre, and Leone 2002). Specifically, we add pro forma R&D capital to net operating assets (NOA) and add back R&D expenses to and subtract amortization of pro forma R&D capital from operating income. While these changes improve the in-sample estimation, absolute forecasts errors using the DuPont model and life-cycle model do not change substantially.

  20. Hedge portfolio returns are regressed on monthly MKTj, SMBj, HMLj, and MOMj risk factor returns, which are obtained from Ken French’s website.

References

Download references

Acknowledgements

We would like to thank Scott Richardson (the editor) and the anonymous reviewer for their helpful comments and suggestions.

Author information

Authors and Affiliations

  1. Haskayne School of Business, University of Calgary, Scurfield Hall, 2500 University Dr NW, Calgary, AB, T2N 1N4, Canada

    Mark Anderson

  2. Bryan School of Business and Economics, University of North Carolina at Greensboro, 516 Stirling Street, Greensboro, NC, 27412, USA

    Soonchul Hyun

  3. Bauer College of Business, University of Houston, 4250 Martin Luther King Blvd, Houston, TX, 77204, USA

    Volkan Muslu

  4. Ted Rogers School of Management, Toronto Metropolitan University (Formerly Ryerson University), 350 Victoria Street, Toronto, ON, M5B 2K3, Canada

    Dongning Yu

Authors
  1. Mark Anderson
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Soonchul Hyun
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Volkan Muslu
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Dongning Yu
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Dongning Yu.

Additional information

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendices

Appendix 1

Table 8.

Table 8 Life-cycle stages
Full size table

Appendix 2

Table 9.

Table 9 Variable definitions
Full size table

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Anderson, M., Hyun, S., Muslu, V. et al. Earnings prediction with DuPont components and calibration by life cycle. Rev Account Stud 29, 1456–1490 (2024). https://doi.org/10.1007/s11142-022-09748-3

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11142-022-09748-3

Keywords

JEL Classification

Search

Navigation