In newsvendor experiments, the pull-to-center effect — average orders lying between mean demand and the profit-maximizing quantity — is a robust aggregate phenomenon, but Lau, Hasija and Bearden (2014) showed it does not describe individuals, who are highly heterogeneous: some order at the optimum, some systematically above or below it, and only some pull toward the demand mean. The surviving open problem their paper opened is to establish a validated individual-level account of newsvendor ordering — measurable factors that predict an individual's ordering strategy relative to the optimal quantity, an individual-level behavioral model that predicts individual orders out of sample across experimental designs and populations, and a criterion for when aggregate pull-to-center summaries are acceptable substitutes for individual-level modeling. Later work has partially advanced each direction (cognitive-reflection and cultural predictors; prospect-theory, adaptive-learning, and reference-dependent models accommodating heterogeneity; a 24-experiment meta-analysis and cross-task replications), but no comprehensive predictor account, no benchmarked out-of-sample individual-level model, and no aggregate-versus-individual diagnostic exists, so the core remains open.

Background

The newsvendor problem is the elementary stochastic inventory model: a decision maker must order a quantity qq of a product before demand DD (a random variable with known distribution) is realized; each unit sold yields margin p−cp-c for unit price pp and unit cost cc, unsold units are salvaged, and unfilled demand is lost, so the expected-profit-maximizing order q∗q^{*} solves the critical fractile condition — the demand quantile at which the marginal benefit of ordering one more unit equals its marginal cost, e.g. F(q∗)=(p−c)/pF(q^{*})=(p-c)/p when unsold units have zero salvage value, where FF is the cumulative demand distribution. Behavioral operations studies this decision experimentally with incentivized human subjects. Schweitzer and Cachon (2000, Management Science 46(3):404-420) documented the pull-to-center (PTC) effect: across subjects and periods, average orders lie strictly between mean demand and q∗q^{*} — too low in high-margin (high critical fractile) conditions and too high in low-margin conditions — and judged explanations based on risk aversion, loss aversion, and other utility preferences inadequate. Proposed mechanisms include anchoring on mean demand with insufficient adjustment, demand chasing (reacting to the most recent demand realization), random decision noise around the optimum (Su 2008, Manufacturing & Service Operations Management 10(4):566-589, whose bounded-rationality noise model reproduces PTC in aggregate), adaptive learning, in which order decisions are updated over rounds in response to realized demand and profits (Bostian, Holt and Smith 2008, Manufacturing & Service Operations Management 10(4):590-608), and reference-dependent preferences in the sense of prospect theory — the descriptive theory of risky choice where outcomes are evaluated as gains and losses relative to a reference point with a value function that is concave for gains, convex for losses, and steeper for losses; Long and Nasiry (2015, Management Science 61(12):3009-3012) showed such a model can be consistent with PTC. Nearly all of this evidence was aggregate. Lau, Hasija and Bearden (2014, Decision Support Systems 58:68-73) re-examined the PTC claim at the individual level, using individual-level data from a previously published study plus a new experiment: PTC appears in group-level statistics, but the population of individual decision makers is highly heterogeneous — individuals differ persistently in their position relative to q∗q^{*} (at, above, or below the optimum) and, on a separate descriptive axis, in how far their orders are pulled toward mean demand — so aggregate summaries misdescribe most individuals. The paper's abstract explicitly flags methodological implications and future research directions, which can be grouped into four linked directions: (1) identifying cognitive, demographic, experiential, or contextual factors that systematically predict individual ordering relative to q∗q^{*}; (2) developing and validating individual-level behavioral models or parametrizations (for example, parameterized newsvendor types) that capture the observed mix of strategies and support prescriptive design robust to heterogeneity; (3) formalizing when aggregate summaries are an acceptable simplification versus when individual-level modeling is required; and (4) replicating the individual-level analyses across designs, populations, payment schemes, and demand environments. The unresolved scientific target is to replace the aggregate PTC description with a predictive individual-level account of newsvendor decision making.

Problem Statement

The problem, as opened by Lau, Hasija and Bearden (2014), is to replace the aggregate pull-to-center description of newsvendor behavior — the robust finding that average orders lie between mean demand and the optimum — with a validated individual-level account. Specifically: (1) identify cognitive, demographic, experiential, or contextual factors that systematically predict whether an individual decision maker orders above, below, or at the expected-profit-maximizing quantity q∗q^{*} — the primary, mutually exclusive outcome axis, distinct from the separate descriptive axis of pull toward mean demand — in newsvendor tasks with known demand distributions — moving beyond single-factor correlations (cognitive reflection, i.e. the tendency to override an intuitive but incorrect first response, as well as culture and personality) to a replicated predictor account of individual ordering strategy that identifies, for each factor, the direction and robustness of its association with the individual-level ordering outcome; (2) develop and validate individual-level behavioral models or parametrizations that capture the observed mix of individual strategies and demonstrably predict individual order decisions out of sample against benchmark models, not merely accommodate aggregate patterns or in-sample fits, across experimental designs, populations, payment schemes, and demand environments; and (3) establish, formally or empirically, a criterion for when aggregate summaries of ordering behavior are an acceptable simplification and when individual-level modeling is required, so that prescriptive designs (contracts, decision support) can be made robust to heterogeneity. These three components jointly serve one scientific objective — an explanatory and predictive individual-level science of newsvendor decision making.

References

  1. Lau, N., Hasija, S., & Bearden, J. N. (2014). Newsvendor pull-to-center reconsidered. Decision Support Systems, 58, 68–73. DOI: 10.1016/j.dss.2012.12.041, https://doi.org/10.1016/j.dss.2012.12.041
  2. Schweitzer, M. E., & Cachon, G. P. (2000). Decision bias in the newsvendor problem with a known demand distribution: Experimental evidence. Management Science, 46(3), 404–420. DOI: 10.1287/mnsc.46.3.404.12070, https://doi.org/10.1287/mnsc.46.3.404.12070
  3. Su, X. (2008). Bounded rationality in newsvendor models. Manufacturing & Service Operations Management, 10(4), 566–589. DOI: 10.1287/msom.1070.0200, https://doi.org/10.1287/msom.1070.0200
  4. Bostian, A. J. A., Holt, C. A., & Smith, A. M. (2008). Newsvendor 'pull-to-center' effect: Adaptive learning in a laboratory experiment. Manufacturing & Service Operations Management, 10(4), 590–608. DOI: 10.1287/msom.1080.0228, https://doi.org/10.1287/msom.1080.0228
  5. Long, X., & Nasiry, J. (2015). Prospect theory explains newsvendor behavior: The role of reference points. Management Science, 61(12), 3009–3012. DOI: 10.1287/mnsc.2014.2050, https://doi.org/10.1287/mnsc.2014.2050
  6. Moritz, B. B., Hill, A. V., & Donohue, K. L. (2013). Individual differences in the newsvendor problem: Behavior and cognitive reflection. Journal of Operations Management, 31(1–2), 72–85. DOI: 10.1016/j.jom.2012.11.006, https://doi.org/10.1016/j.jom.2012.11.006
  7. Köster, C., & Schenk-Mathes, H. Y. (2016). Explanatory and predictive power of the adaptive learning model: Average and heterogeneous behavior in a newsvendor context. Journal of Business Economics, 86(4), 361–387. DOI: 10.1007/s11573-016-0814-8, https://doi.org/10.1007/s11573-016-0814-8
  8. Zhang, Y., & Siemsen, E. (2019). A meta-analysis of newsvendor experiments: Revisiting the pull-to-center asymmetry. Production and Operations Management, 28(1), 140–156. DOI: 10.1111/poms.12899, https://doi.org/10.1111/poms.12899
  9. Uppari, B. S., & Hasija, S. (2019). Modeling newsvendor behavior: A prospect theory approach. Manufacturing & Service Operations Management, 21(3), 481–500. DOI: 10.1287/msom.2017.0701, https://doi.org/10.1287/msom.2017.0701
  10. Vipin, B., & Amit, R. K. (2019). Describing decision bias in the newsvendor problem: A prospect theory model. Omega, 82, 132–141. DOI: 10.1016/j.omega.2017.12.008, https://doi.org/10.1016/j.omega.2017.12.008
  11. Li, X., Chen, L. G., & Chen, J. (2019). Individual and cultural differences in newsvendor decision making. International Journal of Operations & Production Management, 39(1), 164–186. DOI: 10.1108/IJOPM-01-2018-0042, https://doi.org/10.1108/IJOPM-01-2018-0042
  12. Brokešová, Z., Deck, C., & Péliová, J. (2022). Pull-to-center is not just for newsvendors. PLoS ONE, 17(2), e0264183. DOI: 10.1371/journal.pone.0264183, https://doi.org/10.1371/journal.pone.0264183
  13. Brokešová, Z., Deck, C., & Péliová, J. (2025). Experience minimizes the pull-to-center effect in newsvendor decisions. Manufacturing & Service Operations Management, 27(2), 554–568. DOI: 10.1287/msom.2021.0439, https://doi.org/10.1287/msom.2021.0439
  14. Niranjan, T. T., Ghosalya, N. K., Menon, R. R., Rotaru, K., & Gavirneni, S. (2023). Unpacking the cognitive processes of the boundedly rational newsvendor. Production and Operations Management, 32(10), 3138–3157. DOI: 10.1111/poms.14027, https://doi.org/10.1111/poms.14027
  15. Perera, H. N., Fahimnia, B., & Tokar, T. (2020). Inventory and ordering decisions: A systematic review on research driven through behavioral experiments. International Journal of Operations & Production Management, 40(7/8), 997–1039. DOI: 10.1108/IJOPM-05-2019-0339, https://doi.org/10.1108/IJOPM-05-2019-0339
  16. Yamini, S., & Gajanand, M. S. (2021). Inventory decision-making biases: A review and suggestions for future research. Benchmarking: An International Journal, 29(6), 1889–1912. DOI: 10.1108/BIJ-01-2021-0061, https://doi.org/10.1108/BIJ-01-2021-0061
  17. Jammernegg, W., Kischka, P., & Silbermayr, L. (2022). Heterogeneity, asymmetry and applicability of behavioral newsvendor models. European Journal of Operational Research, 301(2), 638–646. DOI: 10.1016/j.ejor.2021.11.014, https://doi.org/10.1016/j.ejor.2021.11.014
  18. Jammernegg, W., Kischka, P., & Silbermayr, L. (2026). Supply chain coordination with a wholesale price contract for behavioral decision-makers. European Journal of Operational Research, 329(1), 112–123. DOI: 10.1016/j.ejor.2025.06.029, https://doi.org/10.1016/j.ejor.2025.06.029
  19. Kirshner, S. N., & Moritz, B. B. (2021). Measuring demand chasing behavior. Decision Sciences, 52(6), 1264–1281. DOI: 10.1111/deci.12518, https://doi.org/10.1111/deci.12518
  20. Akbay, Ü., & Kaya, M. (2022). Effect of personality on newsvendor decisions. Proceedings of the International Conference on Industrial Engineering and Operations Management, 1445–1452. DOI: 10.46254/na07.20220331, https://doi.org/10.46254/na07.20220331
  21. Akbay, Ü. (2023). A cross-national examination of the newsvendor behavior. Proceedings of the International Conference on Industrial Engineering and Operations Management, 1474–1482. DOI: 10.46254/au01.20220337, https://doi.org/10.46254/au01.20220337