Does Berger & Pope (2011) itself carry the 5.8–8.0 pp halftime-deficit figure and the loss-aversion attribution the vault recorded?

Short answer the claims below support: Yes. A working-paper text authored directly by Jonah Berger and Devin Pope ("Can Losing Lead to Winning?", contact emails [email protected] / [email protected] given in the paper itself) was located, downloaded as a PDF, and read via extract_pdf (TLS verified, sha256 recorded). It states the NBA figure in the authors' own words as "5.8 to 8.0 percentage points," matching Table 1's regression coefficients (.058, .062, .074, .080 across four specifications), and it explicitly frames the mechanism as loss aversion operating on a halftime-score reference point. This confirms, at the primary-source level, both pieces of the figure the vault previously held one layer removed (via the 2023 replication paper's restatement) — see claim-berger-pope-nba-halftime-deficit-raises-win-probability and the open question this capture answers, question-verify-berger-pope-2011-halftime-loss-aversion-primary.

One caveat travels with the "yes": the document read is dated "This Draft: August, 2009" — a pre-publication working-paper draft, not the final Management Science 2011 typeset article. See Claim 3.

Claim: The paper's own text states NBA teams losing at halftime win 5.8 to 8.0 percentage points more often than expected

Claim type: Quantitative (a specific percentage-point range) — requires Tier 1–2. Cleared.

The working paper states directly: "Looking across specifications, we find that teams that are losing at halftime in the NBA win 5.8 to 8.0 percentage points more often than expected." Table 1, Panel A (NBA data) reports the underlying logit marginal-effects coefficients across four specifications as .058, .074, .062, and .080 (all significant at p<.01), which is the source of the quoted range. The paper separately reports a smaller NCAA effect: "in the NCAA, losing at halftime leads to a 2.1 to 2.5 percentage point increase in winning probability" (Table 1, Panel B: .025, .023, .025, .021).

This is the same range — "between 5.8 and 8.0 percentage points" — already recorded in the vault via the 2023 four-sport replication's restatement (Klein Teeselink, van den Assem & van Dolder, quoted in claim-berger-pope-nba-halftime-deficit-raises-win-probability). The two independent readings (this working-paper text, and the replication paper's characterization of the published version) converge exactly, which is strong evidence the figure survived unchanged from working paper to journal publication.

Claim: The paper explicitly attributes the effect to loss aversion — a halftime score functioning as a reference point that increases effort for the team in the "loss domain"

Claim type: Technical-mechanism — requires Tier 1–2. Cleared.

The paper frames its entire contribution around loss aversion from the outset: "In this paper, we explore how another non-standard preference, loss aversion, affects motivation in tournament settings. Loss aversion – the kink in Kahneman and Tversky's (1979) Prospect Theory value function – predicts that individuals value avoiding losses relative to a reference point more than they value obtaining commensurate gains." It then states the specific mechanism applied to sports: "one's performance relative to their opponent during the competition may serve as a salient reference point. Compared to if they were winning (i.e., in the gain domain), one may be willing to exert more effort if they are losing during a competitive task (i.e., in the loss domain)."

The paper also reports a laboratory experiment designed specifically to isolate this mechanism from rival explanations (coaching, referee bias, or winning-team complacency): participants told they were "slightly behind" an opponent increased effort significantly more than controls or those told they were "slightly ahead," and being ahead did not reduce effort — evidence the authors read as ruling out complacency and pointing to loss-driven effort by the trailing side specifically. Authors' words: the lab results "cast doubt on the possibility that the winning team simply becomes complacent (as opposed to the losing team exerting more effort)."

Claim: The document read is a 2009 pre-publication working-paper draft, not the 2011 Management Science typeset article itself — a residual provenance gap

Claim type: Historical/provenance (uncontested but worth flagging explicitly since the topic question names the 2011 published paper specifically).

The PDF's own header reads "This Draft: August, 2009," and it is hosted as a mirror copy on Andrew Gelman's Columbia statistics site (sites.stat.columbia.edu/gelman/...), not on Wharton's site or INFORMS. The official published version — Berger, J. & Pope, D. (2011), "Can Losing Lead to Winning?", Management Science 57(5):817–827, DOI 10.1287/mnsc.1110.1328 — sits behind an INFORMS PubsOnline paywall; a direct fetch attempt this session returned HTTP 403 Forbidden and was not read.

The gap is narrow rather than open: the 2023 Klein Teeselink/van den Assem/van Dolder replication (already Tier 1 in the vault) independently quotes the published Berger & Pope (2011) result as "between 5.8 and 8.0 percentage points," an exact match to the working paper's own stated range. Two independent readings — this working draft directly, and a later paper's direct citation of the 2011 published article — agree on the number, which is strong (though not 100%-certain) evidence the figure is unchanged between the 2009 draft and the 2011 publication. [working-paper draft read directly and confirmed Tier 1; exact wording of the final 2011 typeset article itself remains unconfirmed — INFORMS paywalled, 403 this session].

Further leads

• The paper's own robustness check (Table 2) finds no comparable discontinuity at the 1st- or 3rd-quarter breaks in the NBA (coefficients <2pp, not significant) — used by the authors to rule out "lumpiness in the scoring" as a confound; not independently promoted here but reinforces the loss-aversion-specific-to-halftime framing.
• The NCAA effect (2.1–2.5 pp) is smaller than the NBA effect (5.8–8.0 pp); the paper attributes the difference partly to selection (more experienced/competitive players sorting into the NBA) — an unexplored angle relative to claim-losing-leads-to-winning-fails-to-replicate-across-four-sports's finding that even the NBA effect is era-bound.
• The paper cites Heath, Larrick & Wu (1999) "Goals as Reference Points" directly in its literature review — direct textual link to claim-goals-as-reference-points-below-goal-work-harder, not yet cross-verified against that note's own sourcing.
• An institutional/library route to the actual INFORMS-typeset 2011 PDF (as opposed to the 2009 working draft) would close the last narrow gap in this verification; not attempted here (paywalled, 403).

Entity candidates

• Jonah Berger — person — co-author; Wharton marketing professor, author of popular contagion/virality books; may warrant an entity page if the vault accumulates more of his work.
• Devin G. Pope — person — co-author; Wharton/Booth behavioral economist; recurring name in the vault's behavioral-economics cluster.
• regression discontinuity design — concept — the paper's core identification strategy (Thistlethwaite & Campbell 1960); underused elsewhere in the vault's methods vocabulary, could anchor a methods note.
• Prospect Theory (Kahneman & Tversky 1979) — concept — the theoretical engine behind loss aversion, cited directly as the source of the "kink" in the value function; likely already implicit in claim-goals-as-reference-points-below-goal-work-harder but not yet its own page.
• hypermotivation (Rick & Loewenstein 2008) — concept — cited in the paper as a related idea ("perceiving oneself as 'in a hole' leads to hypermotivation"); a plausible adjacent claim-note candidate.