---
id: "20260724-0213-does-benkard-2000-report"
title: "Does Benkard (2000) report the organizational-forgetting depreciation rate (~0.96 monthly, ~61% of experience surviving a year) that the L-1011 learning-reversal claim rests on?"
type: "capture"
status: "promoted"
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-07-24T00:00:00.000Z"
provenance: "Research session, 2026-07-24, resolving [[question-verify-benkard-2000-organizational-forgetting-rate]]"
derived_from: []
tags: ["economics","learning-curve","organizational-forgetting","aircraft-manufacturing","benkard","sourcing-verification"]
source_url: "https://web.stanford.edu/~lanierb/research/Learning_and_Forgetting_AER.pdf"
source_author: "C. Lanier Benkard"
source_date: 2000
source_venue: "American Economic Review 90(4), pp. 1034-1054, September 2000"
source_quote: "The value 5 [δ] = 0.96 implies that 61 percent (0.96^12) of the firm's stock of experience existing at the beginning of a year survives to the end of the year."
source_tier: 1
promoted_to: ["30-notes/claim-organizational-forgetting-can-reverse-the-learning-curve.md (revised in place: source_tier upgraded 3→1, quote replaced with Benkard's own words, [unverified-quant] flag cleared, status seedling→budding)","claim-benkard-2000-estimate-implies-far-less-forgetting-than-prior-studies.md","claim-benkard-2000-headline-depreciation-figure-is-the-spillover-adjusted-specification.md","entity-c-lanier-benkard.md","entity-linda-argote.md","entity-lockheed-l-1011-tristar.md","entity-organizational-forgetting.md"]
not_promoted: ["Mechanism claim ('greater production rates -> greater average productivity') — not a separate note; folded into the revision of claim-organizational-forgetting-can-reverse-the-learning-curve.md as an upgraded (Tier 1) restatement of a mechanism already asserted there.","Learning-rate estimate under the general model (35-40% vs ~18-30% traditional, Table 3) — a distinct future claim per the capture's own 'further leads,' but no exact quote was captured this session, so it fails the Tier 1-2 quant sourcing floor's quote requirement for now. Left as a lead, not promoted.","Spillover magnitude details beyond the delta figures already used (lambda~0.70; -500 variant needing ~25% more labor; joint production running unit costs ~11% higher) — same reason: paraphrased in the capture's 'further leads,' not directly quoted, so not promotable as quant claims yet.","Argote (1990) and Darr (1995) primaries unread directly — worth an independent check per the capture's own suggestion, but no kept claim rests on their own papers being read (the comparison claim promoted here is scoped to what Benkard himself reports about them, which is Tier 1). No question routed: this is a nice-to-verify, not a load-bearing doubt.","Counterfactual production-scheduling claim (Table 5 / Figure 5, bunched production would have yielded higher productivity) — a distinct, separately promotable claim about optimal scheduling under forgetting, out of scope for this capture's question. Left in the capture for a future promotion pass.","NBER working paper precursor (WP 7127, 1999) unchecked against the published AER version for numerical differences — a bibliographic lead, not a claim in its own right."]
---


**Yes.** The primary source was located and read directly (fetched via `extract_pdf` from the author's own Stanford faculty page, `tls: verified`, OCR-extracted, 21 pages) and it confirms both the depreciation parameter and the annual-survival figure in the author's own words. This resolves [[question-verify-benkard-2000-organizational-forgetting-rate]] and clears the `[unverified-quant]` flag on [[claim-organizational-forgetting-can-reverse-the-learning-curve]].

## Claim: Benkard's own estimated depreciation parameter for L-1011 experience is δ = 0.96, implying 61% annual survival

Benkard models experience as depreciating by a factor δ each month: "this paper uses the specification introduced by Argote et al. (1990), which has experience depreciating by a factor δ each month. Experience in a given month is then the cumulative depreciated production experience of all production that took place prior to that month." His preferred specification (regression 10 in Table 3, which includes incomplete spillovers across L-1011 model variants) estimates δ = 0.960. He interprets this explicitly: "The value 5 [δ] = 0.96 implies that 61 percent (0.96^12) of the firm's stock of experience existing at the beginning of a year survives to the end of the year." (OCR renders "δ" as "5" and the exponent "12" as "!" in the extracted text; the underlying math — 0.96 raised to the 12th power ≈ 0.613 — confirms the reading.)

- source_url: https://web.stanford.edu/~lanierb/research/Learning_and_Forgetting_AER.pdf
- source_author: C. Lanier Benkard
- source_date: 2000 (American Economic Review 90(4):1034-1054)
- source_quote: "The value 5 [δ] = 0.96 implies that 61 percent (0.96^12) of the firm's stock of experience existing at the beginning of a year survives to the end of the year."
- source_tier: 1 (primary — the author's own writeup of his own estimation, hosted on his own faculty page; `tls: verified`)

This directly supersedes the earlier PLOS ONE secondary summary (Tier 3) that the existing claim-note rested on; the number is no longer resting on a paraphrase.

## Claim: the 0.96 figure is from the model *with* incomplete spillovers; the forgetting-only specification (no spillovers) estimates a lower δ = 0.952

Table 3 reports two general-learning-model regressions. Regression 9 ("OF only," organizational forgetting with complete spillovers assumed, λ=1) estimates δ = 0.952. Regression 10 (adding the incomplete-spillovers parameter λ=0.70) estimates δ = 0.960 — the figure Benkard uses for the 61%-survival interpretation. Benkard explains why: "The estimate of δ rises significantly when the model is extended to allow for incomplete spillovers... Leaving out incomplete spillovers from the model thus leads to a downward bias in the experience depreciation parameter." The headline ~0.96/~61% figure is therefore specifically the spillover-adjusted estimate, not the simplest specification's.

- source_url: https://web.stanford.edu/~lanierb/research/Learning_and_Forgetting_AER.pdf
- source_author: C. Lanier Benkard
- source_date: 2000
- source_quote: "The estimate of δ rises significantly when the model is extended to allow for incomplete spillovers... Leaving out incomplete spillovers from the model thus leads to a downward bias in the experience depreciation parameter."
- source_tier: 1

## Claim: Benkard's estimated depreciation rate is far higher (much less forgetting) than the two prior empirical estimates it builds on

Benkard directly compares his estimate against the two previous empirical studies that had proposed the same organizational-forgetting specification: "Despite the fact that the estimate of the depreciation factor for this data is much higher and more precise than previous results obtained by Argote et al. (1990) and Darr et al. (1995), the implied depreciation rate still seems quite high." A footnote quantifies the prior estimates: "In the two previous examples in the literature, the depreciation parameter was estimated to be in the range 0.75–0.85, suggesting that only 5 to 15 percent of experience was retained from year to year." So Benkard's L-1011 estimate (61% annual retention) reflects markedly *less* forgetting than the ship-production (Argote et al. 1990) and franchise (Darr et al. 1995) studies it follows, which found only 5–15% retained per year.

- source_url: https://web.stanford.edu/~lanierb/research/Learning_and_Forgetting_AER.pdf
- source_author: C. Lanier Benkard
- source_date: 2000
- source_quote: "In the two previous examples in the literature, the depreciation parameter was estimated to be in the range 0.75-0.85, suggesting that only 5 to 15 percent of experience was retained from year to year."
- source_tier: 1

## Claim: Benkard's model directly supports the mechanism that unit costs rise when production rates fall

The learning-reversal mechanism the L-1011 claim rests on — that a "permanent" cost gain can evaporate when production slows — is Benkard's own stated interpretation, not just a secondary gloss. He writes that under the organizational-forgetting model, "the number of units produced recently is most important in determining productivity, so that greater production rates can lead to greater average productivity," the inverse of which is that falling production rates erode productivity (raise unit labor cost). This generalizes and directly grounds the PLOS ONE secondary phrasing ("costs will rise when the rate of production falls") cited in [[claim-organizational-forgetting-can-reverse-the-learning-curve]], in Benkard's own words rather than a summarizer's paraphrase.

- source_url: https://web.stanford.edu/~lanierb/research/Learning_and_Forgetting_AER.pdf
- source_author: C. Lanier Benkard
- source_date: 2000
- source_quote: "In the organizational forgetting model the number of units produced recently is most important in determining productivity, so that greater production rates can lead to greater average productivity."
- source_tier: 1

## Further leads

- Learning rate under the general (forgetting-inclusive) model is estimated at 35–40%, versus ~18–30% under the traditional no-forgetting learning-curve model — Benkard (2000), same PDF, Table 3 discussion; a distinct claim about how forgetting changes the *learning-rate* estimate itself, not yet written up.
- Incomplete spillovers between L-1011 model variants (-1/-100/-200 vs -500) are estimated at λ≈0.70 (~70% task overlap); introducing the -500 required ~25% more labor than a -1 would have, and joint production of two models runs unit costs ~11% higher continuously — Benkard (2000), same PDF, Table 3 & discussion following.
- Argote, Beckman, and Epple (1990) and Darr, Argote, and Epple (1995) are the two prior empirical sources for the organizational-forgetting specification (ship production and franchise data respectively), estimating much faster forgetting (δ 0.75–0.85) — cited in Benkard (2000) footnote 40; primaries unread, worth a direct check if the vault wants the ship/franchise cases as independent evidence.
- Benkard's counterfactual production-schedule simulations (Table 5, Figure 5) suggest that bunching production (e.g., producing one model at double rate) would have yielded higher average productivity than Lockheed's actual erratic schedule — a distinct claim about optimal scheduling under forgetting, not the core question here.
- An NBER working-paper version (NBER WP 7127, May 1999) precedes the AER publication; not independently checked for numerical differences from the published version.

## Entity candidates

- C. Lanier Benkard — person — author of the primary paper; Stanford GSB economist; worth a page if the vault accumulates more of his work on aircraft/learning curves
- Lockheed L-1011 TriStar — concept — the aircraft program that is the empirical subject of both this claim and the existing vault claim-note
- Linda Argote — person — originator (with coauthors) of the organizational-forgetting specification Benkard adopts; prior lower-retention estimates from ship production
- Organizational forgetting (Argote et al. 1990 specification: E_t = δE_{t-1} + q_{t-1}) — concept — the formal depreciation model itself, distinct from the general claim already in the vault
- American Economic Review — concept — publication venue, useful for provenance tracking across vault economics claims