Do Projects Learn Across Space and Time? Evidence from the Olympics
Pith reviewed 2026-05-10 03:58 UTC · model grok-4.3
The pith
Olympics show no strategic learning across 64 years, with cost overruns staying flat despite repeated hosting.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The central claim is that projects do not learn across space and time. Analysis of Olympic cost overruns from 1960 to 2024 shows no sustained improvement over 64 years. Tactical learning occurs within each individual Games, but the combination of geographic distance, temporal gaps, and the temporary organizational form of each host committee produces a myopia of learning that stops local insights from aggregating into strategic improvement. The authors propose four strategies for overcoming the barrier—incremental, centralising, decentralising, and real options—and conclude that radical reform is required.
What carries the argument
Myopia of learning induced by spatiotemporality, the combination of geographic distance between hosts, temporal gaps between events, and the temporary organizational structure of each host committee that blocks aggregation of tactical into strategic learning.
Load-bearing premise
Cost overruns serve as a reliable proxy for the absence of higher-level strategic learning and that data from Olympics of different scopes, locations, and eras remain comparable without major adjustments.
What would settle it
A sustained decline in average cost overruns across Olympics held after 2000, or documented cases in which lessons from one host were systematically adopted by later hosts and produced measurable cost reductions.
Figures
read the original abstract
Do projects learn across space and time? The Olympics, among the largest publicly funded programmes in the world, offer a unique empirical setting. Theoretically, the Games seem ideal for generating "positive learning curves," driving down costs from one iteration to the next. In practice, they do not. Drawing on the concept of "myopia of learning," we argue that spatiotemporality (geographic distance, temporal gaps, and the temporary organisational form of each host committee) combines to block higher-level learning. Our analysis of cost overruns from 1960 to 2024 reveals no sustained improvement over 64 years. Tactical learning abounds, but none aggregates into strategic improvement. We propose four strategies for overcoming the spatiotemporal barrier (incremental, centralising, decentralising, and real options), arguing that radical reform is required.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims that Olympic Games projects fail to exhibit sustained learning across space and time. Drawing on cost overrun data from 1960 to 2024, it finds no downward trend in overruns despite repeated hosting, attributing this to 'myopia of learning' caused by geographic distance, temporal gaps between Games, and the temporary organizational form of each host committee. Tactical learning is said to occur locally but does not aggregate into strategic improvement; four reform strategies (incremental, centralising, decentralising, real options) are proposed to overcome the spatiotemporal barrier.
Significance. If the no-improvement result is robust to measurement and scope adjustments, the work would usefully extend organizational learning theory to mega-projects by documenting a concrete case where repeated exposure does not produce cumulative gains. It offers an empirical anchor for discussions of public-project governance and could inform policy on large infrastructure programs, though its contribution hinges on demonstrating that the flat overrun series reflects blocked learning rather than unadjusted heterogeneity.
major comments (2)
- [Data and empirical strategy] The central empirical claim (no sustained improvement over 64 years) rests on the cost-overrun series. The manuscript must demonstrate that final costs and budgets are defined and adjusted consistently across Games (inflation, PPP, scope changes such as venue count or infrastructure share). Without explicit harmonization steps or robustness checks, the flat trend cannot be confidently interpreted as evidence of blocked strategic learning rather than measurement artifacts.
- [Results] The regression or trend specification for overruns over time should include covariates for event scale and context (athlete numbers, host GDP per capita, infrastructure intensity). Omission of these leaves open whether the null result on temporal improvement is driven by changing project characteristics rather than the hypothesized spatiotemporal barriers.
minor comments (2)
- [Abstract and introduction] Clarify in the methods how 'tactical learning' is operationalized and evidenced independently of the cost-overrun metric, as the abstract asserts it 'abounds' but the link to the data is not immediately transparent.
- [Discussion and conclusions] The four proposed strategies would be strengthened by brief references to analogous reforms attempted in other mega-projects (e.g., centralized procurement in other international events) to illustrate feasibility.
Simulated Author's Rebuttal
We thank the referee for these targeted comments on the empirical foundation of our analysis. We agree that strengthening the measurement and specification details will improve the paper's ability to isolate the hypothesized spatiotemporal barriers to learning. Below we respond to each major comment and indicate the revisions we will undertake.
read point-by-point responses
-
Referee: [Data and empirical strategy] The central empirical claim (no sustained improvement over 64 years) rests on the cost-overrun series. The manuscript must demonstrate that final costs and budgets are defined and adjusted consistently across Games (inflation, PPP, scope changes such as venue count or infrastructure share). Without explicit harmonization steps or robustness checks, the flat trend cannot be confidently interpreted as evidence of blocked strategic learning rather than measurement artifacts.
Authors: We accept that the current draft does not provide a dedicated section on data harmonization. In the revision we will add a new subsection (and appendix table) that explicitly documents: (i) conversion of all nominal costs to constant 2024 USD using host-country CPI series; (ii) PPP adjustments drawn from World Bank International Comparison Program data for cross-country comparability; and (iii) controls for scope variation via venue count and the share of infrastructure spending. We will also report a robustness panel that re-estimates the trend using only Games with comparable scope definitions. These additions will allow readers to assess whether the flat overrun series survives measurement adjustments. revision: yes
-
Referee: [Results] The regression or trend specification for overruns over time should include covariates for event scale and context (athlete numbers, host GDP per capita, infrastructure intensity). Omission of these leaves open whether the null result on temporal improvement is driven by changing project characteristics rather than the hypothesized spatiotemporal barriers.
Authors: The referee is correct that the baseline specification in the manuscript is a simple time trend. We will revise the empirical section to include a multivariate regression that adds the suggested covariates (athlete numbers, host GDP per capita, and infrastructure intensity) along with host fixed effects. We will report both the original and augmented specifications and test whether the coefficient on the time trend remains statistically indistinguishable from zero. If the null result is robust, this will strengthen the claim that the absence of improvement is not an artifact of changing project characteristics. revision: yes
Circularity Check
No circularity: empirical trend derived from external historical data
full rationale
The paper's core claim rests on an empirical examination of cost overrun data across Olympic Games from 1960-2024, showing no downward trend. This is presented as evidence against aggregate strategic learning, with the distinction between tactical and strategic learning drawn from observed patterns rather than any self-definition, fitted parameter renamed as prediction, or self-citation chain. No equations, ansatzes, or uniqueness theorems are invoked that reduce the result to the inputs by construction. The analysis uses external benchmarks and historical records without internal fitting that forces the flat trend. Minor self-citation risk is absent from the provided text, and the derivation remains self-contained against the data.
Axiom & Free-Parameter Ledger
axioms (1)
- domain assumption Olympics constitute an ideal empirical setting for testing positive learning curves in projects
Reference graph
Works this paper leans on
-
[1]
Learning Curves in Manufacturing
https://archive.nytimes.com/www.nytimes.com/specials/olympics/0723/oly- traffic-art.html . Argote, Linda, and Dennis Epple. 1990. “Learning Curves in Manufacturing.” Science 247 (4945): 920–924. https://doi.org/10.1126/science.247.4945.920 . Argote, Linda, Bill McEvily, and Ray Reagans. 2003. “Managing Knowledge in Organizations: An Integrative Framework ...
-
[2]
“Temporal Adaptive Capacity: A Competency for Leading Organizations in Temporary Interorganizational Collaborations.” Group & Organization Management 49 (1): 114–140. https://doi.org/10.1177/10596011221110080 . Bjerke-Busch, Linn Slettum, and Sebastian Thorp. 2024. “Overcoming the Productivity Paradox in the Public Sector by Managing Deliberate Learning.”...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.