---
title: Foresight vs. Forecasting
description: A primer on the difference between strategic foresight and forecasting.
  Forecasting predicts probable outcomes, foresight explores multiple plausible futures.
  The text traces both fields' intellectual lineage, where forecasting wins and fails,
  the evidence on whether foresight works, and how organizations combine the two.
doc_version: '1.0'
last_updated: 19-Jul-26
canonical_url: https://garden.johanneskleske.com/foresight-vs-forecasting
---

This is a [[Primer|primer]] on the difference between foresight and forecasting.

Two words get used as if they were interchangeable. They are not, and treating them as synonyms is expensive.

When an organization hires a foresight team and then asks it for a number, both sides lose. The team delivers scenarios; the client wanted a prediction, feels shortchanged, and files the work unread. The reverse failure is quieter and does more damage. A forecast that held up nicely for next quarter gets stretched to justify a decision fifteen years out, far past the horizon where its method retains any skill. One mistake wastes good foresight. The other trusts a forecast to carry weight it was never built to hold. Knowing which tool you are holding and what it can honestly do is the whole point of keeping the two words apart.

## Table of contents

- [What is the difference between foresight and forecasting?](#what-is-the-difference-between-foresight-and-forecasting){: .internal-link}
- [What is forecasting?](#what-is-forecasting){: .internal-link}
- [What is strategic foresight?](#what-is-strategic-foresight){: .internal-link}
- [Where do forecasting and foresight really differ?](#where-do-forecasting-and-foresight-really-differ){: .internal-link}
- [Is forecasting reliable? Where it wins and where it fails](#is-forecasting-reliable-where-it-wins-and-where-it-fails){: .internal-link}
- [Does foresight actually work?](#does-foresight-actually-work){: .internal-link}
- [Better together: how organizations combine both](#better-together-how-organizations-combine-both){: .internal-link}
- [A practitioner's view](#a-practitioners-view){: .internal-link}
- [Further reading and sources](#further-reading-and-sources){: .internal-link}

## What is the difference between foresight and forecasting?

**Forecasting estimates the most probable value of something by projecting patterns in past data forward, and it is judged by accuracy. Strategic foresight explores multiple plausible futures under deep uncertainty to improve the decisions made today, and it is judged by the quality of those decisions rather than by whether any single future arrives. Forecasting narrows uncertainty toward one expected outcome; foresight deliberately widens it into a set of alternatives. Forecasting asks what will happen. Foresight asks what could happen and what we should do about it.**

The two are not rivals. They answer different questions, hold different assumptions about how much of the future is knowable, and should be trusted within different limits. The table below sets out where they part company.

| Dimension                    | Forecasting                                                  | Strategic Foresight                                                       |
| ---------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------------------- |
| **Aim**                      | Estimate the most probable outcome                           | Explore multiple plausible futures to inform present decisions            |
| **Treatment of uncertainty** | Reduce it: model variation around a central estimate         | Explore it: map divergent, structurally different futures                 |
| **Time horizon**             | Short to medium, days to roughly five years                  | Long, roughly five to thirty years and beyond                             |
| **Methods**                  | Time-series models, econometrics, structured expert judgment | Scenario planning, horizon scanning, backcasting, causal layered analysis |
| **Output**                   | Point estimates or probability distributions                 | Narratives, scenario sets, strategic options                              |
| **Validation**               | Predictive accuracy, calibration, error metrics              | Decision quality, organizational learning, adaptability, robustness       |
| **Epistemic stance**         | The future as an extension of stable structures              | The future as plural, constructed, and open to choice                     |

Most confusion in the field comes from collapsing one column into the other. The sections that follow take each practice on its own terms, look at the evidence for what it does well and where it breaks, and then show how the two work together.

## What is forecasting?

Forecasting deserves to be taken seriously, not set up as a straw man. It is a mature scientific practice with a century of method behind it and an unusually honest record of self-examination.

### Three research traditions

Its modern backbone is statistical time-series analysis. The systematization of that work by George Box and Gwilym Jenkins, whose 1970 monograph brought autoregressive, moving-average, and ARIMA models into a single framework for analysis and prediction, remains a reference point in almost every empirical forecasting paper written since.[^1] The Box and Jenkins program carried a set of intellectual commitments that still shape the field: treat forecasting as empirical and iterative, check models against out-of-sample data, balance fit against parsimony, and separate genuine structure from noise. Alongside it grew the econometric tradition, from the structural models of Jan Tinbergen and Lawrence Klein through vector autoregressions and general-equilibrium models, which powers the macroeconomic projections that institutions like the International Monetary Fund publish twice a year. A third strand, judgmental forecasting, studies how humans predict, and J. Scott Armstrong's 2001 handbook synthesized decades of comparative accuracy research into a set of evidence-based principles.[^2]

### What the tournaments show

What lifts forecasting above opinion is that its claims get scored. The most influential test bed is the series of forecasting competitions run by Spyros Makridakis since the late 1970s, the M-Competitions, which pit dozens of methods against thousands of real time series and report the errors in the open.[^3] Three findings recur across forty years of these tournaments, and each one is humbling for anyone selling complexity.

- **Simple often beats complex.** Across large sets of heterogeneous series, plain methods such as exponential smoothing match or outperform elaborate models, and in the M4 competition (run in 2018, results published in 2020), only a single hybrid entry, combining machine learning with exponential smoothing, beat the statistical combination benchmark across all series.[^3]
- **Combination beats any single method.** Averaging several forecasts reliably lowers error, which is why weather services run model ensembles rather than one perfect model.
- **Accuracy decays with horizon.** Errors grow as the forecast reaches further out, and beyond a certain point every method converges toward the skill of simple persistence, meaning that guessing “roughly the same as now” becomes hard to beat.

The same discipline shows up in forecasting human judgment about the world. Philip Tetlock's twenty-year study, reported in *Expert Political Judgment* in 2005, collected more than 27,000 dated, probabilistic predictions from 284 experts and scored them. On average the experts barely outperformed chance over a few years, and they were badly overconfident.[^4] But the study is more nuanced than its “experts do no better than dart-throwing chimps” caricature. Tetlock found that cognitive style mattered: “foxes,” who hold many partial models and update often, beat “hedgehogs,” who force events through one big idea. His later work with the Good Judgment Project identified “superforecasters,” ordinary people whose calibrated, frequently updated forecasts outscored intelligence analysts with access to classified material.[^4]

Two boundary conditions are essential, and forecasting researchers state them plainly.

- **Skill is real, but it is near-term.** Tetlock's strong results concern well-defined questions resolving within one to two years, and he warns explicitly that beyond roughly five years the world becomes contingent enough that probabilistic prediction of specific outcomes approaches chance.
- **Human judgment and statistical models are complements.** The best forecasting combines them, and it stops where the evidence says predictability runs out.

## What is strategic foresight?

Strategic foresight grew from a different root and toward a different goal: not to predict the future correctly but to prepare for a range of futures under conditions where prediction fails.

### From *La Prospective* to RAND

The modern term traces to France. The philosopher Gaston Berger reintroduced the word *prospective* in a 1957 article, framing it as a forward-looking attitude, distinct from mere forecasting, that looks past short-term trends toward long-term transformation and carries a duty toward future generations.[^5] A second lineage runs through the RAND Corporation after the Second World War, where Herman Kahn's *On Thermonuclear War* (1960) legitimized the systematic exploration of futures most people refused to contemplate, and where Olaf Helmer and Norman Dalkey developed the Delphi method to structure expert judgment under deep uncertainty.

### Shell and the myth that needs correcting

The practice found its corporate form at Royal Dutch Shell. Pierre Wack, working in Shell's planning group from the late 1960s, turned scenarios into a strategic tool, and his two 1985 *Harvard Business Review* articles remain foundational.[^6] Here the canonical story needs care, because it is usually told wrong. The legend says Shell “predicted” the 1973 oil crisis and profited. Michael Jefferson, a former Shell scenario planner, reconstructed the record in a 2012 paper and found the timing does not support the myth: the scenario framework most often cited as the prophetic one was dated May 1973 and drawn from a scenario book published in January 1973, when concern about oil prices was already rising.[^7] Shell's scenarios did not foresee the shock years in advance. They responded to signals as they emerged. Wack himself never claimed prophecy. His actual argument was subtler and more durable: ==scenarios work by shifting the mental models of decision-makers so that when disruption comes, they recognize it faster and act with more coherence than rivals still trapped in old assumptions.==[^6] The value was preparedness, not prediction.

### The conceptual spine

Two further contributions shaped how the field thinks. Roy Amara, at the Institute for the Future, drew a distinction in 1991 that foresight practitioners still use daily.[^8] Futures come in three kinds:

- The **possible** covers what might conceivably happen.
- The **probable** is what is likely on current trends.
- The **preferable** is what we would want.

Forecasting lives almost entirely in the second category. Foresight insists on all three and, in doing so, makes the future a matter of values and choice, not only of probability. Michel Godet then consolidated the French *prospective* into systematic scenario and structural-analysis methods for strategy.

Academic [[Futures Studies|futures studies]] gave the field depth and a critical edge, with scholars like Richard Slaughter pressing foresight to interrogate power and worldview rather than merely extrapolate trends, a stance now gathered under [[Critical Futures Studies|critical futures studies]]. Its methods reach past prediction: [[Sohail Inayatullah]] developed [[Causal Layered Analysis (CLA)|Causal Layered Analysis]], which reads a problem down through four layers, from surface litany to deep myth and metaphor, and recurring typologies of societal futures push planners to weigh collapse and transformation alongside continued growth. None of this produces a forecast. All of it widens the space of futures a decision can be tested against.

Today the field is institutionalized: the OECD, the European Commission (with annual Strategic Foresight Reports since 2020), Singapore's Centre for Strategic Futures, and UNESCO's futures-literacy program all run dedicated foresight work. Foresight is no longer a fringe practice; it has units, budgets, journals, and a professional body.

## Where do forecasting and foresight really differ?

The surface contrasts (quantitative versus qualitative, short versus long) matter less than the deeper split beneath them. The difference is categorical, not gradual: not one practice doing more of what the other does, but two practices making different kinds of claims.

**The first difference is epistemic.** Forecasting treats the future as a largely knowable extension of stable structures, so its central question is “How likely is X?” Foresight treats the future as genuinely open, plural, and partly the product of human choice, so its questions are “What could happen, what do we want, and how should we act given what we cannot know?” This is Amara's triad doing its work: the probable is one future among the possible, and neither exhausts the preferable.

**The second difference is the treatment of uncertainty.** Forecasting works to shrink it, modeling variation as noise around a central estimate. Foresight works to open it, treating the spread between structurally different futures as the signal rather than the noise.

**The third is singular versus plural futures.** A forecast commits to one expected trajectory, however hedged with confidence intervals. Foresight refuses to collapse to one, on the principle that committing early to a single official future is precisely the failure it exists to prevent. This is why scenarios come in sets and why the discipline warns against the [[Official Future|official future]] that quietly crowds out the alternatives.

**The fourth, and the one most often missed, is the criterion of success.** Forecasting is validated by accuracy: proper scoring rules, calibration, and out-of-sample error. Foresight cannot be validated that way, because it does not claim any scenario will come true. Its criterion is decision quality: Did the work expand what leaders could imagine? Did it surface fragile assumptions? Did it produce strategies that hold up across several futures rather than one? A forecast that proves wrong has failed. A scenario that “proves wrong” but changed a decision for the better has done its job. ==Holding foresight to a forecasting scoreboard is the single most common category error in the field==, and it is where a great deal of good foresight gets thrown away. The deepest reason is not about method at all but about what each practice is really talking about, which the final section takes up.

## Is forecasting reliable? Where it wins and where it fails

Reliability is not a property of forecasting as such. It is a property of the match between a method and its domain, and the evidence draws a sharp line.

### Where forecasting wins

Forecasting wins decisively where the underlying system is governed by stable laws, fed by abundant data, and read over short horizons. Numerical weather prediction is the flagship case. Over four decades it has undergone what researchers call a “quiet revolution,” improving through steady, unglamorous gains until a five-day forecast today is about as skillful as a three-day forecast was thirty to forty years ago.[^9] The atmosphere obeys known physics, the observation network is dense, ensembles combine many model runs, and the horizon is bounded. Short-term demand, load, and retail-sales forecasting occupy the same sweet spot: rich data, fast feedback, processes that are complex but not dominated by structural breaks. So does epidemic nowcasting over horizons of a week or two, where multi-model ensembles gave usefully accurate case and hospitalization estimates during the COVID-19 pandemic.

### Where it fails

Forecasting fails, reliably and expensively, where rare events and structural breaks dominate. Crude oil prices are the standard cautionary tale: after forty years of effort, no forecasting model systematically beats a simple no-change assumption beyond a horizon of a few months, because the big moves come from geopolitical crises, technological shifts, and speculative dynamics that historical regularities cannot anticipate.[^10] Recession prediction is nearly as bad. A study of decades of IMF and private-sector forecasts found that economists rarely call a recession in advance; they revise their numbers down only as the bad news arrives, missing the magnitude until the year is almost over, and official and private forecasters were “equally good at missing recessions.”[^11] Long-range technology forecasts fare similarly: a large share of specific predictions fail to materialize, and many of the actual breakthroughs were unforeseen.

| Forecasting earns its keep | Forecasting breaks down |
| --- | --- |
| Numerical weather prediction | Crude oil prices |
| Short-term demand and load | Recession calls |
| Epidemic nowcasting | Long-range technology forecasts |

### The five-year hinge

The hinge between these two regimes is the horizon, and Tetlock's roughly five-year boundary marks it. This is the striking part: the recommendation to switch tools does not come from foresight advocates but from forecasting researchers themselves. Reviewing methods for anticipating rare, high-impact events, Paul Goodwin and George Wright concluded that every forecasting technique exhibits fundamental flaws in that domain and that the productive responses are to make organizations robust and to stimulate structured critical thinking through devil's advocacy and scenario planning.[^12] When predictability runs low and the horizon runs long, the forecasting literature stops recommending better forecasts. It starts recommending foresight.

## Does foresight actually work?

Honesty cuts both ways, and this is the section vendor content usually skips. If forecasting has documented failure modes, so does foresight, and pretending otherwise is how the field loses credibility with the executives it needs to convince.

### The evidence

The strongest quantitative evidence for foresight's value is the 2018 longitudinal study by René Rohrbeck and Menedjit Kum, which measured “corporate foresight maturity” across firms and tracked their performance over several years.[^13] ==Firms with more mature foresight capabilities showed superior profitability and growth, and the authors read this as evidence that foresight builds competitive advantage by helping organizations anticipate change and adapt.== It is the study most often cited to show that foresight “pays off.” It also has real limits, which the authors acknowledge: foresight maturity is measured through self-reported surveys, firms that invest in foresight may differ systematically in ways the study cannot fully control, and performance is shaped by countless external factors. The finding is genuine but it is an association, not a clean causal proof.

Beyond that single study the evidence base is thinner than the field's confidence implies. A systematic review by Jon Iden, Leif Methlie, and Gunnar Christensen found that strategic foresight research is dominated by conceptual work and case studies, with few rigorous outcome studies and little longitudinal testing.[^14] A review of the scenario-planning literature reached the same verdict: it is methodology-heavy and evaluation-light, rich in descriptions of how to build scenarios and poor in validated evidence that they improve decisions.[^14] There is no agreed standard for measuring whether a foresight exercise succeeded, and reported benefits are usually self-reported perceptions of learning rather than demonstrated changes in performance.

### Two failure modes

Two failure modes recur.

- **Scenario theater.** Elaborate scenarios get produced as a workshop or a communication piece and then decoupled from any actual decision, resource allocation, or contingency plan. The scenarios are interesting and strategically useless, and the exercise stops at the story-generation stage.
- **Unfalsifiable vagueness.** This is the failure Tetlock's work exposes. Long-range foresight statements that lack explicit indicators, time frames, and decision hooks slide into vagueness, becoming narratives so unfalsifiable that any outcome can be reconciled with them after the fact, which permits comfortable post-hoc rationalization instead of disciplined learning. Foresight that will not name what would count as being wrong has quietly abandoned the rigor it claims.

None of this makes foresight worthless. It makes foresight a practice with conditions for doing it well:

- Tie scenarios to real decisions.
- Define the indicators you will watch.
- Cultivate fox-like plurality over hedgehog-like grand narratives.
- Be as honest about foresight's limits as a good forecaster is about the limits of a model.

## Better together: how organizations combine both

The mature answer is not to choose. It is to place each practice where its evidence is strong and to connect them, and organizations that do this well follow a recognizable pattern.

### Wind-tunneling

The central bridging method is **[[Windtunneling|wind-tunneling]]**, also called stress-testing. A set of strategies or forecast-based plans is run against each scenario in turn and rated for how well it holds up, so that robust options (those that perform acceptably across several futures) can be told apart from brittle ones (those that win in one future and collapse in the others). New Zealand's Policy Project, the UNDP foresight toolkit, and the UK Government Office for Science Futures Toolkit all institutionalize wind-tunneling as the standard way to link scenarios to strategy. It is where foresight's plural futures meet forecasting's concrete plans.

### From scenario to model

The second bridge runs in the other direction, from scenario to model. Scenario narratives get translated into explicit assumptions and then into numerical drivers a forecast model can use. A “supply-chain shock” story becomes a defined percentage increase in costs and a defined reduction in market size; a downside scenario becomes a specific set of churn and conversion inputs. The scenario supplies the structural assumptions, and the forecast quantifies their consequences. Governments describe the reverse challenge candidly: the hardest part of running foresight inside a planning system is merging quantitative forecasts and qualitative foresight into one coherent process.

### Should scenarios carry probabilities?

A live debate sits inside this integration: should scenarios carry probabilities? The Shell tradition says no. Wack and his successors argued that scenarios are a way to do planning, not forecasting, and that attaching probabilities tempts users to pick the “most likely” one and revert to single-future thinking, defeating the purpose. Much corporate and financial planning practice says yes, assigning subjective probabilities so that scenarios can be weighted into a single expected value for a decision. Both positions are coherent, and the choice depends on whether the aim is to expand imagination or to feed a specific quantitative decision. The point is to make the choice deliberately rather than by default.

Official guidance now codifies the complementarity. The IMF's 2021 practitioner guide, pointedly titled “How to Implement Strategic Foresight (and Why),” tells policymakers plainly that foresight does not attempt to forecast the future, because long-term forecasting is not feasible, and it lays out horizon scanning, scenarios, backcasting, and policy gaming as the toolkit for what forecasting cannot reach.[^15] The same instinct animates work on robust decision-making, which reframes the goal under deep uncertainty as making better decisions rather than better predictions, and tests strategies for robustness across many futures instead of optimizing for one.[^16] Forecasting handles the regular, short-term, data-rich part of the environment. Foresight takes over where structures break, horizons stretch, and the data runs out. Wind-tunneling and scenario-to-model translation are the seams that hold them together.

## A practitioner's view

The category error named earlier, holding foresight to a forecasting scoreboard, is the surface of a deeper point, and it is the right place to end. In my own practice I treat the distinction as categorical, not gradual. Foresight is not forecasting done over a longer horizon or with fuzzier numbers. It is a different kind of claim about a different kind of object, and the cleanest way I know to see this comes from a distinction the sociologist Niklas Luhmann drew in 1976, between **present futures** and **future presents**.[^17]

Two things get casually blurred here:

- **A future present** is a moment that will one day be the present, such as the actual state of the world in 2050.
- **A present future** is the image of that moment as it exists in someone's head right now: our expectations, hopes, and fears about 2050, held today.

The two are not the same thing, and only one of them is real in a way you can study. The future present has not happened. It is inaccessible, not yet a fact, impossible to investigate. The present future exists now, in language and imagination, and it can be examined. This is the ground under [[The Difference between Present Futures and Future Presents|the distinction]] that runs through my work.

A forecast, read this way, makes a claim about a future present: it asserts something about a future state of the world that is by definition unreachable and untestable until it either arrives or does not. Foresight makes no such claim. It works with [[Present futures|present futures]], the images of the future that exist now and steer the decisions people are making now. Those images are the only part of “the future” that is real and available for study, which is why foresight can be rigorous where long-range forecasting cannot.

The twist, and the part I find most clarifying, is that ==a forecast is itself a present future==. As Armin Grunwald has argued, any statement about the future (a forecast, a scenario, a vision) is a construct assembled from present knowledge, ad hoc assumptions, and normative choices about what matters and what to hold constant.[^18] A forecast is not a window onto 2050. It is a present-day artifact, built from today's data and today's assumptions, that happens to hide its normative and speculative components behind the authority of a number. Quantification does not remove the assumptions. It conceals them.

That is why I do not measure foresight by whether it “comes true.” Coming true is a category that belongs to future presents, and future presents are unreachable. I measure foresight by the quality of orientation it gives in the present: whether it makes today's assumptions explicit, whether it widens the set of futures a decision is tested against, or whether it helps people act well under uncertainty they cannot dissolve. None of this diminishes what a good forecast does inside its own short, data-rich domain, where a weather service or a demand planner earns its keep. It only marks the horizon past which that machinery stops describing the world and starts, quietly, describing us. The most useful thing I ever learned to say about a future statement is not “will it happen” but this: tell me what you think about the future, and I will tell you how you feel about the present. Every image of the future, forecast and scenario alike, is finally a statement about now.

## Further reading and sources

A short, opinionated path into the material. Pierre Wack's two 1985 *Harvard Business Review* articles are the primary texts on what scenarios are for, and Michael Jefferson's 2012 reassessment is the essential corrective to the Shell myth.[^6][^7] On forecasting, Philip Tetlock's *Expert Political Judgment* is the honest reckoning with expert prediction, and the Makridakis M-Competitions are the empirical record of what forecasting methods deliver.[^3][^4] Roy Amara's possible-probable-preferable triad is the one-page conceptual key to the whole distinction.[^8] For the evidence on whether foresight works, read Rohrbeck and Kum for the case in favor and the systematic reviews for the honest limits.[^13][^14] For related methods, see [[Scenario Planning]], [[Windtunneling]], [[Causal Layered Analysis (CLA)]], [[Trends]] and [[Megatrend|megatrends]], [[The paradox of foresight]], and the broader field in [[Futures Studies]] and [[Futures Terminology]]. For the governance side, see [[Anticipatory Governance]] and [[Legal Foresight]].

[^1]: Box, G. E. P., & Jenkins, G. M. (1970). *Time Series Analysis: Forecasting and Control*. San Francisco: Holden-Day. Later editions, revised with Reinsel and Ljung, are published by Wiley. ([Wiley](https://onlinelibrary.wiley.com/doi/book/10.1002/9781118619193)) The canonical systematization of ARIMA time-series forecasting.

[^2]: Armstrong, J. S. (ed.) (2001). *Principles of Forecasting: A Handbook for Researchers and Practitioners*. Boston: Kluwer Academic. ([Springer](https://link.springer.com/book/10.1007/978-0-306-47630-3)) The empirical synthesis of forecasting principles, including forecast combination and the case for simple models.

[^3]: Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2020). “The M4 Competition: 100,000 time series and 61 forecasting methods.” *International Journal of Forecasting*, 36(1), 54-74. ([DOI](https://doi.org/10.1016/j.ijforecast.2019.04.014)) Source of the findings that simple methods and combinations are hard to beat and that accuracy decays with horizon. The M-Competitions run from M1 (1982) through M6 (2024).

[^4]: Tetlock, P. E. (2005). *Expert Political Judgment: How Good Is It? How Can We Know?* Princeton: Princeton University Press. ISBN 978-0691128719. The twenty-year study of expert forecasting, the fox/hedgehog distinction, and the roughly five-year horizon beyond which prediction approaches chance. See also Tetlock, P. E., & Gardner, D. (2015). *Superforecasting: The Art and Science of Prediction*. New York: Crown. ISBN 978-0804136716.

[^5]: Berger, G. (1957). Article reintroducing the term *prospective*, published in *La Revue des deux mondes*. On Berger's role in founding French *la prospective*, see Futuribles International's history of the field. ([Futuribles](https://www.futuribles.com/en/la-prospective/histoire-et-memoire-de-la-prospective/histoire-de-la-prospective/))

[^6]: Wack, P. (1985). “Scenarios: Uncharted Waters Ahead.” *Harvard Business Review*, September-October 1985. ([HBR](https://hbr.org/1985/09/scenarios-uncharted-waters-ahead)) And Wack, P. (1985). “Scenarios: Shooting the Rapids.” *Harvard Business Review*, November-December 1985. ([HBR](https://hbr.org/1985/11/scenarios-shooting-the-rapids)) The foundational statement that scenarios shift mental models rather than predict events.

[^7]: Jefferson, M. (2012). “Shell scenarios: What really happened in the 1970s and later.” *Futures*, 44(8), 710-721. ([DOI](https://doi.org/10.1016/j.futures.2012.04.001)) The historical reassessment showing the “prophetic” scenario framework was dated May 1973, drawn from a January 1973 scenario book, and that claims of clear competitive advantage are not robustly supported.

[^8]: Amara, R. (1991). “Views on futures research methodology.” *Futures*, 23(6), 645-664. ([DOI](https://doi.org/10.1016/0016-3287(91)90085-G)) The distinction between possible, probable, and preferable futures.

[^9]: Bauer, P., Thorpe, A., & Brunet, G. (2015). “The quiet revolution of numerical weather prediction.” *Nature*, 525(7567), 47-55. ([DOI](https://doi.org/10.1038/nature14956)) Documents the steady, cumulative gains in forecast skill, including the roughly one-day-per-decade extension of useful horizon.

[^10]: Baumeister, C., & Kilian, L. (2016). “Forty Years of Oil Price Fluctuations: Why the Price of Oil May Still Surprise Us.” *Journal of Economic Perspectives*, 30(1), 139-160. ([DOI](https://doi.org/10.1257/jep.30.1.139)) On the failure of oil-price forecasts to beat a no-change benchmark beyond a few months.

[^11]: An, Z., & Loungani, P. (2018). *How Well Do Economists Forecast Recessions?* IMF Working Paper WP/18/39. ([PDF](https://www.imf.org/-/media/files/publications/wp/2018/wp1839.pdf)) Finds that forecasters miss the magnitude of recessions until the year is almost over, and that official and private forecasters are “equally good at missing recessions.”

[^12]: Goodwin, P., & Wright, G. (2010). “The limits of forecasting methods in anticipating rare events.” *Technological Forecasting and Social Change*, 77(3), 355-368. ([DOI](https://doi.org/10.1016/j.techfore.2009.10.008)) Concludes that all forecasting methods have fundamental flaws for rare events and recommends robustness and scenario planning instead.

[^13]: Rohrbeck, R., & Kum, M. E. (2018). “Corporate foresight and its impact on firm performance: A longitudinal analysis.” *Technological Forecasting and Social Change*, 129, 105-116. ([DOI](https://doi.org/10.1016/j.techfore.2017.12.013)) The strongest quantitative evidence that foresight maturity is associated with superior firm performance; the authors themselves state the caveats on measurement and causality.

[^14]: Iden, J., Methlie, L. B., & Christensen, G. E. (2017). “The nature of strategic foresight research: A systematic literature review.” *Technological Forecasting and Social Change*, 116, 87-97. ([DOI](https://doi.org/10.1016/j.techfore.2016.11.002)) On the dominance of conceptual and case-based work and the scarcity of rigorous outcome studies. See also Amer, M., Daim, T. U., & Jetter, A. (2013). “A review of scenario planning.” *Futures*, 46, 23-40. ([DOI](https://doi.org/10.1016/j.futures.2012.10.003)) on the methodology-heavy, evaluation-light character of the scenario literature.

[^15]: International Monetary Fund (2021). “How to Implement Strategic Foresight (and Why).” *IMF How-To Notes*. ([IMF eLibrary](https://www.elibrary.imf.org/view/journals/061/2021/010/article-A001-en.xml)) Official guidance stating that foresight does not attempt to forecast the future because long-term forecasting is not feasible.

[^16]: Lempert, R. J. (2019). “Robust Decision Making (RDM).” In V. A. W. J. Marchau et al. (eds.), *Decision Making under Deep Uncertainty: From Theory to Practice*, 23-51. Springer. ([DOI](https://doi.org/10.1007/978-3-030-05252-2_2)) Reframes the goal under deep uncertainty as better decisions rather than better predictions.

[^17]: Luhmann, N. (1976). “The Future Cannot Begin: Temporal Structures in Modern Society.” *Social Research*, 43(1), 130-152. The origin of the distinction between present futures and future presents. This foundational essay predates digital object identifiers and is cited here in the standard bibliographic form.

[^18]: Grunwald, A. (2019). “The objects of technology assessment. Hermeneutic extension of technology assessment reflections.” *Journal of Responsible Innovation*, 6(1), 96-101. ([DOI](https://doi.org/10.1080/23299460.2019.1647086)) On futures as present-day constructs of knowledge, assumptions, and normative choices. See also Grunwald, A. (2009). “Wovon ist die Zukunftsforschung eine Wissenschaft?” in *Zukunftsforschung und Zukunftsgestaltung*, Springer, the source of the framing that a forecast is itself a present future.

## Sitemap

- [Notes Index](https://garden.johanneskleske.com/notes-index)
- [Futures Terminology](https://garden.johanneskleske.com/futures-terminology.md)
- [Sitemap](https://garden.johanneskleske.com/sitemap.md)