Secular vs Cyclical: Why it’s hard to tell in real time, and what the EV write-down wave reveals

In February 2026, Stellantis disclosed charges of about €22.2 billion (roughly $26.5 billion) tied to a strategic “reset,” after the company concluded it had overestimated the near-term pace of Battery-Electric Vehicle (BEV) adoption and needed to realign its product and supply chain plans. The company characterized the move as expanding “freedom of choice” across powertrains (EVs, hybrids, and advanced internal combustion engines). It said the charges included roughly €6.5 billion of cash payments expected over the next four years.

In its own breakdown, Stellantis said roughly €14.7 billion related to re-aligning product plans with customer preferences and new U.S. emissions regulations, explicitly “reflecting significantly reduced expectations for BEV products”, including write-offs for cancelled products and platform impairments, plus projected cash payments connected to both cancelled and “other ongoing” BEV products that were now expected to run far below earlier projections. It also recorded about €2.1 billion tied to resizing the EV supply chain (including steps to rationalize battery manufacturing capacity), alongside additional operational changes.

The market reaction was immediate and severe: Stellantis shares fell as much as ~24% in Milan on the day and were down by a similar magnitude in Paris, an episode Reuters characterized as potentially the company’s most significant one-day drop on record, with trading halts triggered early in the session. In U.S. trading, reporting widely described a roughly mid‑20% decline in the U.S.-listed shares as well.

The “strategic reset” also had a concrete industrial footprint. Stellantis agreed to sell its 49% stake in the NextStar Energy battery joint venture in Ontario, a plant across the border from Detroit, to its partner LG Energy Solution for a nominal $100, while remaining a customer for the facility’s output.

This is not just one company’s self-inflicted wound.

Reuters estimated that global automakers booked roughly $55 billion in write-downs over the prior year as they scaled back EV ambitions, pointing to a familiar pattern: demand, policy, and competitive dynamics moved differently than many corporate plans implicitly assumed.

The same Reuters roundup highlighted comparable retrenchments: Ford announced a $19.5 billion write-down in December 2025 tied to a pivot away from specific EV programs; General Motors disclosed a $6 billion charge in January 2026 (after an earlier $1.6 billion EV-related charge in October 2025); and Volkswagen flagged a €5.1 billion (~$6 billion) hit in 2025 associated with Porsche delaying parts of its EV rollout in favor of hybrids and internal combustion engines.

In a widely circulated framing, Haig Partners managing director John Murphy called the industry’s aggressive EV push “the single biggest capital allocation mistake in the history of the automotive industry” and suggested EV write-downs could reach at least $100 billion.

Whether one endorses that rhetoric or not, the underlying puzzle is real: when you are living through a shift, how do you tell whether the data are showing a durable, structural (secular) re-pricing of the future, or a transient (cyclical) air pocket that will reverse?

What “secular” and “cyclical” mean in markets, and why the definitions don’t solve the problem

The inference problem is not a niche academic detail. It is central to any founder deciding which tailwind to ride, to investors determining their strategy, and to operators making investment decisions.

Classic time-series work decomposes observed series into components meant to represent “permanent” (trend) and “transitory” (cycle) movements. Still, the resulting decomposition depends on the model used and on the assumptions made about persistence, shocks, and the forecasting structure. Even when you adopt a well-known decomposition method (e.g., the permanent/transitory decomposition formalized in the early 1980s), the practical meaning of “permanent” is tied to statistical assumptions about the data-generating process.

The reality is quite simple: “secular” and “cyclical” are not labels that nature stamps on a data printout. They are conclusions drawn from (often implicit) models. In a rapidly changing industry like autos, where technology, regulation, consumer preferences, and financing conditions all move at once, reasonable models can yield very different answers from the same short-run evidence.

Separating trend from cycle is a ‘filtering’ exercise that is notoriously unstable at the end of a data sample, precisely when decisions are made. Furthermore, cyclical shocks often ‘bleed’ into the trend through hysteresis, in which a temporary downturn (such as high interest rates) causes permanent structural shifts in a company’s capacity and labor force.

There is also a pure identification problem: Some trend-cycle decompositions mechanically imply correlations between trend and cycle innovations, and research has shown that distinguishing whether “trend shocks enter the cycle” or “cyclical shocks enter the trend” can require additional identifying restrictions; reduced-form methods may fit both stories and cannot tell you which is the “true” one.

In other words, even if you accept a trend-cycle framework, the causal interpretation is not automatic.

So, in real time, you face a three-layer uncertainty:

1) the data (which get revised and are noisy at turning points),

2) the statistical separation of trend and cycle (which is model-dependent and unstable at endpoints), and

3) the economic mapping from “trend/cycle” to structural reality (which can be non-unique).

The EV transition as a case study in overlapping secular forces and cyclical headwinds

The EV transition is a particularly instructive case because it contains both strong secular drivers and substantial cyclical drag.

Secularly, EVs are associated with long-run trajectories of technology diffusion and costs. Adoption of innovations often follows an S-shaped path: slow early uptake, a faster “takeoff” phase as costs fall and infrastructure improves, and then eventual saturation. The empirical diffusion model introduced in Frank Bass’s classic work formalizes how “innovation” and “imitation” processes can generate exactly this kind of nonlinear adoption curve.

Cost dynamics provide another secular tailwind: decades of evidence suggest that lithium-ion battery prices have fallen along learning curves, where costs decline as cumulative production increases. Syntheses of battery learning rates commonly place the decline at ~19% per doubling of cumulative deployment over long horizons.

At the global level, the International Energy Agency reports rapid expansion in electric-car sales in recent years; its 2025 outlook noted that global electric-car sales exceeded 17 million in 2024 and reached a sales share above 20%, with significant geographic variation (China far ahead of most markets).

Cyclically (and institutionally), however, EV demand is unusually sensitive to macro-financial and policy conditions because EVs often carry higher upfront prices and because the economics for many buyers depend on financing terms, incentives, and charging convenience.

By late 2025 and early 2026, multiple automakers explicitly linked EV pullbacks to weakening demand and to shifting U.S. policy support. For example, Reuters reported that Ford cited market changes and policies that reduced federal support for EVs when it announced its $19.5 billion EV-related write-down and program cancellations, noting that U.S. EV sales had fallen sharply after the expiration of a long-standing $7,500 consumer tax credit in late September 2025.

Likewise, General Motors’s $1.6 billion write-down in October 2025 as it reshaped EV plans after policy changes, and a later $6 billion write-down to unwind certain EV investments, with the company expecting additional, but smaller, charges and warning of slower EV adoption.

The Stellantis episode sits inside the same mosaic. Reuters’ article above mentioned that Stellantis’ CEO described the writedowns as reflecting “the cost of over-estimating the pace of the energy transition,” and linked the broader industry pullback to the more challenging U.S. market environment and competitive pressures, including from China. The company’s own release emphasized realigning product plans with customer preferences and U.S. emissions regulations, which implies not only a change in demand forecasts but also a shift in the firm’s view of how regulation will bind across its fleet and time horizon.

What makes “secular vs cyclical” especially treacherous is that different geographies can be in various phases of the same diffusion process at the same time.

In Norway, EV adoption is so advanced that the market is effectively near saturation: Reuters reported that in 2025, 95.9% of new car registrations were fully electric (with nearly 98% in December), driven by a long-standing mix of incentives for EVs and penalties for combustion vehicles. Yet Reuters also reported that, more broadly across Europe, the EV market was still in a competitive, policy-sensitive phase; for example, fully electric vehicles accounted for 22.6% of EU registrations in December 2025, roughly tied with petrol, and hybrids were the largest category.

Meanwhile, Reuters reported that Stellantis pegged EVs at 7.7% of U.S. new-car sales and 19.5% of European new-car sales at the time, underscoring how far the U.S. was from an “inevitable near-term” transition path.

A practical implication follows: the EV transition can be secular in direction (long-run share rising, battery costs falling, regulations tightening over decades) but cyclical in speed (purchase timing, mix shifts between hybrids/EVs, pricing pressure), with policy acting as a “regime switch” that can re-time adoption.

That combination makes it easy to misread a temporary deceleration in one geography as evidence that the entire secular shift has stalled, or, conversely, to extrapolate early-adopter acceleration into an overly steep global trajectory.

In fact, we treat the slowdown as a mix of macro-financial ‘noise’ and policy shifts. However, for Western OEMs like Stellantis and Ford, the threat is not just a cyclical air pocket; it is a secular loss of comparative advantage.

While U.S. and EU legacy makers are writing down assets, Chinese competitors (e.g., BYD, Xiaomi) are scaling. This suggests that for Stellantis, the ‘reset’ isn’t just about waiting for interest rates to drop; it is an admission that their first-generation EV platforms are secularly uncompetitive on a cost-per-range basis.

Thus, we must distinguish between a Market Slowdown (cyclical) and a Market Takeover (structural).

How executives and investors get fooled: incentives, accounting, and irreversibility

The secular/cyclical confusion is not just about statistics; it is also about how organizations encode beliefs into irreversible commitments.

A useful lens is the economics of irreversible investment under uncertainty. Robert Pindyck and Avinash Dixit emphasize that when investment is mainly irreversible, and future demand/costs are uncertain, the option to wait has value; committing capital too early can destroy that option and may rationally require higher investment thresholds than a simple net-present-value rule would suggest.

In corporate finance, Stewart Myers similarly framed many growth opportunities as real options, whose value depends on the ability to make (or avoid) follow-on investments as uncertainty resolves.

Translate that to EVs: platform bets, battery joint ventures, supplier tooling, and factory conversions are quintessentially lumpy and partially irreversible. If leaders treat the EV ramp as an imminent secular certainty, they will rationally compress timelines, invest at scale, and sign long-dated commitments (to secure capacity and first-mover advantage). If the ramp then turns out to be slower, whether due to cyclical financing conditions, policy changes, or consumer-preference frictions, the same commitments become stranded, and the option value of having waited becomes visible only in hindsight.

Accounting then forces the economic reassessment into a discrete number, a write-down, often at precisely the moment when the cycle is least forgiving. Under IAS 36, the impairment standard used in IFRS reporting, recoverable value is assessed as the higher of fair value less costs of disposal and “value in use,” which is explicitly based on expected future cash flows discounted using a rate reflecting market assessments of the time value of money and risk.

That structure creates a channel through which “cyclical” changes, such as higher discount rates or a deteriorating near-term cash-flow outlook, can trigger impairments even if the long-run secular story remains unchanged. (Advisory interpretations of IAS 36 explicitly flag rising market interest rates as a potential external impairment indicator because they raise discount rates and can reduce value-in-use calculations.)

Stellantis’ own disclosure demonstrates this interaction: it described both non-cash elements (write-offs, impairments) and a substantial cash component (projected payments over four years), reflecting both valuation reassessments and absolute contractual obligations from cancellations and supply-chain resizing.

This is one reason write-downs are an unusually harsh scorekeeper for “trend mistakes”: they convert a probabilistic forecasting error into an accounting event, concentrated in time and often occurring during a weak part of the business cycle.

The institutional incentives are equally important.

Large organizations tend to coordinate around narratives that enable capital allocation, especially when the alternative is to admit, “We don’t know the slope of the curve.” Once a narrative becomes consensus (“EVs are an unstoppable near-term trajectory”), it can drive correlated decisions by boards, suppliers, and governments, creating feedback loops (capacity races, subsidy commitments, and supply-chain build-outs) that appear to confirm the narrative. The result is a classic real-time identification trap: behavior that is rational conditional on a belief can itself generate data that seem to validate the belief, until the system hits a constraint (consumer affordability, charging bottlenecks, policy shift, or competitive price shock).

Beyond statistical filtering, executives fell victim to Narrative Reflexivity. Between 2021 and 2023, equity markets applied a ‘Tesla Multiplier’ to any legacy OEM announcing aggressive EV targets, while penalizing those preaching ‘optionality.’ This created a feedback loop: high stock valuations (driven by the narrative) lowered the cost of capital for EV projects, which incentivized boards to greenlight billions in CAPEX to maintain those valuations. The write-downs of 2025–2026 mark the ‘Correction Phase’ of this reflexivity, in which the market finally decoupled the narrative from the unit economics.

A rigorous toolkit for distinguishing secular trends from cycles without pretending certainty

The $100 billion error was fundamentally a Linearity Bias. Planners took the ‘Innovation’ phase of the Bass Diffusion Model (the steep upward slope). They projected it as a straight line, failing to account for the ‘Chasm’, the point where early adopters are exhausted, and the mass market requires infrastructure parity that didn’t yet exist.

Because trend-cycle separation is inherently uncertain, “rigor” is less about finding a single magic statistic and more about disciplined triangulation and decision design that is robust to being wrong. The most rigorous approach treats “secular vs cyclical” as a hypothesis-testing and model-comparison problem, with explicit attention to endpoint uncertainty.

A methodological baseline is to treat trend changes as potential structural breaks and cycles as regime-dependent deviations. So the question becomes: what can you do that still works when your trend estimate is fragile?

A practical, high-rigor checklist looks like this:

First, separate “direction” from “speed.” In EVs, the long-run direction (higher electrification share over decades) can remain intact while the near-term pace slows, producing enormous capital allocation errors if the pace assumption is embedded in capacity plans. This is precisely what Stellantis’ charge description implies: expectations for BEV volumes and profitability were revised down enough to impair platforms and cancel products, even as the company still framed its strategy as spanning EVs, hybrids, and advanced ICE (Internal Combustion Engine, not the other ICE).

Second, triangulate with cross-sectional evidence rather than relying on a single aggregate time series. If a slowdown is purely cyclical, you often see it broadly across discretionary categories and geographies with similar financing exposure. If it is structural, you see persistent share shifts, cost re-ranking, or durable consumer-preference changes in specific segments. The EV landscape provides a natural experiment: Norway’s near-saturation of EV share, the EU’s hybrid-heavy mix with rising BEV shares, and the U.S.’s lower penetration create a cross-sectional map that is hard to reconcile with any single “one-size-fits-all” story.

Third, model policies as regimes, not as noise. The Reuters reporting around Ford, GM, LG battery investments, and Stellantis consistently treats U.S. EV incentives and regulatory posture as first-order demand determinants. Whether you view that as desirable or not, it is analytically decisive: a policy regime shift can mimic a cyclical downdraft in the short run, while actually changing long-run expectations about adoption timing and profitability.

Fourth, focus on accounting triggers as “stress tests,” not as trend declarations. Impairments and write-downs are signals that expected discounted cash flows have been revised downward, but they are not, by themselves, proof that the secular story is false. Under IAS 36 logic, a mix of lower near-term margins and higher discount rates can force impairments even if long-run volumes recover. The right inference is narrower: the market is now paying to unwind commitments made under an adoption-speed assumption steeper than current conditions justify.

Finally, redesign decisions to be robust to misclassification by using “optionality” in capital allocation. The real-options literature argues that, under uncertainty and irreversibility, staged commitments and flexibility have value because they preserve the ability to adapt as information arrives. Concretely, for an EV transition, that means emphasizing flexible manufacturing, modular platforms, and contracts that avoid one-way lock-in, so the firm can pivot between hybrids/EVs as the cycle and policy regime evolve, rather than encoding a single trajectory into the cost base.

Final Words

The more interesting lesson of the February 2026 write-down wave is not “EVs were a mirage” or “EVs are inevitable.”

It is that the most complex and most expensive errors happen when companies confuse a secular direction with cyclical speed, and then make irreversible investments whose payoff requires the steep version of the story.

The rigors of trend-cycle econometrics, the history of real-time output-gap mismeasurement, and the option value of waiting all point to the same conclusion: you usually cannot know the answer with confidence at the turning point, so the real test of rigor is how well your strategy survives being wrong.