Butterfly Effect and Forecasts

  John Eichberger |  July 20, 2022

A friend of mine who worked for a market analysis and consulting firm at the time, and who personally produced many forecasts about the transportation market (including some for the Fuels Institute), once told me – a forecast is only as good as the assumptions used and the results are outdated as soon as they are released. This is because there are so many factors that can influence the evolution of the market that it is impossible to account for all of them and it is equally impossible to factor in potential changes in each.

From this I conclude that forecasts are subject to the butterfly effect – “small changes in initial conditions can lead to large-scale and unpredictable variation in the future state of the system” – and therefore must be viewed with interest but with caution, especially when seeking to build strategies around such data. Nowhere in the transportation industry is this more relevant today than with regards to forecasts concerning the growth of the electric vehicle (EV) market.

For example, I recently have seen projections from highly regarded global consulting firms who conclude that EVs will represent 50% of light-duty vehicle sales in the U.S. by 2030. Another analysis said the U.S. has recently passed the tipping in EV sales – which they define as account for 5% of sales. Such projections and analyses from these firms carry significant weight – they form the basis of news stories, influence investment decisions and often contribute to the development of government policies. Yet how much can we rely on these forecasts and how should we effectively use them to evaluate the direction of the market?

Former Secretary of Defense Don Rumsfeld said famously:

“There are known knowns. These are things we know that we know. 
There are known unknowns. That is to say, there are things that we know we don't know. 
But there are also unknown unknowns. There are things we don't know we don't know.”
 

Known Knowns
I assume some of these highly regarded forecasts are largely predicated on the stated government policies and priorities and publicly announced production intentions of global automakers, some of whom have pledged to cease production of combustion vehicles by a certain date in the future. Many OEMs are investing billions of dollars in EV and battery production facilities. Governments throughout the world and within the states are contemplating a future market in which new combustion engine sales will be prohibited. These are “known knowns” and, against this backdrop, it is understandable that observers would accept these forecasts as indisputable fact. But remember what my “forecast for a living” friend told me about forecasts – we must ask what assumptions are being incorporated, which elements are not being factored into the model and how uncertainties are being addressed, especially if we are going to use this information to inform our plans for the future.

Known Unknowns
Do these forecasts take into consideration “known unknowns?” These might include developments we know might happen but we do not know if, when or how significant they might be should they occur. For example, the impact of inflation on market development; the impact of recent price increases for leading EV models; the extent of a possible recession and its impact on consumers’ ability to afford a new vehicle; implications of a prolonged supply chain disruption or microchip shortage that has already hindered vehicle production and availability; global energy disruptions that have elevated consumer prices and threaten energy shortages in key markets throughout the world; and a whole host of other issues.

For example, according to a July article in the Wall Street Journal, EV market leader Tesla has experienced some significant challenges recently and the author noted that some of these challenges “have been largely out of the company’s control.” This is the key point – so many variables that will influence the development of a market are beyond the control of individual entities and impossible to incorporate accurately into forecast models. Forecast models should be qualified and include reference to or some method for factoring such known unknowns to account for a degree of uncertainty.

Unknown Unknowns
Because we don’t know what we don’t know, we cannot predict these. What we can assume with certainty is that there will always be unexpected market developments that will plague the accuracy of forecasts. We cannot account for these developments, but we must acknowledge them and realize that when they do happen they can significantly change the trajectory of the market. Consequently, forecasts and projections must be considered in the appropriate light – they provide great insights into possible future market scenarios, but they must be handicapped by the reader to account for the unknown unkowns.

A Better Way
Market analysts who claim to “know” the future and issue forecasts in such a way to infer that their model’s results are a veritable certainty are doing the market no favors. They are influencing investments and policies based upon the assumptions and variables they have included in their model, and there is no possibility they have captured every potential flap of a butterfly’s wings. Such “results” are intended to draw attention and make headlines, and that is worrisome.

For example, several years ago a “thought leader” claimed that within 10 years of autonomous vehicles being approved for U.S. travel, 95% of vehicle miles traveled would be through electric, autonomous ride hailing services and that personally owned vehicles would be a rarity. This captured a lot of headlines, but was the forecast truly reflective of potential market conditions and did it take into consideration uncertainty? Finally, was it reviewed by anyone outside of the firm that published it to comment on the hypotheses, inputs, methodologies and analyses?

Rather than publishing reports designed to shock and awe and drive headlines, it would be much more reasonable, responsible and reliable to run several models with different assumptions to create a range of potential market development scenarios based upon those variabilities. In addition, researchers should clarify that their results are based upon certain assumptions and explain how variability within those assumptions, uncertainty about future developments and influential variables outside of their model could affect the results. And finally they should seek the additional perspective of diverse experts through a collaborative peer review process to ensure that the assumptions and conclusions are reasonable, objective and based upon facts. “Definitive” studies that have not been reviewed by anyone but the authors should be viewed with additional caution.

Additional Insights on the EV Market
The Fuels Institute this summer will publish a new report, written by IHS Market (now a part of S&P Global), seeking to better understand how many EV charging stations we might need, when and where we might need them and what type of chargers we might need in different use case scenarios. To reach these results, they had to factor in the expected size of the EV market over time. Therefore, within this report are their own forecasts for how the EV market may develop. Contrary to recent published forecasts, their projection is much more conservative and estimates that by 2030 only 17% of light duty vehicles sold will be electric as will 6% of vehicles on the road. (The forecasts for this study were developed in mid-2021, so we must acknowledge that market developments since then were not incorporated into the model.) This is not to say these outcomes are guaranteed, but it served as a baseline against which they evaluated the needs for EV charging stations and it serves as an example of the variability that exists within the art of modeling.

We published a paper in January comparing the life cycle carbon emissions of EVs, hybrids and internal combustion engine vehicles. We released the report with a focus on a model utilized in the report that assumed the carbon intensity of the average U.S. utility grid. A friend quickly challenged me, explaining there is no “average” in the U.S. utility mix. (This is a good time to point out a very important observation – we must not rely upon headlines, infographics or executive summaries to provide us with the critical insights that underly forecasts and projections. By nature, these are designed to deliver top-line results – but for us to truly understand how those results were achieved, and more importantly how they pertain to the development of the market, we have to look at the methodology that should be described in the actual reports.)

I responded to my friend that we had to establish a benchmark against which to compare variability, and then turned his attention to other sections of the report where we adjusted the carbon intensity of electricity generation to see what effect that might have on life cycle results. We also adjusted the model for more than 10 other variables to see how each independently could affect the results. We did not account for every potential variation nor did we run a complex series of combinations – that would have been unwieldy and cost prohibitive, not to mention impossible to avoid missing something. But we sought to provide a series of data points to further the discussion along – not to publish “the” answer.

So, with all that said - what might the next ten years hold for the electrification of the transportation sector? Maybe we will see EVs represent somewhere between 17% - 50% of sales by 2030. Maybe the floor will drop out of the market and EVs won’t achieve even the 17% forecast or maybe we will experience a breakthrough in technology that will create such a compelling consumer value that sales will eclipse 50%. The bottom line is we simply don’t “know,” and these examples only pertain to the vehicles market itself and do not even reference the general state of the global economy. Therefore, we must prepare for a variety of future market scenarios.

If the objective is to reduce carbon emissions, we must prepare to offset the potential impact of a swarm of butterflies throwing the market in multiple directions. To prepare we should support new technologies, improve existing ones, and look up and down the supply chain and throughout the life cycle of each to identify opportunities to reduce emissions. There is no one potential outcome and there is no one silver bullet solution. As we seek to understand the market, it is important that we rely upon multiple resources and perspectives, ask critical questions and avoid being trapped in a spider’s web right next to those pesky butterflies.