Cost witout benefit

In 1970, the economist George Akerlof published what would become a classic paper called, “The Market for Lemons: Quality Uncertainty and Market Mechanisms.” The “lemons” he was referring to were low-quality used cars. His insight was that, in markets where buyers cannot evaluate products, they will not pay a premium for purported high quality. This, in turn, will give sellers little incentive to provide high quality products. Consequently, the average quality of products in the market declines, further reducing the amount buyers are willing to pay. In economics jargon an “information asymmetry” between buyers and sellers results in “adverse selection”. The process identified by Akerlof is now a textbook example of market failure.

A buyer of short-range weather forecasts does not need to know about meteorology to evaluate them.

The used car market is not the only one susceptible to the “lemon” problem. (Ironically, in the market for actual lemons, buyers can easily assess freshness and taste). A rapidly growing market that is also vulnerable is the one for information about long-range climate risks. A buyer of short-range weather forecasts does not need to know about meteorology to evaluate them. They just need to know some statistics and be able to collect a sufficiently large sample of forecasts and subsequent outcomes to estimate the skill of the provider. For forecasts 1–2 days ahead, this sample might be obtained in a matter of months. However, companies increasingly need forecasts of climate-related risks on horizons of years, and even decades. Obviously, on these time scales, waiting to see how forecasts turn out before choosing a provider is not practical. Some providers want to keep their models and methods proprietary, but even with full transparency it can be hard for non-specialists to judge their quality. When a market has these characteristics, buyers need to be alert to the danger that bad products will crowd out good ones.

There are theoretical reasons why the quality of climate information may deteriorate, but is there evidence of this happening in practice? In 2019, an article in Science suggested that some climate information products might be “…operating outside of the bounds of scientific merit” and, if you work in the field of climate risk, you have probably heard stories of providers pushing precision and granularity beyond what is scientifically plausible. But is there systematic evidence that some climate information has a lemon flavour?

Shortcomings in long-range climate-related predictions might not become apparent for years but there are already indications of problems. Hain et al. recently looked at six different metrics of firm-level exposure to physical climate risks. Three of the measures, from TruCost (part of S&P), Carbon4 Finance and South Pole, were based on the output of climate models, while the other three, from Truvalue Labs (part of FactSet) and two academic groups were constructed by analysing the content of social media, earnings calls and 10-K filings. The correlations between the metrics were generally low, both overall and within specific sectors. Indeed, after allowing for the multiplicity of correlations tested, there were arguably no significant correlations between different metrics. Think about this for a moment: six metrics all claiming to measure the same thing, or at least to be proxies for the same thing, yet they don’t correlate with each other. Even the most generous interpretation implies that most of the metrics contain little information about the common factor they claim to measure. Perhaps one of the metrics is a good measure, but which one? Perhaps none of the metrics are informative. Consider an analogy in which there are six different blood tests to measure the same substance, but when hundreds of subjects are given all six tests there is no correlation between the results. Would a doctor base clinical decisions on any of these tests? Would a medical regulator allow such a situation?

By the time the accuracy becomes apparent all those involved are likely to have moved on.

The problem of information asymmetry between buyers and sellers might be compounded when the buyers of climate information are not concerned with the accuracy of the information either. Much of the demand for forward-looking climate information is driven by regulators and shareholders expecting firms to disclose their climate-related risks. The exercise is one of “putting numbers in boxes”, and by the time the accuracy of the numbers becomes apparent, all those involved are likely to have moved on. This is not conducive to accurate disclosure.

At this point, you might protest that there are many people involved in climate-risk who want to provide accurate and rigorous information that is supported by science, and there are many people who want such information to make accurate disclosures of risk to shareholders. This is true, but it misses the point: the problems in the market for climate-risk information are structural. The used-car market did not become a market for lemons because it was populated by untrustworthy salesmen; it attracted and created the stereotype of the used-car salesman because of its intrinsic information asymmetries.

If information asymmetry and the potential for adverse selection in the market for forward-looking climate information is not addressed, it is likely that more evidence will accumulate that undermines the credibility of climate-related risk disclosures. This will make mandatory disclosures a cost to businesses without any offsetting benefit to society.

This article originally appeared on Medium as How climate-related risk disclosures could become a cost without a benefit