European Summer Temperature Average Correlation Matrix

Weather derivatives are financial instruments that can be used by organizations or individuals as part of a risk management strategy to reduce the risks associated with adverse or unexpected weather conditions. Risk takers are typically large insurance/reinsurance companies and specialized hedge funds. Some banks were also involved in this market until the 2008 financial crisis when the majority of them closed their power trading desks.

At the heart of any climate risk trading strategy for risk takers lies a good diversification of the risk. Risk takers carefully balance their books by looking at the following:

-?????????Geographical diversification: Europe, North America, South America, Australia, Japan…

-?????????Time diversification: months, seasons, years…

-?????????Element diversification: Temperature, Rain, Wind…

-?????????Type of risk: Excess heat, drought, cold spell…

-?????????Market: pure weather, hurricane, energy, wildfire…

Part of the issue that risk takers have to deal with is that the risk they can take from a given market segment is often in the same direction. For example, all European fossil fuel energy companies have a similar risk profile. They may not have the same sensitivity to the risks, but gas demand related to heating purposes in the winter depends mostly on temperature.

Time diversification is a widely used technique because weather variables are mean reverting variables. Hence, if a risk taker absorbs monthly temperature risks laid off by an energy company, it absorbs the volatility of a ‘zero-sum’ game in exchange for a small risk premium. Time diversification is a good risk to take, and everyone is happy. However, this is not always possible. Geographical diversification is the most common way a risk taker can diversify his/her risk with.

So, it is imperative to look at the correlation between various average temperature indices of large European cities when building a portfolio of weather risks. As always, when studying the correlations between weather indices, the quality of the data and the trend-adjusting technique are key. Here, I am using the Speedwell Environmental System to perform this mini-study. The period is June-August. I used our unique database of cleaned weather data. Data is trend-adjusted using a kernel method.

In the matrix above, the first percentage value is the linear correlation. The second value inside the square brackets is the number of data points (years) used to calculate this value.

And as always, data keeps surprising me. Below I have highlighted the correlations that are in excess of 85%:

I knew that Paris was highly correlated with Essen and London. However, I did not know that Bern was so highly correlated with Barcelona!

Before anyone asks, yes the Speedwell Environmental System stress tests the correlation using the Jacknife’s resampling technique and no, the high extreme value (~27 ; ~21.5) has virtually no impact on the value of the correlation.

?

This reminds me of an article I read about Melissa Mayer a few years ago in which she essentially said: “do not think or believe, prove it with good data instead”. This has stayed with me ever since. So, there you go, if you can, do not go long temperature swap contracts for Bern and Barcelona at the same time when building your portfolio. The correlation is higher than for Barcelona vs Madrid. Who would have thought that?

?

Speedwell Climate is the leading provider of data for the climate risk transfer business. We also provide the Speedwell Environmental System for pricing and managing climate derivative contracts. Speedwell Climate also provides independent valuations of weather derivatives, quanto pricing, and portfolios of climate risks.