Valuation of Weather Derivatives 13 June 2001 © 2001 Weather Risk Advisory. All rights reserved Temperature About a year ago, there was a rumour that Coca-Cola the soft drink company, had plans for a vending machine, which varies the price for a drink according to outside temperature. The company denied the plans but we can see the logic behind the idea. One of the main causes of uncertainties in cash flow is the weather. Temperature in this case. In the hot summers of 1976 and 1995, the machine could have charged a fortune A regulated power utility company must fix their price for electricity for a certain period. Prices can only be increased with permission of the regulators. Yet, power consumption will be low in warm winters. There is a source of uncertainty of incoming cash flow: temperature © 2001 Weather Risk Advisory. All rights reserved Rain Agriculture companies have been aware of weather risk for a long time. The major climatic factor which influences crop growth is rainfall “Excess moisture continues to be a problem, particularly to later seed crops.” “… but rainfall over the weekend brought haying to a standstill.” “Excess moisture has taken its toll in some areas as crops in low lying lands have drowned out…” “…heavy rains causing some localized crop damage” © 2001 Weather Risk Advisory. All rights reserved Wind One source of alternative energy is on the rise: Wind farms are being built, not only in this country but also in Europe. Their exposure to lack of wind (and also an excess of wind) is obvious No wind, no electricity, no money © 2001 Weather Risk Advisory. All rights reserved Weather Derivatives © 2001 Weather Risk Advisory. All rights reserved Example One: Description Swap • I shall pay you USD 5,000 for every degree Celsius the average temperature at Heathrow airport is below 18 on any day in the observation period (1 Nov 2001 - 31 Mar 2002) • In exchange, you pay me a fixed amount • Of course, these two payments are netted and the difference paid at the end of the contract period • Netted pay-out is limited to USD 1,000,000 © 2001 Weather Risk Advisory. All rights reserved Example One: Description 2 Option • As above, however, cash flow only occurs when the differential is positive (or negative) • Plus an up-front premium payment • Pay-outs limited to USD 1,000,000 © 2001 Weather Risk Advisory. All rights reserved Example One: Index • Heating Degree Days (HDD) • On each day of the observation period 1. Calculate the number of degrees the average temperature is below 18 degrees (or other reference temperature as defined in contract) 2. Sum them up (and maybe round according to contract details) • Formula HDD 18 Tk n k 1 © 2001 Weather Risk Advisory. All rights reserved Index Calculation: Numerical Example Date 01-Nov-99 02-Nov-99 03-Nov-99 04-Nov-99 05-Nov-99 06-Nov-99 … 29-Mar-00 30-Mar-00 31-Mar-00 © 2001 Weather Risk Advisory. All rights reserved Max T 17.3 13.8 14.6 17.2 21.3 16.6 Min T 12.2 9.9 5.4 11.7 16.3 12.9 7.1 8.2 10 3.7 5 5 Avg T Daily HDD 14.75 3.25 11.85 6.15 10.00 8.00 14.45 3.55 18.80 0.00 14.75 3.25 … 5.40 12.60 6.60 11.40 7.50 10.50 ======= 1664.7 Example One: Trade Swap Calculate Differential = Fixed swap level - HDD index • When differential is positive, receive USD 5,000 * differential • When negative, pay USD 5,000 * differential Put Option Calculate Differential = Fixed swap level - HDD index • When differential is positive, receive USD 5,000 * differential • When negative, no cash flow Up to the value of the limit © 2001 Weather Risk Advisory. All rights reserved Example One: Payout Formula Swap Put Option max min HDD strike, limit ,limit max min HDD strike , limit ,limit © 2001 Weather Risk Advisory. All rights reserved Example One: Underlying • Daily average temperature or • The index (HDD) A model is needed for either the temperature process or the index © 2001 Weather Risk Advisory. All rights reserved Example Two: Description • Pay me USD 500,000 for every day in the observation period (1 Jul 2001 - 31 Aug 2001) on which rainfall is above 0.1 inch at Des Moines, Iowa, International Airport • In exchange for a fixed up-front premium payment (call option) • Pay-out limited to USD 2,500,000 © 2001 Weather Risk Advisory. All rights reserved Example Two: Index • Count the number of days in the observation period on which the rainfall is above the specified threshold © 2001 Weather Risk Advisory. All rights reserved Index Calculation: Numerical Example Trigger Date 1-Jul-00 2-Jul-00 3-Jul-00 4-Jul-00 5-Jul-00 6-Jul-00 7-Jul-00 8-Jul-00 9-Jul-00 10-Jul-00 © 2001 Weather Risk Advisory. All rights reserved 0.10 Rainfall Event Cnt 0 0.00 0 0.00 0 0.00 1 0.74 1 0.41 0 0.00 1 0.61 0 0.07 0 0.00 0 0.00 ======= 3 Example Two: Trade Call Option Call option with strike 0 and tick size USD 5,000 • • Pay an up-front premium payment Calculate Differential = rain index – strike 1. When differential is positive, receive USD 5,000 * differential 2. When differential is zero, no cash flow • Maximum pay-out USD 2,500,000 © 2001 Weather Risk Advisory. All rights reserved Example Two: Underlying • Daily rainfall or • The index Again, a model is needed for either rainfall process or the index © 2001 Weather Risk Advisory. All rights reserved Exotic Structures Further complications and additions • Dual trigger event: temperature and rain • Multi-station trades: baskets or multi-station events • Event only counted when it occurs on two or three consecutive days © 2001 Weather Risk Advisory. All rights reserved Models © 2001 Weather Risk Advisory. All rights reserved Burn Rate Analysis • Clean, reconstruct and de-trend the data • Calculate the index (HDD, etc.) from historical annual observations • Calculate the resulting trade pay-off for every year • Calculate the average of the trade pay-offs • Discount back from settlement date to today • Add risk premium © 2001 Weather Risk Advisory. All rights reserved Data Modification • Clean: fill in missing data and correct errors • Reconstruct: modify data to account for change of equipment, change of location • De-trend: remove trends due to global warming or urbanisation © 2001 Weather Risk Advisory. All rights reserved Burn Rate Analysis : Numerical Example Strike Season 1990/1991 1991/1992 1992/1993 1993/1994 1994/1995 1995/1996 1996/1997 1997/1998 1998/1999 1999/2000 Mean © 2001 Weather Risk Advisory. All rights reserved 1700 Index Swap Option 1846 -146 0 1781 -81 0 1735 -35 0 1766 -66 0 1572 128 128 1948 -248 0 1807 -107 0 1538 162 162 1675 25 25 1665 35 35 ======================= 1733.3 -33.3 35 Burn Rate Analysis : Pros and Cons Positive + Simple to implement + Easy to understand + Suitable for portfolio valuation Negative - Low probability events ignored or unreliably estimated - General problem is that in such a low sample, a single observation can strongly influence the option value - Unclear selection and calculation of risk parameters © 2001 Weather Risk Advisory. All rights reserved Stochastic Models for Weather Paths • Choose a stochastic model for daily observations • Estimate the model parameters using weather data • Simulate a weather path • Calculate index and option value for this path (=simulated observation) • Repeat many times and calculate mean • Discount back to today © 2001 Weather Risk Advisory. All rights reserved Stochastic Models: Examples • Auto-Regression (AR): use the weighted sum of previous days’ temperatures for the estimation of next day’s temperature • Mean-reverting diffusion (MR): a Markov model with pull-back to the mean • Cao-Wei: a variation of AR • Bob Dischel: a variation of MR © 2001 Weather Risk Advisory. All rights reserved Estimators • Moment estimator find parameters based on observed and model moments, e.g. mean and standard deviation • Maximum-likelihood estimator find parameters which are most likely, based on observed data © 2001 Weather Risk Advisory. All rights reserved Example: Auto-regression Equation T n n 1T n 1 2T n 2 n © 2001 Weather Risk Advisory. All rights reserved Example: Mean-reverting Diffusion • Equation dT t a t T t dt dW t or • alternatively dT t t dt at T t dt dW t © 2001 Weather Risk Advisory. All rights reserved Stochastic Models: Pros and Cons Positive + One (model) fits all (indexes) Positive/negative ± Portfolio risk management possible but may have associated problems (multi-location correlation) Negative - Slow or inaccurate (in particular for low-event contracts) - Harder to implement © 2001 Weather Risk Advisory. All rights reserved Weather Forecasts © 2001 Weather Risk Advisory. All rights reserved Long-range Weather Forecasts • So far, the approach has been to base an estimation of the future on past observations • With weather, can we do better and use forecasts? • And how could we integrate them into pricing models? © 2001 Weather Risk Advisory. All rights reserved Ensemble Forecasting - Probability Forecasts • Ensemble forecasting has been developed over the past decade and has lead to a large increase in the accuracy and usability of medium-range forecasts • The technique is now being applied to longer range forecasts • Ensemble forecasting uses the sensitivity of forecasts to the initial conditions to provide probability forecasts Ensemble Forecasting By perturbating the initial conditions of a forecast and rerunning it multiple times, a range of forecasts for the same time period are produced. These can be used to produce probability distributions of possible outcomes. © 2001 Weather Risk Advisory. All rights reserved Climate Prediction Center Ensemble Forecasts • They have been found to be reliable in El Nino and La Nina years for predicting the US climate, but have a tendency to be too cold • CPC predicts whether temperature and rainfall will be above, below or near normal for 30 and 90 day periods • But El Nino does not effect Europe so this approach cannot be used © 2001 Weather Risk Advisory. All rights reserved Forecast Format • Forecasts are expressed as the probability anomaly of the observation • Three classes - above - near - or below normal • Forecast probability anomaly is the difference between the actual forecast probability of the verifying observation falling in a given category and its climatological value of 33.3% • Mean temperature and total precipitation © 2001 Weather Risk Advisory. All rights reserved Forecast Format: Example CL A5 A 10 B5 B 10 Below Near Above 33.3% 33.3% 33.3% 28.3% 33.3% 38.3% 23.3% 33.3% 43.3% 38.3% 33.3% 28.3% 43.3% 33.3% 23.3% © 2001 Weather Risk Advisory. All rights reserved Integration of Forecasts into Pricing Models • Adjust model parameters so that model probability distribution matches forecast • Read our paper for more details © 2001 Weather Risk Advisory. All rights reserved Summary and Conclusions © 2001 Weather Risk Advisory. All rights reserved Theoretical Considerations of Modelling Approaches • Arbitrage-free pricing approach (risk-neutral valuation) as used in standard derivatives industry price = risk-neutral value (+profit margin) • Actuarial approach price = statistical value + risk premium (+ profit margin) Actuarial approach currently used in weather derivatives industry and insurance • Shadow-price approach © 2001 Weather Risk Advisory. All rights reserved Summary • A market for weather derivatives is being established • Typical examples of currently traded products • Untradable weather risk will probably remain, in particular extremes. They will remain insurance cases • Two modelling approaches, both have their strengths and weaknesses • Forecasts • Actuarial approach © 2001 Weather Risk Advisory. All rights reserved Two Challenges Ahead • Transition to arbitrage-free approach (when markets become liquid, take future contracts and back out a risk-free temperature curve, take options and work out a volatility surface) • Portfolio management of mixed books (e.g. power and weather), to give you an idea of that challenge, even portfolio management of weather books is not fully solved © 2001 Weather Risk Advisory. All rights reserved References © 2001 Weather Risk Advisory. All rights reserved References • M. Cao and Wei, J.: Equilibrium Valuation of Weather Derivatives, Working Paper. • B. Dischel: At Last: A Model for Weather Risk, EPRM Mar 1999. • G. Considine: Introduction to Weather Derivatives, Working Paper. • I. Nelken: Weather Derivatives – Pricing and Hedging, Working Paper. • L. Zeng: Weather Derivatives and Weather Insurance: Bulletin of the American Meteorological Society, Vol. 81, No.9, Sep 2000. © 2001 Weather Risk Advisory. All rights reserved www.WeatherRiskAdvisory.com info@WeatherRiskAdvisory.com tel: +44 (0) 1954 206246 fax: +44 (0) 1954 206250 © 2001 Weather Risk Advisory. All rights reserved