Making Money Whatever the Weather

Summary

• A new Journal of Banking and Finance paper shows weather can affect global stock markets.
• We can use temperature conditions to predict daily returns in cold countries.
• Combining measures of sunshine, rain, temperature and wind, the authors show how investors can construct weather-based trading strategies that offer up to 15% gross annualised returns.

Introduction

You rarely think of weather as having any impact on the stock market. Yet studies have shown that sunshine influences investors’ propensity to buy stocks, thereby affecting asset prices. The problem, however, is that these studies are quite primitive – they focus mainly on sunshine and fail to consider seasonal weather patterns.

A new Journal of Banking and Finance paper goes further. Alongside sunshine, it asks whether other weather conditions, such as wind, rain, snow and temperature affect markets. They do, and the authors show how a weather-based trading strategy can offer abnormal returns for investors.

Data

Weather data is collected for 1973-2012 from the National Climate Data Centre (NCDC). For each country, the authors collect data from the weather station nearest the main stock market. To determine the weather on any given day, they take the average hourly observations of sky cover (SKC), temperature (TEMP), wind speed (WIND), precipitation (RAIN), and snow depth (SNOW) between 6am and 4pm local time.

Daily index returns are collected for all countries in Datastream’s Global Equity Index. This is 49 countries from 1973-2012. All returns are nominal, in local currency terms and include dividends. Any daily returns above 2.5% are excluded because, according to work cited by the authors, this is the best filter to capture the effects of weather on returns.

Methodology

To determine the profitability of a weather-based trading strategy, the authors divide the sample into two windows: 1973-1992 and 1993-2012.

The strategy is ‘trained’ in the first window using standard OLS regressions to estimate the weather-return relationship. From this relationship, the authors form ‘daily-rebalanced portfolios over one year of the trading period’. In other words, using the information collected from 1973-1992, they create a trading strategy for 1993, then from 1974-1993 for 1994 etc.

To reduce the complexity of the task, the authors group the 49 countries by their average annual temperatures. They create three groups – cold, mild and hot (Table 1). They then estimate the weather-return relationship for each country within a region and aggregate.

For example, how did sunshine, wind, rain, snow and the temperature affect returns in the US every day in June 1990? Similarly, how did weather conditions affect Austria, Belgium, Canada etc. in June 1990? All these countries are in the ‘cold’ group. The authors then aggregate across all 16 countries in this group to get an average weather-returns relationship in June 1990 for all cold-region countries.

They do the same for all mild and hot countries for every month from 1973-1992. I have attached the results in the appendix below to aid interpretation. If they did not do it this way, they would have to create a trading strategy that would differ for all 49 countries. Given that weather conditions can be quite similar across groups of countries, this seems a sensible way to reduce the dimensionality of the trading strategy.

Trading Strategy

The authors create both a hedge strategy and a long-only strategy by temperature regions. It assumes that the weather-returns relationship can accurately forecast and make use of same-day pre-market weather information for all countries in the same temperature region

For the hedge strategy, they go long in the country whose return is predicted to be highest based on the weather conditions from 5-9am on that trading day. And they go short in the country with the lowest predicted return. For the long-only strategy, the authors do the same as the long leg of the hedge strategy. Both portfolios are rebalanced daily.

To evaluate the performance of both strategies, the authors examine the average annualised returns, Sharpe ratio and Sortino ratio. However, the results do not factor in transaction costs.

Finally, the authors also construct a ‘global’ hedge and long-only portfolios for the world region by pooling all countries. They select the long and short positions based on the predicted return for all temperature regions on each day.

Portfolio Results

Below are the paper’s results. The figures exclude transaction costs. But based on the rebalancing rate, the authors can estimate trading costs lower the returns by roughly 2-4pp.

Global Portfolio

The global hedge strategy is highly profitable. It generates a mean annual return of 15.2% between 1993 and 2012, implying weather meaningfully impacts stock returns (Chart 1).

For context, the strategy outperforms a simple buy-and-hold strategy of Datastream’s world index – both the absolute returns (15.2% vs 6.2%) and risk-adjusted returns (Sharpe Ratio of 0.46 vs 0.24) are higher. It also outperforms a US index (CRSP value-weighted index), whose mean average return was 8.8.%.

The global long-only portfolio also does well. It has the potential to be more profitable as it saves on transaction costs. The strategy produces a mean annual return of 13.4% with a Sharpe Ratio of 0.36.

Temperature Region Portfolio

When picking countries based on their temperature grouping, the hedge strategy is most profitable in mild regions. It produces an annual return of 14.5%. Cold regions also do well (13.3%). But in hot countries, annual returns are zero. By time zones, the weather strategy offers the highest realised profits in Europe-African countries. The authors’ rationale is that this region has the toughest weather conditions, meaning weather is likely to have the strongest effect on investor mood.

Similar results hold for the long-only strategy. And, if anything, the results are more statistically significant despite offering slightly lower returns. The long-only strategy in cold countries yields 10.2% mean annual returns and 14.2% in mild countries. The trading strategy still fails in hot regions.

Key Weather Variable

If you could only choose one weather condition on which to build a trading strategy, temperature is the most important one. The authors find that a one-variable hedge strategy using only TEMP generates the highest and most significant profits, but only when focusing on cold countries (16% annual return) or Europe-African countries (10.3% annual return).

Bottom Line

The paper shows how weather-based trading strategies can be profitable even after factoring in transaction costs. It means weather has economically important effects on global stock markets. Perhaps this is because it affects people’s moods or because it can cause disruption. Nonetheless, it provides an alternative dataset that could give insights into future returns.

Appendix

These are the weather-return relationships published by the authors for the three groups of countries.

Interpretation: A negative SKC value in February for cold countries implies that cloudier conditions, on average, lower index returns within the 16 cold countries. In March, windier and snowier conditions have the same effect.