Just an idea, of course.
But it comes via JPMorgan’s Seamus Mac Gorain, and it’s part of a 12-page study covering more than a decade of intraday trading data around big economic releases.
Here’s his thinking:
The unusually severe weather in Europe and North America this winter contributed to some unexpectedly weak economic data releases, of which the most striking was perhaps the 4%ar downside surprise in UK GDP. This begs the question of whether there is a systematic and exploitable relationship between the weather and economic surprises.
To explore this question, we construct a simple measure of the severity of the US weather: that is, the average temperature each month, less the average temperature for that month in the previous ten years. … We then compare this temperature gauge to the level one month later of our US Economic Activity Surprise Index (EASI), which measures the net balance of positive less negative economic surprises over the previous six weeks.
We find that the EASI is indeed predicted by the weather, though the relationship is statistically significant only in the coldest months (November to March for the US), which intuitively is when we would expect the weather to impact economic activity the most (see Chart 7).
Chart 8 shows the correlation of surprises in the most market-moving releases with our temperature gauge between November and March; the relationship is almost always positive (Durables is the exception), but statistically significant only for the Michigan and Conference Board Consumer Confidence Surveys, the Philly Fed survey and Retail Sales.
As ever, it is harder to predict market movements than economic surprises, but we do find a statistically significant relationship between our temperature gauge and bond returns around the Philly Fed and Nonfarm Payrolls releases.
Using the temperature as a trading signal –– going long Treasuries and short the S&P around the payrolls release if the weather was unusually cold the previous month, and vice versa –– would have generated an information ratio of 0.5 for bonds, with a success rate of 55%, and 0.6 for stocks, also with a success rate of 55%, since 2000. Note that these returns are based on employing this strategy only for the November-March payrolls releases.
That suggests that the weather does provide a useful signal. That is particularly so given that our characterisation of the weather is a very simple one, and more complex measures of weather severity (e.g. incorporating, say, snowfall) may provide better results.
As for that economic data consensus, Mac Gorain has some interesting insights there too.
His basic thesis; economist forecasts for economic releases like the NFPs can help predict the likelihood of a data surprise. For instance, if economists forecast a rise from the previous month, the actual release is above the forecast 57 per cent of the time, according to JPM’s data set.
But tracking consensus won’t tell you much about how the market will actually react to the surprise …
… it is much harder to predict the market reaction than it is to predict the data outturn relative to economists’ forecasts. That implies that economists’ forecasts do not really represent the market consensus. Instead, the market assimilates information more quickly and more efficiently, and is therefore much harder to beat. This paper discusses what signals do and do not help to beat the market.
In summary, we find that trading US bonds and equities on the directionality of data surprises –– more positive surprises when economists forecast a rise in the data, and vice versa –– does not make money, suggesting that the market is aware of this bias in the consensus. Focusing on the forecasts of economists with a better track record is more profitable, but still offers only modest returns. Extrapolating surprises from recently released, related data is profitable when applied to ISM on the basis of European PMIs released earlier that day. Weather patterns also provide good returns as neither forecasters nor the market seem to recognise sufficiently the impact of unusually low temperatures on activity. Finally, we get good returns from going against the grain of recent surprises, going short USTs when data have been surprising on the upside, and long on a slew of negative surprises.
Weirdo weather. And markets.
Trading data isn’t just beating consensus: study – MarketWatch
The economic impact of bouts of severe weather are easily exaggerated – The Economist