Oct 10 2011

The Old Poor Richard’s Farmers’ Almanac(ck)

It would be nice to have an accurate prediction of the weather. Although for me personally the ability to predict the weather mostly affects my recreation – not my livelihood or my safety. Perhaps it would help prepare for a particularly harsh winter if I knew one was coming, but then again I could just prepare for a harsh winter in any case.

In years past, and for many people and in many parts of the world today, weather prediction can make a much bigger difference than planning their activities for the weekend. Farmers, in particular, depend on the weather for their living, and perhaps even their survival.

It is no wonder, then, that there is a big market for long range weather prediction. In the early US, almanacs that provided information on weather were many and popular. One of the most famous is Poor Richard’s Almanack, because it was published by Benjamin Franklin, who published it from 1732 to 1758. (Incidentally, Franklin used the pen name Richard Saunders for this publication – with no relation to the famous Australian contemporary skeptic of the same name.)

These almanacs were like household handbooks for the colonies – they contained calendars, astronomical information, statistics on precipitation and temperature, witty and entertaining aphorisms – and long range weather forecasts.

Two of these almanacs survive until this day. The Farmers’ Almanac and the Old Farmers’ Almanac. Until I was researching for this post, I had no idea these were two separate almanacs. The Farmer’s Almanac was founded in 1818 and is published in Lewiston, Maine. The Old Farmers’ Almanac was founded in 1792 and is published in Dublin, New Hampshire.

Both almanacs engage in, among other bits of useful information, long range weather prediction. The Old Farmers’ Almanac states:

Based on his observations, Thomas used a complex series of natural cycles to devise a secret weather forecasting formula, which brought uncannily accurate results, traditionally said to be 80 percent accurate. (Even today, his formula is kept safely tucked away in a black tin box at the Almanac offices in Dublin, New Hampshire.)

“Thomas” refers to the first editor, Robert Thomas. What they are describing in the paragraph above is a typical process of mining streams of data for patterns and trends. In the stockmarket this is called technical analysis – using past trends to predict future trends. This is, essentially, pseudoscience.

The fact that it is secret is also a clue. Legitimate science must be transparent. If there were scientific knowledge about long range weather prediction then that would result from published research that was evaluated and discussed by a community of experts. A body of scientific knowledge is built over time, in the research literature. It’s not kept in a black tin box.

The reference to the black tin box also suggests to me that the modern incarnation in the almanacs of long range weather prediction is a bit tongue-in-cheek. It is meant for fun and entertainment (similar to the justification given by newspapers for including astrological readings). But neither do they admit it is not legitimate, because they know a large part of their customers take it seriously.

They therefore can have it both ways – not really claiming to be scientific or accurate, but getting the benefits of making such claims.

As further evidence they say that the predictions are “traditionally said to be 80 percent accurate.” What does that mean, exactly?” They are not citing scientific evidence, nor making the claim themselves, just citing “tradition.”

The Farmers’ Almanac is similar. A Consumer Reports story writes:

The Farmers’ Almanac bases its predictions on “a secret mathematical formula using the position of the planets, tidal action of the moon and sunspots,” says the AP. The scientific community hasn’t embraced its methods, still it’s fun to see what the 194-year-old almanac predicts each year.

Don’t confuse this with science, but it sounds sciencey, and it’s fun.

I could find one reference to an actual study of the accuracy of the Old Farmers’ Almanac. The Newshour Online reports:

In the October 1981 issue of Weatherwise, pages 212-215, John E. Walsh and David Allen performed a check on the accuracy of 60 monthly forecasts of temperature and precipitation from the Old Farmer’s Almanac at 32 stations in the U.S. They found that 50.7 percent of the monthly temperature forecasts and 51.9 percent of the precipitation forecasts verified with the correct sign. These may be compared with the 50 percent success rate expected by chance.

No better than chance is what we would expect from a mysterious method of weather prediction kept in a tin box.

But how good is the current state of the science of weather prediction? Anyone who pays attention to the weather news has probably figured out that meteorologists are pretty good at predicting weather over the next 2-3 days (at least with a fair probability). But 5-10 day forecasts, while better than chance, are pretty dodgy. Beyond 10 days it is difficult to impossible to say anything specific about the weather.

Everyone recognizes that there is no method that can predict that it will snow 8 inches in Connecticut on January 7th 2012. That kind of precision that far in advance is completely impossible. But what about general trends – it will snow more than average this winter?

It is more accurate to refer to such predictions as climate predictions, rather than weather predictions. Climate is highly predictable in terms of seasonal variations in particular regions. I can tell with a very high degree of accuracy and confidence that it will be cold this winter with a high chance of snow.

Long range (decade to century) climate modeling is another matter, not relevant to this post except to note it as a distinct category of climate prediction.

In the middle there is 3-13 month climate trend prediction – which most closely resembles the types of forecasts made in the almanacs. In other words – will this winter be typical, or will it be warmer or colder than usual, and will it have less or more precipitation than usual? It is not impossible to make these kinds of predictions.

The National Weather Service, in fact, has its Climate Prediction Center that does just that. This is a close as modern science can get to the magical mysterious methods that the Almanacs keep so secret. There are, in fact, several climate patterns that climatologists have learned to recognize that do influence short term (1-13 month) seasonal climate. You have probably heard of El Nino and La Nina. They describe El Nino:

The term El Niño refers to the large-scale ocean-atmosphere climate phenomenon linked to a periodic warming in sea-surface temperatures across the central and east-central equatorial Pacific (between approximately the date line and 120°W). El Niño represents the warm phase of the El Niño/Southern Oscillation (ENSO) cycle, and is sometimes referred to as a Pacific warm episode. El Niño originally referred to an annual warming of sea-surface temperatures along the west coast of tropical South America.

La Nina is essentially the cool phase. There is also ENSO neutral, when neither El Nino or La Nina is present. Based upon what ENSO pattern we are in scientists can make general statements about the seasonal climate. You can see the kinds of data they can provide here – maps of whether temperature and precipitation will be above average, below average, normal, or equal chance of either.

These are still probabilistic statements, not really predictions.

Still, I hear many people quoting one almanac or the other about what kind of winter we are in for. They don’t seem to realize that the almanacs are using 200 year old pseudoscientific methods that have never been validated. Despite the coy marketing of these predictions, many people take them as legitimate.

As evidence for this, one editor of the Old Farmers’ Almanac, Roger Scaife, tried to bring the almanac more in line with modern science:

Scaife also committed the greatest of all Almanac blunders: He dropped the weather forecasts! In their place, he substituted temperature and precipitation averages. The public outcry was so great that he reinstated the forecasts in the next year’s edition, but it was too late to save his reputation.

It seems that the public did not want the scientific information – they wanted the predictions made by mysterious methods. I can understand, for marketing reasons, why future editors of the almanac would not consider dropping the predictions. But here is a recommendation – why not get rid of the two century old dubious methods and replace them with the climate forecasts made by the National Weather Service?

6 responses so far

6 thoughts on “The Old Poor Richard’s Farmers’ Almanac(ck)”

  1. daijiyobu says:

    Re.: “mysterious methods […like] ‘a secret mathematical formula using the position of the planets, tidal action of the moon and sunspots'”, to quote Morrissey’s “Ouija Board”:

    “Ouija board
    Would you work for me? […]
    Would you, ouija board
    Would you help me ? […]
    The table is rumbling
    The glass is moving
    ‘No, I was NOT pushing that time’.”


  2. splicer says:

    The new almanac? This guy “draws on 60 years of weather data” and sells it to energy companies so they can lock in fuel prices during big temperature swings. He uses statistical relationships in the data and worked with the “Scripps Institutution of Oceanography to find a commercial use for its academic research.”

  3. SARA says:

    “They found that 50.7 percent of the monthly temperature forecasts …. verified with the correct sign. ”

    I’m wondering how they calculated that. I haven’t seen an almanac for years, but I don’t see how 50% is just chance with temperature. With precipitation- its a yes/no answer, or so I assume. So the 51% is obviously chance. Unless they have to get snow/rain/sleet/hail correctly.

    But temperature range must have more than options. And the variability of Spring and Fall would in particular seem to make the odds of being right by chance lower than 50%? I’m not good with stats so, I perfectly willing to be corrected on that point.

    Of course temperature is easier to predict generally. I know its not going to be avg of 80F in CT in January. Its likely to be averaging in the 40’s (I’m guessing here – I didn’t check the history) and so my odds are increased by knowing the history of the area.

    I suppose they must have factored in that in some way.

    I don’t assume that the Almanac has any predictive qualities, I’m just curious about how the 50% was considered chance.

  4. tmac57 says:

    SARA- That’s a good question. If I had to guess, it would be that the predictions that the Almanac editors use are derived from long historical trends that produce mean temperatures,rainfall etc.,such that predicting the mean trend would put you roughly in the midrange of probable possibilities.Falling on either side of the mean would be about 50/50 so your average would be close to that. I am not well versed in stats.,so maybe someone who is could expand or correct that notion.

  5. bachfiend says:

    I don’t know where else to put this comment.

    On that site of supreme stupidity, “Egnorance”, that I read for endless amusement, Michael Egnor has managed to outdo himself with his latest comment:

    “‘Yes, you are a Paleyist. William Paley attempted to prove the existence of a deity by claiming that there is design in nature, and that is exactly what you do’.

    Paley’s design argument was based on complexity.

    My argument is that of Aquinas, which is based on teleology, irrespective of complexity.

    The teleological behavior of a single electron in a hydrogen atom is as much proof of God’s existence as is the teleological behavior of the genome.

    This is a major debate in the ID/Thomist community. I thought you had some insight into it. My mistake”.

    Teleological behaviour in a single electron? How does that relate to the Heisenberg uncertainty principle?

  6. Jaunty says:

    SARA and tmac57,
    Yeah, I think the almanac gives more qualitative weather forecasts (eg. ‘stormy’ or ‘cooler’) rather than specific temperatures for specific days. If this is all you have to go on, then you can simply compare their prediction of cooler/warmer with how the day’s temperature actually varied from the mean temperature.

    “They found that 50.7 percent of the monthly temperature forecasts …. verified with the correct sign. ” would mean that 50.7 percent of the time the almanac’s prediction fell on the same side of the mean as the day’s actual temperature. If this is all the almanac predicts, then you can’t do too much better than this. If they predicted actual temperatures, then you could make a more quantitative comparison between the forecast and real temperatures (such as with the least squares method).

Leave a Reply