I dare you to come up with an explanation for this

Add the following study to the long list of ones I don’t believe. The article claims that astrological signs are better predictors of accidents than age or postal code.

We’ve reported on bizarre patterns in outcomes related to month of birth before (like overrepresentation of certain birth months in the NHL or World Cup), but those patterns were readily related to cutoff dates for age eligibility. This one, on the other hand, makes no sense to me.

I dare a blog reader to come up with a reasonable explanation other than errors in the study. Note that the claim is not that accident rates are higher only just after you get your license, these are rates for people of all ages.


That's hysterical! Between the two Leos and the one Aries that I know well, only one (Leo) was involved in a 2-party accident, ruled the other driver's fault (no, I don't know that driver's sign). The only other accidents for the other two were stick-shift cars deciding to slip out of gear while parked. Oddly enough, that particular Leo is an "I own the road!" sort while the Aries there is more of a passive driver, so long as absolute control of the car belongs to her.

I'd be more interested then in the specifics between 2+ -party accidents versus no-fault insurance claims.

Chris 99

A key point is that the month is almost certainly what is significant, not the astrological sign -- if in fact the study is legit. Once one leaves the astrology angle out, this loses its preposterousness and becomes an interesting puzzle to contemplate.



"The most common birthday in the USA is October 5, whereas the least common birthday is May 22."

It has to do with the most common months people are born of which I was too lazy to look up (I did find the above statement fom the above link). I was also too lazy (I have a newborn at home) to see if in the study they randomly selected 100,000 people or slected 8,333.33... people from each of the 12 Zodiac signs and tested their driving records. I have a feeling that it was the former and that certain signs are weighted based on the most common times to be born. I win.. YAY ME!!


The info on InsuranceHotline never actually states any numbers or whether the results were statistically significant. I think they just ordered them and, wha-la, instant press. Guess you'd have to buy the book to find out...


The study claims that drivers born in Sept-Oct and Jan-Feb are pretty bad, while drivers born in May-July are pretty good - if most people get their license on/around their birthdays, maybe it has something to do with the weather conditions when you first learn how to drive => more or less subsequent aggresiveness on the road (just like if you first learn to drive on a crappy car, you'll ruin subsequent cars 'cause you'll pull too hard on the clutch/press too hard on the pedals, etc... )


Perhaps it's because age and postal code are already bad indicators of accidents, so it doesn't take that much random variance to get something else to look like a better indicator. The problem is confounding factors. Age is related to general health, reflexes, memory, range from home, visual acuity, experience, and a thousand other things at one or two removes, for example.

I will believe it when a worldwide study is done and for any given population of humans (e.g. all males born in 1974), it can be shown that birth sign varies continuously and similarly with accident rate and intensity. Why those stipulations? Because birth sign is supposed to impact the individual in such a way that as you approach the cusp, you take on mixed properties, so it isn't quite a discontinuous variable.


It could be correlated in the same way that schizophrenia is correlated with birth month. Some environmental factors that varies seasonally affecting gestation. I notice that winter an dearly spring births are more likely to develop schizophrenia. This has been suggested as a result of either vitamin D deficiency during gestation or as a result of the mother getting the flu during gestation. A similar mechanism could be involved in this, affecting the risk-taking tendency of people born in those months.

Just an idea.


(cross outs unintended)


Waiting for the results to come back from the peer review... :)

Did you click through the news story to the "study?" He has the signs stack-ranked but no numbers. I'd be curious what the proportional difference between the tops and bottoms of his lists are.

Note also that he's promoting his already-published book, the thesis of which is that astrological sign has something to do with how you drive... Hmmmmm!

Doug Nelson

He's selling a book and promoting a web business. Therefore there are no errors in the study (the data is quite deliberately rigged).


Could it have something to do with influences of others in the same grade? Most drivers get their licenses during High School. In many school districts, birthdays in Sept-Oct are unlikely to have many older friends in the same school year. Therefore, they have fewer cautionary tales of classmates who made bad driving mistakes. Later in a semester, caution is likely to be in mind due to memories of mistakes made by other students. It seems that the correlation of birth month and age within a school class may have some effect.


Here's a better link: http://www.hamiltonspectator.com/NASApp/cs/ContentServer?pagename=hamilton/Layout/Article_Type1&c=Article&cid=1166050213436&call_pageid=1020420665036&col=1014656511815

Solve away!


For any atrological significance, you have to demonstrate that star sign is a better correlator than month of birth.


Maybe your missing the point here. Perhaps what we should be taking from this is that not every statistical correlation means anything. If you do a greater sample, or someone else repeats it, it may simply vanish. Any statistic should have some kind of "confidence" value, but that doesn't say any particular test won't give miss leading results. This would be an interesting topic for a blog post or 2.


This has been known about for years - it's been reported in the UK women's press for about 8 years, and by various insurance companies (Directline I believe, and then Diamond)

There are various possibilities - gestational differences - but I sometimes wonder if things like the moon really do affect things.


There is plenty of evidence for annareaven's idea. Your chance of becoming schizophrenic certainly vary with the month you are born. There was a flu epidemic in Denmark in the 50's or 60's; a study found that children who were in the 2nd trimester during the epidemic had an unusually high rate of schizophrenia.


While I would obviously like to play with the real data, reading between the lines I see some funny stuff right away. Take the following quote, for example:

"I'd rather get into a car with a 24-year-old Leo than a 25-year-old Aries," Romanov said.

So Astro sign shows a greater difference than a single year of attained age, according to Romanov. I don't doubt that at all. I work in health related insurance and I would expect health to vary much more by age than the age impact on driving record. Even in health insurance a single year shows very little marginal difference in expected claim cost... except at the two tails. Very old and very young show greater change in claim cost per year. With auto insurance I would expect something similar. My intuition says that close to starting driving (16-18) claims costs are much higher and then I suspect that claims costs *might* be high again in the old ages. But one would probably have to normalize by the number of miles driven. Although a 98 year old lady might be more dangerous per mile driven, she likely only drives a few miles a week. So her probability of claiming in a given year is probably lower than mine, because I commute 50 miles a day.

So back to Lee Romanov. I suspect that her data does, indeed, show that one astro sign vs. another is a better predictor than adding or subtracting a single year of age. I also suspect that relationship would change if age was bucketed into something more meaningful, like 5 or 10 year groups. I am suspicious of what the very first comment made by pvarner said: this is just a spurious distinction. If one plays with data very much one sees this all the time. It does not indicate that it is meaningful, but it does help to get hits to your web site and to sell books. Good PR does not equal good statistics.



Wow. Someone linked The Hamilton Spectator. Don't see that very often. Unless you want to find information on why Stelco or the Ti-Cats suck. But I kid. Nice to see a fellow Hammer reading an intelligent blog.

It would be interesting to see what the variance was from best to worst. Does anyone know of a link for the data? Maybe I should check Swivel.

Ryan Fox

Of course the moon affects things. Ever heard of the tide? Anyway, astrology is concerned with the stars, not so much the moon.


It's scary that the Hamilton Spectator nowhere states "according to a (fill in blank) study".

Very funny, however. If you check out the guy's page, at the bottom he applies his results to "my name is Earl", the t.v program. Probably a big indicator of how rigorous his study is. I think his sample of 100,000 drivers was accidentally mixed up with my wife's 100,000 tickets and accidents--she happens to be a winter birth. Reminiscent of sun spot activity and market cycles.