Gedankensexperiment

January 30, 2009 9:13 PM

People say condoms are 95-98 percent effective, if used correctly. How does that figure compare to other risky things, like commuting to work, riding a motorcycle, riding a bicycle in traffic, etc?

Here is my understanding: If randomly chosen couples each have sex 100 times (or if me and her use a condom 100 times, in parallel universes) then the number of pregnancies observed, divided by the number of couples, approaches a certain number, as the number of couples goes to infinity (presumably 2 or 5.)

If we were to replace “two people having sex with a condom, for X amount of time” by “two people riding on a motorcycle, for X amount of time” and “get pregnant” by “get injured” how would that final number change? Ditto for “riding a bike in traffic”, “skateboarding in traffic”, and similar things.

I'm just trying to get a feeling for how significant 95-98 percent really is, so I can take that into consideration. Of course the more important question is how much do I like her, but for now, Hit me with the statistics! Thanks Hive mind.

posted by anonymous to science & nature (11 comments total)

At worst, most psychological research accepts a result (as much as you can accept a result with this sort of thing) with a 95% level of accuracy.
posted by theichibun at 9:32 PM on January 30

The "percentage effective" for birth control methods is per year of use, not per session. Used correctly, condoms have a 95-98% chance of preventing pregnancy for a given year of sexual activity.
posted by NMcCoy at 9:44 PM on January 30 [1 favorite]

First, you're misunderstanding the effectiveness rate as they're usually reported. It's nothing to do with number of times a couple has sex; it's just the percent of couples who get pregnant per year.

Second, if it were randomly chosen couples, the rate would be drastically higher than 2--5%, as that is the failure rate for perfect use. Not normal use. The failure rate in normal use is that 10-20% of couples using condoms get pregnant per year of use.

If you want a handy touchstone for what 2--5% is:

2--5% is like flipping a coin five times, and having them all come up heads.*

But remember with condoms, 2--5% is perfect use. Perfect. Always on, each and every time. No rubbing your cock on her external junk, or her thighs, EVER. Always pulling out promptly. Always stopping if you think it might have broken. Never using petroleum-based lubricants. Always using a new one, every time. Always throwing it away and getting a new one if you put it on the end of your schlong and realize you have it the wrong way around (or unrolling it the wrong way).

If you want a handy touchstone on what a more realistic 10% is:

That's more or less like flipping a coin three times and having it come up all heads.**

*1/32, or 3.125%, to be precise
**1/8, or 12.5%
posted by ROU_Xenophobe at 9:48 PM on January 30 [3 favorites]

The pregnancy percentages aren't calculated the way you describe. Condom and other contraceptives pregnancy rates are calculated by getting a sample population to have sex using contraceptives for one year and the percentage that have no unplanned pregnancies is the number you see contraceptive companies publish.

This complicates the statistic; some people may have sex only a few times per year while others may have sex several times a day.


This article has a few neat tables of stats on motorcycle safety. Also check this article out it has some comparisons on risk of airplane vs. car vs. train.
posted by gregr at 9:52 PM on January 30

I think you are misreading the condom failure rate numbers. The 98% "effectiveness" for condoms does not mean 2 pregnancies out of 100 sex acts. If that was the case, they would barely be useful. It actually means 2 people out of 100, who use condoms as a sole form of contraception for a year, end up pregnant in that year. (My understanding of the methodology is that they take groups of people who say they're using a particular birth-control method, and then see how many of them have unintended pregnancies over time.)

So if you wanted the per-sex-act failure rate, you would have to divide the 2% annualized failure rate by the average number of sexual encounters a person has in a year. It makes the number a lot lower, obviously, but it's not really any more useful in everyday life.

If we want to compare this to automobile accidents, first you start off with the number of deaths per billion passenger-miles, allegedly 11.8 for cars. Then you'd need to figure out the average number of miles traveled by a person in a year. The total number of passenger-miles traveled in the US, in passenger cars over public roads, is about 2.7x10^12 miles according to this. That's roughly 8800 miles per person in the country, if the miles are evenly distributed among the ~300M residents. (Which of course they almost certainly are not... if you wanted to look at "upper-middle class car-owners" I bet the number of miles would be much higher. So this is where you want to start varying with numbers more appropriate to your target.) If you plug and chug, and assuming I kept my orders of magnitude straight, I think this works out to something like a 0.01% chance of becoming a fatality per year.

A good way to check the reasonableness of this calculation would be to multiply it by the population and see if the result (30k/yr, again assuming I did my math right) anywhere near the actual number of reported road deaths (~40k/yr). This suggests the number is lower than it ought to be.
posted by Kadin2048 at 9:57 PM on January 30 [1 favorite]

Hmm, 5% is the chance of catching the flu in a not-bad year of it.
posted by losvedir at 4:45 AM on January 31

It is seems odd to try to measure such disparate risks. If condoms fail, someone can come to life. In a motorcycle accident, someone can die. The way that security people handle this is to compute what is known as the annualized loss expectancy. You figure out what the cost of the bad result would be, and what the chance of it happening is to provide an estimate of the likely cost per year of that event. Then you figure out what preventative measures will be most cost-effective in reducing the losses. I will try and give a very simple example using totally made-up numbers.

Let's say that you ride a motorcycle 10,000 miles per year; that motorcyclists have one accident per 1,000,000 miles on average; that an accident costs $100,000 in lost wages and medical bills and motorcycle damage on average. You can figure that for any one year, your expected loss will be

(10,000 miles) x (1 accident/1,000,000 miles) x ($100,000/accident)

which, if you work out the numbers and units is $1,000 per year expected loss.

Let's also say that you are out having lots of protected sex with condoms. There is a 2% chance of having a child per year. Now say that the cost of a child is $5,000 in birthing expenses and $15,000 per year until age 18. Without getting into the time value of money, we can say that the cost is therefore about:

$5,000 + (18 * $15,000) = $275,000

then annually you can expect:

.02 * $275,000 = $5,500 expected loss

So in this example, the condoms are more dangerous than the motorcycle - at least in the annualized loss expectancy realm. Of course, in this simple example we are ignoring all sorts of factors: you might be an above or below average motorcycle rider; you might get special training or wear highly visible clothing that reduces your chance of accidents; you might die, and that is really bad. You might have multiple births, or a special needs child that is more costly, or be sleeping with someone incredibly wealthy who will then support you and the child in great luxury. So the estimates can be made more accurate with all that information.

So I don't think we can accurately analyze your case here without lots more information. Also remember, that what is expected and what happens can very significantly - averages are the results of a wide variety of results.
posted by procrastination at 9:19 AM on January 31 [1 favorite]

Let me first state the caveat that I don't believe these are very good comparables for a variety of reasons, but since you asked . . .

Here are the stats for bicycling, motorcycling, and driving for a whole year--since that is how the condom failure rates are calculated--over a whole year of "normal use":

Bicycle: In the U.S., about 27% of the population bicycles regularly (defined as at least once/month during good-weather months). About 800 bicyclists are killed annually. That works out to ~0.001% annual chance of fatality.

Motorcycle: In the U.S. there are about 5,370,000 motorcycles registered and in 2006, 4810 fatalities. That works out to about 0.09% fatality rate.

Motor vehicle: I'll assume that pretty much everyone in the U.S. rides in a motor vehicle at least occasionally and roughly 30,000 motor vehicle occupants are killed annually. That's an annual 0.01% fatality rate.

More comparable might be the motor vehicle injury rate: About 2.9 million are injured by motor vehicles annually, meaning the annual injury rate is about 1.0%

Of the comparisons, bicycle is probably worst (most people who own a bicycle ride it only occasionally; most people in a relationship seem to have sex more than occasionally) and motor vehicle probably the best (most Americans drive regularly, though a few drive almost constantly and a few don't even want to drive at all--yup, you could pretty much substitute "people in a relationship" for "Americans" and "sex" for "drive" right there).

Regardless, your "risky" behaviors you chose as comparables are actually all a couple of orders of magnitude (or more) less likely than condom failure.

(Though as others pointed out up-thread, though the chances are lower, the consequences are quite a lot more severe--an unexpected pregnancy one can deal with; being dead, not so much . . . )
posted by flug at 11:58 AM on January 31

FWIW and IMHO, comparing condom failure risk to something like the chance of hurting yourself skateboarding really isn't that helpful. Except maybe in the sense of realizing, "If I go skateboarding there is a pretty high chance that sooner or later I'll get injured, like scraped up pretty bad or maybe break a wrist sprain and ankle. But I'm prepared to accept that risk and deal with the injuries if/when they happen."

In somewhat the same way, annual condom failure rate is high enough (something like one in ten or one in twenty couples each year, if you look at the statistics for "real-world" rather than "ideal conditions") that you really shouldn't engage in this activity unless you AND your partner are prepared to deal with the consequences of pregnancy.

If you and your partner are not ready to do that, your not ready to do it . . . IMHO.

Obviously, there are a number of ways to deal with failure if it happens (day after pill might help in some cases; abortion; carrying baby to term).

And you can also take steps to reduce the risk--combining birth control methods, for instance, reduces the failure rate by quite a lot.
posted by flug at 12:19 PM on January 31

And you can also take steps to reduce the risk--combining birth control methods, for instance, reduces the failure rate by quite a lot.
posted by flug at 2:19 PM on January 31 [+] [!]

This. I know it wasn't your question, but it'll do whole lot more to allay your fears (and actually significantly reduce the risk of pregnancy) than thinking about motorcycle accidents ever could; try running the stats on failure rates for hormonal birth control plus good condom use, or some other combo. The more important question is not how much you like her, but rather how you can get your sex on while minimizing the chance of babymaking in the process.
posted by you're a kitty! at 12:49 PM on January 31

Oh also - the post title is genius.
posted by you're a kitty! at 1:25 PM on January 31