Wedding bells... again...?

April 24, 2011 8:02 AM

What's the mean number of marriages for Americans?

How many times, on average, do Americans get married within their lifetime, and what's the standard deviation? I absolutely cannot find this parameter anywhere!

Please note I'm not looking for average number of marriages/1000 or anything of that nature; rather, how many times one American will be married in their lifetime.

Thanks very much, social scientists of Mefi!

posted by WidgetAlley to Society & Culture (20 answers total)

I don't know the answer to your question, but I wanted to point out a couple of statistical problems with answering the question.

First, if you want to know what's typical, you probably want the median, not the mean, since the mean will be pulled up by the Larry Kings and Elizabeth Taylors and so even though it's the mean, it will not be at all common to have that many or more marriages.

Second, inside you phrased your question as how many times the average american WILL get married, not how many times they HAVE been married. Knowing how many times they WILL get married requires something like the kind of calculations used to do life expectency, since you need to know not just what has happened, but what is likely to happen in the future. So with life expectency you take what has happened "given that there are 15999 people in this city who have lived to be 43 at this point, how many of them will die at 44? At 45? etc." and do that for every age. With the marriage you need to figure out "given that there are X females who are Y years old and never married, what's the probability they'll marry next year, the year after, etc. etc." And do that for other ages and other marital statuses (e.g. given 458782 men who are 32 and currently married, what's the probability this marriage will end and they will marry again (next year, the year after, etc. etc.). Then from THOSE tables you can calculate the median or mean.

That said, the answer to the median is surely less than 1 since most people only marry once and many people never marry.
posted by If only I had a penguin... at 8:14 AM on April 24, 2011 [2 favorites]

Response by poster: Oooh, good call on the "will get married," I was speaking in the hypothetical and not thinking about it. Have been married is fine too, this is for a quick-and-dirty calculation so absolute accuracy is not essential.

As far as mean v. median, I am fine with right-skew due to high-lying outliers, and I will definitely need the mean since I am trying to (roughly) calculate probability based on a z-score.
posted by WidgetAlley at 8:29 AM on April 24, 2011

Mean should be sufficient for this question. I imagine the distribution is normal or approximated so; accordingly the median and mean will be roughly similar.

Medians are useful for non-normal distributions like net worth or income.
posted by dfriedman at 8:34 AM on April 24, 2011

I don't know if you've seen te links at this google search but there may be some useful links here: http://www.google.com/search?q=average+number+of+marriages+per+American&ie=UTF-8&oe=UTF-8&hl=en&client=safari
posted by dfriedman at 8:37 AM on April 24, 2011

This data may not exist. A report from 2002 gives probability values for divorce and remarriage.

"...the probability of remarriage among divorced women was 54 percent in 5 years--58 percent for white women, 44 percent for Hispanic women, and 32 percent for black women. However, there was also a strong probability that 2nd marriages will end in separation or divorce (23 percent after 5 years and 39 percent after 10 years)."

One would need to adjust for race, gender, age, SES, and childhood family life and then calculate probability of marriage, probability of divorce, and probability of additional marriages/divorces... Don't look at me, I'm only training to be a statistician.
posted by hammerthyme at 8:40 AM on April 24, 2011

There's no way the distribution is normal. Zero is probably the second-most-common value, following one, and you can't have fewer than that. I don't think z-scoring the value is really applicable. Distributions with long right tails and no left tail will suggest the proper stats.

Is your underlying question something along the lines of, "Given a random American, ignoring age/race/whatever, what is the expected value for the number of times he/she has been married?" Or is it, "Donald Trump has been married X times; how far out of the 'norm' (colloquially speaking) is he?" The first is easier than the second.
posted by supercres at 8:44 AM on April 24, 2011 [4 favorites]

Best answer: I imagine the distribution is normal or approximated so; accordingly the median and mean will be roughly similar.

There's no way the distribution is normal. A normal distribution is symmetric. If the distribution were normal and the mean were 4 (which is a pretty high guess), then for every person married 6 times there would have to be a person married -2 times. Obviously that's not possible. The mean will be higher than the median, but if you just want the mean, that's easier.

Ok, here's data from the GSS (national probability sample):

Dataset: General Social Surveys, 1972-2006 [Cumulative File]
Variable marnum : R MARRIED MORE THAN ONCE?
Literal Question
490. Have you been married more than once?

Values 	Categories 	N 	NW
1 	YES 	310 	292 	 24.2%
2 	NO 	889 	916 	 75.8%
0 	NAP 	49817 	49808
8 	DK 	0 	0
9 	NA 	4 	4 

Variable marital : MARITAL STATUS
Literal Question
4. Are you currently--married, widowed, divorced, separated, or have you never been married?

Values 	Categories 	N 	NW
1 	Married 	27998 	31629 	 62.0%
2 	Widowed 	5032 	3369 	 6.6%
3 	Divorced 	6131 	4472 	 8.8%
4 	Separated 	1781 	1334 	 2.6%
5 	Never married 	10064 	10209 	 20.0%
9 	No answer 	14 	8 

So based on this, 20% are never married, 60% are married once, and 20% are married some unspecified number of more than once. According to my calculations, the marriied more than once pop would have to have been married 5 times on average to bring the total mean to 1.5 (rounding to 2). THey would have to be married an average of 7 times to bring it up to actual 2. Remember for them to be married an average of 5 or 7 times, would require some proportion to be married much more than that. Unlikely.

So your answer is surely 1.x. Where X is less than 5.
posted by If only I had a penguin... at 8:53 AM on April 24, 2011 [9 favorites]

It is very difficult to imagine any z-score calculation you could do in this context that would not be meaningless or worse.
posted by escabeche at 9:03 AM on April 24, 2011 [1 favorite]

To add to the problem, IMO, what you might be looking for is not marriage, but committed relationships that are not described legally.

I have a sister-in-law who has been married 4 times. Another one has never been married but is in a committed relationship of more than 20 years. (The former is a born-again nutcase and I love the whole 'solemn vow' thing in her case. It's like talking to a woman has spinach in her teeth! We see it; she doesn't.)

Anyway, I think you can figure this out, but the granularity of "marriage" corrupts the concept of long-term, semi-permanent, ostensibly-monogamous, legal/non-legal, gay/hetero couples who might not qualify for the label "married", but who are most assuredly occupying the same social slot.

Of course, if you opt for my concept, you'll have to set a threshold of some sort... how long is long enough to qualify for long-term?

I suspect that the number is well in excess of 1... probably more like 3 to 5. I do not know many couples who weren't coupled with someone else before their current ball-and-chain. I know none that would call those associations trivial and who did not enter into them hoping for lifelong commitment.

I will watch this thread with interest. It combines two of my favorite subjects... relationships and math. Now if you could only work in a squirrel or bacon, I'd be psychically orgasmic.
posted by FauxScot at 9:36 AM on April 24, 2011

What are the probabilities of first marriage by various ages for men and women?
posted by milkrate at 12:03 PM on April 24, 2011

One problem you'll find looking into this is that we're not quite sure how many people are actually married: apparently lots of people tell the Community Survey "Yes, my partner and I, we're, um, LEGALLY MARRIED, yep, that's right", and aren't as likely to tell the same thing to the decennial Census.

I'd have to look up the numbers, but one of these surveys shows more married men than women, and the other shows more married women than men. There's a couple percent - on the order of a million? couples - difference between them.

(We have the total number of marriage licenses, and the total number of legal dissolutions, but good luck making sense of it from that starting point.)

(Digression: There are probably a lot of reasons why the numbers don't jibe. Some people have been telling people "This is my spouse" for so long that that's what they tell everybody. Some of the difference between the two surveys must be people "keeping up appearances/not admitting to survey-takers that their marriage is actually over"; but I have a hunch that there are also some number of same-sex marriages hidden in the marriage data, too.)
posted by AsYouKnow Bob at 1:14 PM on April 24, 2011

That said, the answer to the median is surely less than 1 since most people only marry once and many people never marry.

Off the top of my head, I'd guess it's slightly OVER 1.0: 75-80% of Americans get married in their lifetime; and about 40% of marriages are not-first-marriages (though I forget if that's for just one of the partners) At a complete guess, I'd say there are 1.25 marriages per adult American.

If 20% never marry, and 40% of marriages are re-marriages (two or more marriages), I'd think both the mean and median should be slightly above 1.0 .
posted by AsYouKnow Bob at 1:39 PM on April 24, 2011

I don't know what definition of median you all are using, but as far as I know, the median can only take on values represented in the actual distribution. No one is ever married "slightly above 1.0" times. The median is exactly one.
posted by mr_roboto at 1:55 PM on April 24, 2011

mr_roboto, I don't think so. The median just means that you precisely at the point that splits the probability distribution into 50% one either side. Consider the following data set: {1, 2, 2, 3, 3, 14}. The median is 2.5, because that is the number that would split the set into two subsets, {1, 2, 3} and {3, 3, 14}.

You may be thinking of the mode, instead, which does have to be a value present in the original set.
posted by tickingclock at 2:03 PM on April 24, 2011

(In my example above, the mode would be 2 and 3; unlike the mean and the median, a distribution may have a single mode, multiple modes, or even no mode!)
posted by tickingclock at 2:04 PM on April 24, 2011

Argh. In my first comment, that was supposed to say, "... into two subsets, {1, 2, 2} and {3, 3, 14}." Too much coffee and stress make me weird and jittery.
posted by tickingclock at 2:05 PM on April 24, 2011

Percentile statistics (of which median is one) are probably the correct way to think about this. Been married seven times? Congrats! You're in the top 1% (99th %ile)!

With a whole number dataset like this, the median also has the be a whole number, unless the 50% division falls exactly between two values. In that case, it's the mean of those two values.
posted by supercres at 2:08 PM on April 24, 2011

With a whole number dataset like this, the median also has the be a whole number, unless the 50% division falls exactly between two values. In that case, it's the mean of those two values.

Good point! I typically only deal with continuous distributions, so I did not think of this at all.
posted by tickingclock at 2:27 PM on April 24, 2011

This interactive map from the Pew Research Center might be helpful if you'd like a state-by-state or gender-based breakdown.

These numbers are from the CDC based on the 2009 report:

Marriage and Divorce

(Data are for the U.S.)

* Number of marriages: 2,077,000
* Marriage rate: 6.8 per 1,000 total population
* Divorce rate: 3.4 per 1,000 population (44 reporting States and D.C.)

FYI, if the man and a woman and both are over age 28 at the time of their first marriage, their likelihood of divorce is cut in half - that gets reduced to 11% chance of divorce for women and 15% chance for men over age 35 that marry for the first time.

So, your answer will change based on the age, gender and location of the American.
posted by Unicorn on the cob at 3:54 PM on April 24, 2011 [1 favorite]

Response by poster: Thanks everyone! I got what I needed. I was merely using an example of a friend of ours who has been married several times to illustrate how Z-scores work to a third party; but then I couldn't find average number of marriages and became curious as to whether the information even existed. It seems it doesn't, not definitively in any case. Thanks very much!
posted by WidgetAlley at 7:43 AM on April 25, 2011