top of page

Long marriage: The right time

Updated: Jun 19, 2024


At the beginning of the year, I spoke to a good friend about getting married. We both wondered whether a "special" wedding day like 22 February 2022 would help us to have a long (and happy) marriage. For various reasons (not least the fact that all the registry offices in Munich were already fully booked), conducting our own experiment was out of the question. Fortunately, as a statistician, you have other ways of approaching such questions...


Superstition has always played a major role in wedding customs: the groom is not allowed to see the bride's dress before the wedding; if the wedding ring falls to the ground during the ceremony, this will cause disputes in the marriage, and a lucky cent in the bride's shoe protects against money worries.

All these traditions are intended to protect against misfortune and ensure the longest possible marriage. When it comes to the date of the wedding, some brides-to-be also rely on special "lucky days", with 22 February 2022 being a prime example. On average, more than seven times as many couples get married on these lucky days compared to a normal day.

This is based on a dataset of all marriages that took place in Switzerland between 1987 and 2010. But is there any truth to these lucky days?

Do marriages that take place on special dates really last longer?

Statistically speaking, yes. At least in Switzerland. This is the astonishing result of a survival time analysis that we carried out on the basis of these married couples.

For each individual couple, it is known whether a marriage still existed (at least on paper) in 2010. In a good 250,000 cases, love had not survived, 225,000 couples had been divorced by a judge and just under 30,000 by death. It is also possible to calculate what percentage of couples were still married after one, two or seven years and how long a divorced marriage had previously lasted.


Repdigits bring marital bliss


A simple, initial approach compares the relative frequency of divorced marriages with a snap number wedding day and those married on a "normal" day. The proportion of divorces is higher for marriages contracted on any day. A chi-square test is carried out to check whether the wedding day actually has a significant influence on the divorce rate. The result is highly significant.

However, this procedure does not utilise the information about the date of divorce.

After all, the risk of divorce does not remain constant over the course of a marriage.

Furthermore, because there are relatively few "long-term marriages" and relatively many "short-term marriages" in the data set, the significance of the divorce figures after a long marriage is much lower than in the early years of a marriage. However, modern statistical methods can take such special data situations into account.


Quickly down the aisle - those who marry today are less likely to get divorced

A study has found out which date couples should choose for a long marriage.

Even the BILD newspaper took an interest in our study...


Cox regression is typically used to analyse survival probabilities in medical issues. For example, it is used to analyse whether a certain therapy or preventive measure leads to a longer life. In our case, the preventive measure is the wedding anniversary on a particularly "lucky" date. The difficulty with such data is that observations are right-censored. This means that at the end of the observation period (i.e. on the right-hand side of the timeline), no statement can be made about the remaining marriages as to whether the marriage will end at some point due to death or divorce.

The Cox regression model now calculates the so-called hazard rates of the censored divorce data, taking into account the increase and decrease in the number of all married persons. Widowhoods are therefore also included here. The hazard rate here refers to the probability of divorce in a certain arbitrarily small period of time. The Cox coefficient of marriages on "lucky days" is 0.7999 and is highly significant. This means that

the risk of divorce over the maximum 24 years observed is reduced by a good 20 per cent on average if people get married on a "repdigit" day.

The Cox regression thus confirms the result of the "naive" calculation of divorce rates.


When love leaves


The survival curves can also be represented graphically in a so-called Kaplan-Meier curve. The coordinates of the curve are derived from the Kaplan-Meier estimator. It indicates how many marriages did not end in divorce by 2010. The survival probability here therefore refers to the probability of still being married after one, two, seven or even 24 years.

In other words, the risk of divorce is calculated for each individual year of marriage and for the total duration of 24 years. Insurance companies do nothing else when they draw up mortality tables, for example to calculate life insurance premiums. The probability that any Swiss marriage concluded on a "normal" day will survive for at least 23 years is therefore 63.2 per cent. For marriages with a "special" wedding anniversary, however, this probability is 69.3 per cent. (Since the first Schnapszahl date was 8 August 1988, there can be no corresponding marriage with a duration of 24 years in this data set).

Diagramm zu Scheidungsrisiko für jedes einzelne Ehejahr in Abhängigkeit des Hochzeitsdatums

Survivalkurven Ehe. Bildquelle: STAT-UP


Incidentally, the greatest risk of getting divorced in a particular year of marriage if you were still married at the beginning of the same year is actually in year seven for both groups.

The risk of divorce in the seventh year of marriage is about half as high as the average risk of divorce in all other years!

The "darned seventh year" therefore applies regardless of whether you got married on a "lucky day" or not.


... and the stork brings the babies?


Correlation does not equal causation - otherwise it could be statistically proven that the stork brings the babies. This is because the number of births has fallen in parallel with the stork population in Central Europe. So perhaps the reason for the longer-lasting marriages is the luck of the draw. But perhaps it is also due to the reduced potential for arguments, because it is easier to remember a date...

If you want to start your own marriage with the best possible conditions from a statistical point of view, but don't want to wait until 3 March 2033 to get married, you can count on the (bad) luck of Friday the 13th. This day shows a coefficient of 0.9597 in the Cox regression, meaning that the predicted risk of divorce is reduced by 4 per cent compared to the average marriage. And the next date would be 13 May 2022...

For further reading: Hoppe E, Schüller K: Analysis of Marriage Durations. Applying the Cox Proportional Hazards Model to Analyse the Effect of Wedding Dates on the Duration of Marriages. DOI: 10.13140/RG.2.2.14091.16167

bottom of page