1 Introduction
The question of whether and to what extent superstitions affect people’s choices has become a concern of social and marketing researchers in the last decades in Asian countries as it is the area where people seem to be influenced by superstitious omens the most (Reference Pratt and KirillovaPratt & Kirillova, 2019; Reference Pratt and KwanPratt & Kwan, 2019). In some Asian countries people even consult with numerologists before opening a business or signing an important contract (Reference Chinchanachokchai, Pusaksrikit and PongsakornrungsilpChinchanachokchai, Pusaksrikit & Pongsakornrungsilp, 2017). Surprisingly, even people who are involved in market trading activities are responsive to such superstitions. Reference Fluke, Webster and SaucierFluke, Webster and Saucier (2014) reported that solar eclipses are correlated with lower-than-average returns on four American stock indices and these slumps were reversed the day after the eclipse. In Europe people’s choices are also sometimes dictated by superstitious habits that today may seem as cultural traditions, particularly in Czech Republic, drinking beer from the ‘wrong’ mug is considered to bring a ‘bad’ luck (Seignovert, n.d.). Thus, when buying a new mug people may be especially interested in the ‘right’ type of a mug. Reference De Paola, Gioia and ScoppaDe Paola, Gioia and Scoppa (2014) conducted an experiment (in need of replication) to determine whether the “lucky” and “unlucky” seats influence test results of students at an Italian University. The study was conducted among 700 students during important exams. The authors selected sixty-one students who had to write the exam on the seat number 17 which the Italians consider “unlucky”. Also, for more accurate results, one hundred and eight students were assigned to seats 13 and 30 (“lucky” numbers in Italy), while others performed tasks on the exam on neutral seats. As a result, the researchers showed that there was no relationship between the “lucky” and “unlucky” numbers of seats and exam grades. At the same time, women were more confident in high scores when they sat on successful seats and were more upset of the fact that their grades did not depend on the “luckiness” of the seat number (Reference De Paola, Gioia and ScoppaDe Paola, Gioia & Scoppa, 2014). Not accidentally, there are no seats 17 in Italian cinemas, rooms 17 in Italian hotels and even row 17 in “Alitalia”’s aircrafts. Many instances when superstition-driven expectations cause consumers to make purchase decisions that run counter to economic rationality have been presented in the literature (Reference Block and KramerBlock & Kramer, 2009; Reference Kramer and BlockKramer & Block, 2011).
While sometimes superstitions manifest merely in preferences without leading to substantial benefits or losses for those who account for these superstitions, in some cases they affect market outcomes and overall welfare. For example, superstitious investors submitting more limit orders at 8 (lucky number in the Chinese culture) than at 4 (unlucky number) apparently have higher trading losses (Reference Bhattacharya, Kuo, Lin and ZhaoBhattacharya, Kuo, Lin & Zhao, 2018). Buyers pay less for homes with addresses containing more unlucky numbers and fewer unlucky numbers and more for homes with lucky addresses in Singapore (Reference He, Liu, Sing, Song and WongHe, Liu, Sing, Song & Wong, 2020). Reference Antipov and PokryshevskayaAntipov and Pokryshevskaya (2015) studied numerological superstitions on people’s buying behavior in the apartment market of Saint-Petersburg, Russia. Using unique sales data for the primary market, authors compared the quantity of bought apartments on the 7th (“lucky”) floor with ones on floors 6 and 8, the same comparison was made for floor 13 (“unlucky”) with floors 12 and 14. Using the fact that all the floors were constructed alike, the researchers were able to identify the expected effects of the “lucky” and “unlucky” floors. Reference BurakovBurakov (2018) showed that a 10% discount was sufficient to mitigate the impact of such numerological superstition and help to boost the sales of “unlucky” apartments in Moscow.
The problem of childbirth scheduling for non-medical reasons (including but not limited to superstitions) is another example of biased decision-making associated with welfare losses. According to Reference Schulkind and ShapiroSchulkind and Shapiro (2014), scheduling births for non-medical reasons has become an increasingly common practice in the United States and around the world. Using data on all births in the U.S. from 1990 to 2000 they confirmed that families respond to the financial incentives (tax benefits) by electing to give birth in December rather than January and found that most of the manipulation comes from changes in the timing of caesarian sections. Scheduling births has negative health consequences (lower birthweight, a lower Apgar score, and an increase in the likelihood of being low birthweight) for the newborn from accelerating deliveries, including short-term movements within “full-term” pregnancies (Reference Schulkind and ShapiroSchulkind & Shapiro, 2014; Reference Tita, Landon, Spong, Lai, Leveno and VarnerTita et al., 2009). There have been several research studies related to auspicious birth dates. A study based on the national US data from 1996 to 2006 found that on Valentine’s Day there systematically was an increase and on Halloween – a decrease of both spontaneous and cesarian births (Reference Levy, Chung and SladeLevy, Chung & Slade, 2011). Two other studies concentrated on Chinese superstitions: the first one used microdata from individual vital statistics natality records covering all live births in California from years 1991–2002, collected and maintained by the California Office of Statewide Health Planning and Development (Reference Almond, Chee, Sviatschi and ZhongAlmond, Chee, Sviatschi & Zhong, 2015). The second — as well as 2014–2016 birth certificates data Guangdong Province of China (Reference Huang, Ma, Zhang and ZhaoHuang, Ma, Zhang & Zhao, 2020). The impact of Chinese numerological superstitions associated with numbers 8 (lucky) and 4 (unlucky), as well as of the thirteenth, especially Friday the 13th, on timings of birth has been confirmed in both studies for children born from Chinese parents. Reference Lin, Xirasagar and TungLin, Xirasagar and Tung (2006) used data from Taiwan on all singleton deliveries during 1997–2003 and showed that cesarean delivery rates were significantly lower during the inauspicious “ghost month” of July, and higher than normal during June, representing pre-emptive c-section to avoid delivering in July. These results agree with those obtained by Reference Halla, Liu and LiuHalla, Liu, and Liu (2019)
While preferences towards certain wedding dates are unlikely to have such serious impact on one’s welfare as birth date manipulations, some interesting patterns and even long-lasting consequences have been identified in the literature using data on marriages as well. Using Dutch marriage and divorce registries from 1999 to 2013, Reference Kabátek and RibarKabátek and Ribar (2018) identified several distinct types of popular wedding dates including Valentine’s Day and numerically special days (dates with the same or sequential number values, e.g., 09.09.99, 01.02.03), showing that on an adjusted basis, the incidence of weddings on such dates was 137–509% higher than ordinary dates. After statistically controlling for couples’ observable characteristics, special-date weddings were more vulnerable, with 10–17% higher divorce odds compared to ordinary dates. These relationships are even stronger for couples who have not married before.
The numbers of marriages and, especially, of births on each day of the year are surprisingly rarely disclosed publicly. For example, today this information is not publicly disclosed even by the traditionally generous US agencies. Not surprisingly, we have found no recent research on daily variations in births or marriages based on European data. European researchers have considered birth seasonality patterns using only monthly data (Reference Balan and JabaBalan, Jaba & others, 2016; Reference CypryjańskiCypryjański, 2019), but did not study numerological superstitions. To the best of our knowledge, among European countries daily numbers of births and marriages are publicly disclosed only for Denmark.
Our study is one of the first that uses European data to shed light on how the willingness of people to avoid certain dates associated with numerological superstitions and symbolism is reflected in birth and marriage statistics from a European country – Denmark. The triad “7–9–13” corresponds in Denmark to English “knock on wood” (Reference RusselRussel, 2013). Interestingly, numbers 7 and 9 are considered lucky, while 13 alone – unlucky, but the combination of 7, 9 and 13 is considered to bring luck. However, whether and how Danish numerological superstitions are reflected in important choices such as those related to birth and wedding planning have not been studied in academic literature before. In Section 2 we describe our dataset and methods. In Section 3 we present the results of our statistical analysis. Section 4 concludes and outlines some directions for future research.
2 Materials and methods
The dataset of daily marriages and births in Denmark in 2007–2019 was obtained from “Statistics Denmark” – a web-interface to Denmark’s official statistical data.Footnote 1 In Danish, like many other European languages, the written date format is “dd.mm.yyyy”, which is important to account for when looking for special dates. As all dates are from the same century we will sometimes use a shortened “dd.mm.yy” format.
Our initial data exploration has shown that Danes really believe in good luck associated with the combination “7–9–13”, even more than in the power of an internationally recognized “lucky” combination “7–7–7” (Figure 1). The largest number of marriages was on 07.09.13 (3136 marriages), followed by 07.07.07 (1967 marriages), and 18.08.18 (1948 marriages) with the third one being merely a beautiful date rather than a superstition-related date. At the same time on 13.09.07 the number of marriages was lower than average (42 marriages), suggesting the importance of the sequential order of the numbers comprising the “lucky” triad. Other numerically special dates that were in the top-10 most popular marriage dates were 12.12.12 (1382 marriages), 11.11.11 (1379 marriages), and 08.08.08 (1072 marriages). Birth dates can naturally be targeted to a smaller extent, which is why the top list of most popular dates is not so insightful: 06.08.08, 31.03.10, and 28.07.08 (Figure 2).
The dynamics of the daily number of marriages in Denmark (2007–2019)
The dynamics of the daily number of live births in Denmark (2007–2019).
Key variables recorded in our daily dataset, containing data for 4748 consecutive days are presented in Table 1.
Using the data on daily number of marriages and live births we tested the hypothesis that Danes favor the 7th and the 9th and avoid the 13th and tested the effects of some other special days. Following most previous studies (Reference Almond, Chee, Sviatschi and ZhongAlmond et al., 2015; Reference Huang, Ma, Zhang and ZhaoHuang et al., 2020; Reference Kabátek and RibarKabátek & Ribar, 2018) dependent variables were log-transformed so that regression estimates are interpreted in ratio rather than in interval terms. In the case of marriages, the transformation was especially useful as it decreased its skewness and made the distribution closer to normal (Figure 3).
Density plots: empirical distribution of the log-transformed daily numbers of marriages and live births.
Categorical variables d7, d9, d13, valentines, halloween, apr1, and feb29 were constructed in such a way that they have 3 levels instead of 2 levels commonly used in the case of special event indicators. With this coding scheme the coefficient of the second level (corresponding to the date of the event) will show the difference between the number of marriages or births on the day of interest compared to the average of one day before and one day after the event. This prevents a situation when a researcher uses a single binary variable to account, for instance, for the Halloween’s effect, but then it turns out that the estimate would be the same if instead of a dummy variable for October 31 a dummy variable for October 30 was used. In such a situation the Halloween’s effect could have been attributed not to the Halloween, but to the lower attractiveness of this part of the year, which is otherwise not captured by main effects of months and days of the week.
For each of the dependent variables (log(marriages) and log(births)) we estimated a multivariate regression model, which included the following explanatory variables: fixed effects of year, weekday, holiday, valentines, halloween, apr1, feb29, and special. Factor variables d7, d9, and d13 were included to allow for testing differences between days 7, 9 and 13 against the days right before and right after them. Variable fri13, which is essentially the interaction between a binary indicator of weekday=5 and day=13, has been included as well to test the hypothesis that the 13th is avoided more when it is a Friday.
Even though our dataset allows inferring the effects of official public holidays as well, we prefer to use them merely as control (nuisance) variables and focus on the effects of irrational beliefs, as there are objective reasons for lower number of births on such days. For example, few doctors are ready to work on holidays, which decreases the number of cesarian or induced vaginal labors scheduled to such days.
3 Results
The explanatory power of various date-related predictors is naturally higher for the case of marriages (R2=0.895) than for the case of births (R2=0.645), as the exact date of birth can be planned only to a limited extent (Table 2). Danish couples tend to avoid getting married on the 13th even if it is not on Friday. Other things equal, 25% fewer marriages occur on the 13th (other than Friday) than on the 14th. This negative effect is comparable to the effect of the Halloween (−27%). However, the difference reaches 39% for Friday the 13th.
*** p < 0.001
** p < 0.01
* p < 0.05
There are no extra marriages on the 7th compared to the 6th and the 8th, and the 9th is, on average, less popular than nearby dates. By estimating an additional regression not presented here, we have checked that there is a premium for some round numbers – the 10th compared to the 9th and the 11th and to the 20th compared to the 19th and the 21th, but have not identified the same pattern for the 30th vs. the 29th and the 31st.
On average, 2% fewer children are born on the 13th than on days before and after the 13th (p<0.001). Unlike in the case of marriages, no additional fear associated with the 13th being Friday was identified for the case of childbirths. Contrary to our expectations, the 9th is also systematically around 2% less popular than the nearby dates (p=0.025), but the neighborhood of the round number 10 explains this.
Halloween is avoided both as a wedding date (around 27% fewer people married) and a childbirth date (7.5% fewer newborns). Controlling for other factors, significantly more marriages occurred on April Fool’s day and on February 29 (by 81% and 139% compared to an average nearby day). However, people avoid giving birth to their children on February 29 and are reluctant to April 1. The effect of February 29 (-15%, p<0.001) is stronger than the effect of the 13th. The favorite numerically special dates both for the marriage and for the birth giving were those when the day, the month and the year coincided, as well as the unique date of 07.09.13 associated with the “7–9–13” superstition. Such dates were associated with more than 10 times more marriages and 6–8% more childbirths than we would have expected on such days if they had not been numerically special. Interestingly, in the case of childbirth St. Valentine’s day had about the same positive premium, while Halloween – about the same negative premium as the above mentioned special dates.
4 Conclusion and directions for future research
The research contributes new evidence that Danes tend to be superstitious when choosing their wedding day. They avoid the 13th, especially if it is a Friday. Other things equal, they also prefer to move the birth date from the 13th. We have not found a positive premium neither for the 7th nor for the 9th. In the case of birth timing, the 7th and the 9th are even significantly less popular than the nearby dates, but the effect’s magnitude is small. This can be explained by the importance of the “7–9–13” combination in Danish culture with less emphasis on 7 and 9 being lucky numbers on their own. When put in the context, the effects of numerological superstitions associated with numbers “7”, “9” and “13” individually are small compared to many other effects of salient holidays and special dates. Somewhat surprisingly, February 29 and April 1 are favored as wedding dates, while February 29 is avoided as the birth date and April 1 is neither avoided nor favored, implying that cultural stereotypes are context-dependent. The unique date 07.09.13 and dates where day, month, and year equal one another were associated with dramatic increase in marriages and 6–8% increase in childbirths. Halloween’s negative effect on childbirths was about the same as the Valentine’s Day’s positive effect (around 7%).
Preference towards particular dates of marriage and — to some extent — births (when mothers insist on shifting in birth timing despite medical recommendations) revealed in our study can potentially be monetized through price differentiation to decrease pressure on medical staff on especially popular dates. In a recent discrete choice experiment the average marginal willingness to pay for a spring birth was 877 USD among married US women aged 20–45, which also implied a willingness to trade-off 560 grams of birth weight in the normal range to achieve a spring birth (Reference Clarke, Oreffice and Quintana-DomequeClarke, Oreffice & Quintana-Domeque, 2019). Policy makers should be alert to cultural beliefs associated with delivery to enable informed delivery choices by mothers. Disincentivizing the scheduling of births for non-medical reasons can also decrease adverse health outcomes associated with elective caesarian sections and inductions.
The fact that it was confirmed that people were especially willing to get married and give birth to children on “special” dates makes it interesting to assess long-lasting consequences of such decisions. In addition, in the absence of more detailed statistics, excess childbirths on special dates implies that indicators of being born on such special dates can potentially serve as proxy or instrumental variables correlated with the likelihood that the person was born with the help of a cesarian section or induced labor.
Open access
