Digit preferences in marriage formation in Sweden

Abstract

AbstractBACKGROUNDDigit preferences are normally seen as potentially harmful biases in respondents' reports. Possibly such preferences might also be the cause of some patterns found in Swedish marriage formation, thus affecting actual demographic behavior.OBJECTIVEDigit preferences in marriage formation in Sweden are examined - more specifically, the additional propensity to marry for the first time during the year 2000 or at ages ending with 0 - and their demographic and socioeconomic correlates.METHODSEvent-history analyses are applied to Swedish register data covering 3.5 million men and women in 1991-2007.RESULTSFirst-marriage risks clearly increase for both men and women at exact ages 30, 40, 50, and 60 and in the year 2000. These patterns exist across demographic and socioeconomic groups and are not due to measurement error or random variation.CONCLUSIONSThe timing of marriage is not strictly determined by conventional demographic or socioeconomic factors. Whether the findings are idiosyncratic to contexts like the Swedish, where there are small differences between marriage and cohabitation, remains to be answered.1. Introduction1.1 Digit preferences in data and in real lifeIn the study of demographic behavior, temporality can have important and unexpected influences. Demographic patterns found across various aspects of time, such as calendar time, age, or durations, can be an artifact of the data, methods, or methodology (see Mills 2000 for discussion). For example, digit preferences - preferences for certain types of numbers - normally appear as a form of measurement error found in some data. When respondents themselves report age and timing of demographic events there is a risk that they round off to an even number or a number ending with 5 or 0 (e.g., Budd and Guinnane 1991; Mills 2000). In the register data used in the present study, collected by authorities and agencies, this is not an issue.Mills (2000) stresses that actual demographic behavior can also be influenced by temporality. Some examples are seasonal cycles in weather conditions or social norms about the proper timing of life events. She also states that everyday life is shaped by institutional calendars and even something as arbitrary as the date on a calendar can influence life course behaviour (p.105).In the present study the effects of digit preferences on demographic behavior in Sweden are examined. More specifically, I study the propensity to marry for the first time in the year 2000 or when being of an age that ends with 0, and also if there are any differences across demographic or socioeconomic groups. Andersson (2004) has reported a peak in 2000 for first marriages among women of all parities as well as for second marriages in Sweden. It is not known, however, if the peak was equally pronounced across other demographic, or socio-economic, distinctions for women or for men. Marriage peaks among Swedish women are also visible at exact ages 30 and 40 when previous divorces, parity, and calendar years 1971-1993 are taken into account (Andersson 1998). It is not known whether this pattern also exists for later years when Swedish marriage trends have turned from declining to increasing (Ohlsson-Wijk 2011), or for higher ages when marriage formation generally is not as common. In addition, it would be interesting to know if such patterns exist across various demographic and socio-economic groups and if there are any gender differences.In Sweden, turning an age that ends with 0 is generally associated with larger celebrations and might therefore be seen as specifically suitable to combine with other celebrations, such as weddings. These specific birthdays could also symbolize distinct transitions in life, and trigger considerations of statuses or other transitions in life, such as marriage formation.The peak at the turn of the millennium indicates that this was viewed as a special time to marry. …