Abstract
Divorces have become common, but reliable longer-run historical data with several years of follow-up are scarce. This study investigates divorce trends and divorce risk based on prospective data for all Norwegian different-sex marriages formed from 1886–2018, with yearly follow-up continuing until 60 years after the wedding (N = 2.7 million). First marriages and remarriages are investigated separately, as are marriages in rural and urban areas. The results indicate a general decline in divorce risk, but not in rural areas. The previously suggested impacts of age-at-marriage and remarriage on divorce risk are questioned.
Divorces have become a common experience in many countries. 1 Given this high prevalence of divorce, many studies conducted over several decades have examined patterns of divorce. The main finding is that divorce rates have changed over time. The crude marriage and divorce rates—the number of marriages and divorces a given year per 1000 people in the population—are available for many countries and are known to be a useful measure of divorce rates. 2 These numbers indicate that marriage is more popular in the US than in the EU but that these regions follow the same trends, with an increase in divorce risk from the 1950s and more recently a decline. Indeed, when dividing the crude divorce rate by the crude marriage rate for the US, EU, and Norway, the results are nearly equal (46.4%, 45.5%, and 44.2%, respectively in 2018).3,4 However, updated data on marriages and divorces are scarce in many countries, leaving studies to rely on old data and a small number of data sources. 5 Moreover, Kennedy and Ruggles state that accurate data on divorces are both difficult to find and not always reliable. 6 However, for decades, Norway Statistics has been compiling annual data on every resident in an ongoing survey on a range of issues (e.g., marriage status, residency, and educational level). Thus, analyses of Norwegian marriages, which build on reliable and detailed data, might be valuable when seeking to understand marriage and divorce patterns over time.
This study aims to broaden the understanding of divorce trends and risk by investigating an extended period of time from several perspectives.
The First Research Question Entails: What are the Divorce Rates for Each Cohort of the Total Sample of Married Couples in Norway from 1886 Until 2018 and What are the Trends?
Several factors are theorized to influence divorce trends. Those chosen for this study— increasing age-at-marriage, the proportion of remarriages, urbanization, and the gender revolution—were discussed in the special issue of the Journal of Marriage and Family, A Decade in Review, published 2020.
Remarriage
As many experience divorce, remarriage has become common. Remarriages are associated with a higher risk of divorce, at least in the US 7 and Norway. 8 Indeed, Amato identifies prior marriage as one of nine consistent predictors of divorce. 9 In recent decades, remarriages for one or both partners are more frequent, but the trends suggest that shrinking proportions will remarry in future cohorts. 10 These changes are presumed to affect the divorce trends in the married population. Nevertheless, a lack of previous research on the effects of prior marriage experience on divorce risk means that prospective studies with longer-run historical data including several years of follow-up are scarce. This prospective study includes 37 years of follow-up on remarriages.
The Second Research Question Entails: How do Remarriages Differ from First Marriages Regarding age-at-Marriage, Divorce Risk at Different Years of Follow-up, and Current Trends?
Urbanization
Urbanization has long been a demographic trend worldwide, and Sassler and Lichter posit that it is a common misperception that rural areas represent a “safety deposit box” that stores traditional family values. 11 They indicate that the retreat from marriage and rise of cohabitation have been more rapid in rural than urban areas during the past few decades. However, whether these trends also apply to divorce trends in rural and urban areas has not been verified. This study includes data on marriages formed in rural and urban areas.
The Third Research Question Entails: How do Rural Marriages Differ from Urban Regarding age-at-Marriage, Divorce Risk at Different Years of Follow-up, and Current Trends?
Age-at-Marriage
Several researchers who have examined age-at-marriage determined that it is strongly negatively correlated with divorce risk,12,13 and thus a predictor of divorce. However, none have tested whether it is a predictor of divorce early or late in a marriage. Indeed, age-at-marriage has risen, largely because premarital cohabitation has become the norm. Although most cohabiting unions do not end in marriage,14,15 70% of women who first married between 2010 and 2014 in the US cohabited with their spouses beforehand. 16 Cohabitation before marriage is also the dominant trend in Norway. 17 While contemporary studies confirm the long-known correlation between premarital cohabitation and increased divorce risk over time, 18 some researchers hypothesize that premarital cohabitation ultimately reduces divorce risk in two ways:19,20 (i) by elevating age-at-marriage, which, as mentioned, is known to be a protective factor and (ii) due to a selection effect, 21 i.e., unstable cohabitations tend to dissolve, whereas more stable unions eventually transition into marriage. 22 As cohabitation before marriage has increased in popularity and led to higher age-at-marriage, we should expect to find that as age-at-marriage rises, divorce rates decline at the national level. However, this theory is not empirically tested.
The Fourth Research Question Entails: Is There Strong Evidence That age-at-Marriage is a Predictor of Divorce Early and Late in a Married Life?
The Gender Revolution
The gender revolution, which has enabled women to obtain a higher education, enter the workforce, and strive for equal rights, has also been theorized to contribute to marriage outcomes.23,24 The degree of gender equality has varied over time. According to several theorists,25,26 gender equality may help explain the decline in the number of divorces in recent cohorts. Scandinavian countries are leaders in the transition to gender equality in the areas of values, education, work, and living conditions. 27 From this perspective, the divorce patterns in those countries are especially interesting to investigate, as they might forecast trends for countries less advanced in the gender equality transition. The results from the research questions will be discussed in light of the gender revolution.
The objective of this study is to better understand historic and possible future trends regarding divorce risk in general, with a special focus on age-at-marriage and differences in subgroups such as first versus remarriages and rural versus urban marriages.
The present study
This study aimed to understand divorce trends and risks based on data for 133 cohorts of marriages formed in Norway from 1886 until 2018. During such a long time, several factors might have affected divorce risk, such as the general economy in the nation, labor opportunities, educational level of men and women, world history (e.g., world wars), moral values, and legislation. Since not all readers are familiar with the Norwegian context, nor able to read documents in the Norwegian language, I will provide a brief overview of important aspects in relation to marriages and divorces. During the span of time entailed by this study, Norwegian law on adult relationships changed. In the 19th century, marriage was viewed as an arrangement that lasted for life. However, in 1909 the first Norwegian law defining divorce was legislated. Small changes were added in 1918, but it was not until 1990 that the law of divorce more clearly defined marriage as something that could be dissolved. Modern divorce procedures were introduced, making divorce more available to women. Women's economic obstacles to divorce were addressed through arrangements by state feminism 28 in the Norwegian welfare state which, beginning in the 1980′s, increasingly supported single and divorced mothers. 29 Another important issue was that cohabitation was prohibited in Norway until 1972, but has since become more and more common. As of 2021, 65% of Norwegian adults live in a domestic partnership; 30% of those are cohabitating and the rest are married. However, for those under 30 years of age, cohabitation is dominant. Same-sex unions have been legal in Norway since 1993, first legalized under the law of partnership, which provided same-sex couples legal rights and obligations similar to those of married couples. In 2009, same-sex couples were included in the definition of legal marriage in Norway and received equal rights to different-sex couples. Same-sex marriages only represent 1–2% of the married population and are not included in this study.
The popularity of marriage in Norway is in decline, especially when seen in relation to population growth. During 1981–2018, the population in Norway increased by almost 30%, from 4.1 to 5.3 million, yet the number of marriages remained around 20,000 each year. This decline in the marriage rate is similar to that found in the US 30 and other countries 31 and can be understood by reference to the observed correlation with the increased popularity of cohabitation in Norway.32,33 Long-established cohabitations might now be considered to be the new marriage, even though the legislation on aspects of taxation and inheritance differs substantially between marriage and long-term cohabitation in Norway, as it does in other countries. Norway has the highest proportion of (unmarried) cohabiting couples between the ages of 25 and 29 in Europe. 34
Method
Data
The dataset consists of all marriages formed from 1886 until 2018 in Norway (N = 2,698,632). All marriages were investigated for each successive year to see if the pair was still married. They were identified by their unique identification number in the database of Statistics Norway, a method that made it possible to detect divorces of people that divorced and remarried in the same year, something that would not be detected if only the people individually were controlled for being married each successive year. When investigating divorces, some argue for using reports of separation instead of divorce; 35 however, that state is both reversible and sometimes not reported in the Norwegian database. In this study, separated couples are considered to be married and not divorced. The data used in this present study were attained from two sources. Data on the oldest cohorts (1886 until 1980) stem from a report by Norway Statistics. 36 Data on the cohorts from 1981 until 2018 were attained through special order from Norway Statistics and not from the publicly accessible data published on its website (www.ssb.no). The data are register data and are available for download 37 and described in detail in a connected study. 38 The data for the cohorts from 1981 until 2018 contain more details such as age-at-marriage, whether it is a first marriage (both marrying for the first time) or a remarriage (where one or both had been married before), and whether the marriage was conducted in a rural or urban area. This sample is called the detailed sample hereafter. The size of the population of marriages varies significantly from one year to the next. The period under study started in 1886 with 12,819 marriages. It increased with several peaks and troughs, and the all-time high was reached in 1947 (29,923 marriages). However, the population in Norway has increased every year in this period, resulting in a declining marriage rate (marriages per 1000 inhabitants) since 1947 (9.5). It was at its lowest in 2018 (3.2), the last cohort included in this study.
Analysis
Jasp (version 0.13.1) was used for the descriptive and Bayesian t-tests and regression analyses. R-studio (version 1.3.1073) was used for time series analysis and forecasting. Bayesian analysis was applied because it provides results indicating the strength of the evidence for one hypothesis (e.g., age is not a predictor of divorce) in relation to an alternative hypothesis (e.g., age is a predictor of divorce). The main outcome of interest from the analysis is the Bayes Factor (BF), which quantifies the evidence for each of the tested theories. Schönbrodt and Wagenmakers present a rough heuristic classification where BF = 1 is no evidence, 1–3 and 1/3–1 is anecdotal evidence, 3–10 and 1/10–1/3 is moderate evidence, 10–30 and 1/30–1/10 is strong evidence, 30–100 and 1/100–1/30 is very strong evidence, and a BF >100 and <1/100 is defined as extremely strong evidence. 39
Results
This study investigated marriage and divorce trends and risks by seeking answers to four research questions:
What are the divorce rates for each cohort of the total sample of married couples in Norway from 1886 until 2018 and what are the trends? How do remarriages differ from first marriages regarding age-at-marriage, divorce risk at different years of follow-up, and current trends? How do rural marriages differ from urban regarding age-at-marriage, divorce risk at different years of follow-up, and current trends? Is there strong evidence that age-at-marriage is a predictor of divorce early and late in marriages?
Research Question One – Total Sample
Divorce rate
The highest divorce rate among all 133 cohorts in this study was reached by the 1986 cohort (40.8%) after being married for 32 years. As illustrated by the straight line in Figure 1, no cohort of married couples before or after has surpassed them. Since prospective data with yearly follow-up for several years are rare and the total number of marriages and divorces per year are easily accessible in most countries, I added an analysis based on that register data. The calculations of the divorce rate (i.e., the total number of divorces divided by the total number of marriages) for each year are illustrated by the dotted line in Figure 1. These data are only accessible from 1969 from Statistics Norway and the highest divorce rate using this method of calculation was 57.7%, reached in 1993.
The total divorce rate and yearly divorce rate divided by the yearly marriage rate.
Divorce risk
Almost no marriage ended in divorce during the same year that the marriage occurred. The highest divorce rate among newlyweds was reached in 1959 where 0.06% of those who married that year ended up divorced during the same year. The peak of divorces happened between the 4th and 7th years into the marriage, followed by a successively declining trend, as exemplified with the 1981 cohort in Figure 2. Of the total population of 22,035 marriages, 8613 ended in divorce before the end of the 37th year of follow-up.
Divorces for each year of follow-up for the 1981 cohort.
Comparing recent cohorts with early cohorts, we find a substantial rise in divorces late in marriage. The percentage in the rise from early cohorts to later ones was high, but the number of marriages that ended in divorce after 30 years of marriage was still lower than those who ended their marriage after the first 5 to 10 years, as presented in Table s-1 in the supplemental material. The risk of divorce for the original cohort of married couples differs from the divorce risk for couples who have lived together for several years. For couples who passed the point at which the peak of divorces occurred, the chance of divorce decreased every year. I used survival analysis on the 1981 cohort as an illustration to construct Figure 3. For the total 1981 cohort, the chance of celebrating their silver (25th) anniversary was 64% but for those couples who had already passed the 10th anniversary, the chance was 77%. For those who celebrated their silver anniversary, the chance of celebrating their 37th anniversary (the maximum length of follow-up for that cohort) was 93%, providing both were still alive at that time.
Chance of continuing together: 1981 cohort.
Divorce trends
The total divorce rate was informative for comparing older cohorts; however, as the years of follow-up were different, it was not useful to investigate trends in the data. Older cohorts have 60 years of follow-up and the 2018 group only one. A comparison at fixed time points (i.e., 5 and 10 years after the wedding) was more useful to investigate the differences between cohorts and time trends. Divorces steadily increased from 1886 to 1950 (1.3% to 11.2%, respectively). However, the increase was not as steep as during the period from 1950s until the mid/end of the 1980s. Since that time, there has been a decline in divorce risk. I applied the Holt method of forecasting 40 using the method suggested by Hyndman and Athanasopoulos 41 on the data 5 and 10 years after the wedding predicting the following 10 years. The results with 80% and 95% prediction intervals are displayed in Figure 4. The inspection of the residuals identified white noise (p = .078 for the 5-years-after-wedding series and .209 for the 10-years-after-wedding series), indicating that the model was satisfactory.
Divorce rates 5 and 10 years after the wedding, with forecasts.
Research Question Two – First Marriages and Remarriages
The detailed sample, with data from 1981 until 2018, was used to compare the first-time-married to the remarried. In 1981, it was 19.4% of remarriages in the total married population. The proportion of remarried peaked in 2006 with 30.5% but then leveled out and stayed close to 28% from 2009; see Table s-2 in the supplemental material for more details.
Difference in age
Age-at-marriage increased steadily, but more for the remarried (12.4 for men and 12.2 for women) than for the first-time-married (8.2 and 8.5, respectively). See Table s-3 in the supplemental material for more details. The SD of age also increased for the first-time-married (3.3 for men and 3.1 for women) but not for the remarried in this period. Mean age-at-marriage for the first-time-married was 34.1 (SD = 7.6) for men and 31.8 (SD = 6.9) for women in 2018, while it was 52.8 (SD = 9.7) and 48.2 (SD = 9.9) for the remarried. In 2018, the difference in age between men and women was 2.3 for the first-time-married and 4.5 for the remarried. In 1981, the age difference was 2.6 and 4.3, respectively.
Difference in divorce risk
The divorce rate was higher in the remarried sample for all cohorts consistently through the years of follow-up, with the highest identified in the 1985 cohort (43.2%). Bayesian t-test analysis showed extreme evidence of this difference 5 years (BF10 = 2.108 e + 11), 10 years (BF10 = 8.520 e + 9), 15 years (BF10 = 1.754 e + 10), 20 years (BF10 = 1.157 e + 6), and 25 years (BF10 = 8884.341) into the marriage. To assess the size of the elevated risk of divorce for remarried couples compared with the first-time-married, I used Bayesian A/B analyses on the 1981 cohort, as presented in Table 1. The results were contrasted with the 2013 cohort to identify a possible trend. The results indicated that the relative risk was higher in the early years of marriage, with extreme evidence found for the difference between the two groups (BF10 ranged from 1962.272 to 3.390e + 11). The relative risk 5 years after the wedding was higher in 1981 than in 2013, thus indicating a change over time.
Relative Risk of Divorce for the First-Time-Married and Remarried.
Note: Group 1 = First-time-married, Group 2 = Remarried. CI = Credibility Interval.
Difference in trends
Since the relative risk analysis indicated a trend, I further assessed the possible differences in divorce trends between those marrying for the first time and those remarrying, focusing on the data 5 years after the wedding. As depicted in Figure 5, both groups had an overall decline from 1981 until 2018, but with an increase at the beginning of the time series. The decline in the divorce rate during the last 25 years was 5.5 percentage points for the remarried and 2.1 for the first-time-married. Two models were used to investigate whether the two slopes were different. First, an autoregressive model inspecting the data with a three-year lag obtained BF10 = 5.268 e + 130. Second, a linear regression investigated the trends and obtained BF10 = 2.190 e + 7839. Hence, both these analytical techniques found extreme evidence that the change in the divorce rate in this period was higher for the remarried than the first-time-married.
Divorce rate trend for first marriages and remarriages at the 5-year follow-up.
Research Question Three - Urban and Rural Marriages
The third focus of the analyses was on the differences between urban and rural marriages in the detailed sample using data from 1981–2018. The analysis started by assessing the age differences in and between those samples; it then continued with an investigation of the difference in divorce risk and finally inspected the trends observed in the urban and rural samples.
Difference in age
Age-at-marriage increased in both urban and rural areas, but far more in rural areas. In 1981, men in rural areas were 27.8 years (SD = 7.3) compared with 28.8 years (SD = 7.8) in urban areas. In 2018, rural areas surpassed urban areas in age-at-marriage (39.4; SD = 11.6 vs. 37.3; SD = 10.6). When investigating first marriages separately, I observed the same trend. Mean age-at-first-marriage was 25.9 (SD = 4.5) in 1981 in rural areas and 26.3 (SD = 4.2) in urban areas. In 2018, it was 34.6 (SD = 8.1) for rural areas and 33.7 (SD = 6.9) for urban areas. The difference in age between spouses did not change in the period from 1981 to 2018 and was 2.6 for urban areas and 3.1 for rural.
Difference in divorce risk
The total divorce rate for the urban sample peaked with the 1982 cohort (42.4%) and thereafter declined. The rural sample peaked four years later, with the 1986 cohort (39.4%). When analyzing the difference in divorce rates between the rural and urban sample for all the cohorts together, I found no evidence of a difference. Bayesian t-test analysis also presented no evidence of a difference between the groups 5 years (BF10 = .317), 10 years (BF10 = .508), 15 years (BF10 = .324), and 20 years (BF10 = .684) into the marriage. The 1981 and 2013 cohorts were investigated by Bayesian A/B analyses to assess the relative risk of divorce for rural couples compared with the urban, as presented in Table 2. There was no evidence of different relative risks for the urban and rural areas, but the relative risk 5 years after marriage was lower in 1981 than in 2013, thus indicating a change over time.
Relative Risk of Divorce for Urban and Rural Marriages.
Note: Group 1 = Urban, Group 2 = Rural, CI = Credibility Interval.
Difference in trends
Since the relative risk analysis indicated some change over time, I performed two additional steps in the analysis. First, I investigated only the first 5 cohorts (1981 until 1985) and found strong and very strong evidence that divorce risk between the rural and urban samples was different 5 years (BF10 = 43.419), 10 years (BF10 = 25.658), 15 years (BF10 = 91.598), 20 years (BF10 = 19.572), and 25 years (BF10 = 54.635) into the marriage. In the final step, I investigated the cohorts from 2000 and found extreme evidence that the urban and rural samples were also different in that period, but found the opposite trend to the previous analysis, BF10 = 8868.226 (5-years-after-wedding), 577.184 (10-years-after), and 480.141 (15-years-after).
When investigating the trends, I used the same method as for the total sample, investigating the divorce rate 5 and 10 years after the wedding for cohorts from 1981–2018. Analysis of all cohorts with 10 years follow-up (1981–2008) in the 10th year of the marriage showed that the divorce rate was higher in urban areas until the 1993 cohort. From that cohort and until the last, the divorce rate in the rural areas was higher, as depicted in Figure 6. I identified a downward trend in urban areas (-5.5 percentage points for the 10-years-after-wedding) but an upward trend in rural areas ( + 1.5 percentage points). The differences in trends were bigger in the 10th year of follow-up than in the 5th, indicating that the divorce rate in rural areas is increasingly higher for each successive year of marriage compared with an urban marriage. Using the same two models as used to analyze first and remarriages, I investigated the two slopes to assess whether they were different. Using an autoregressive model to inspect the data with a three-year lag, BF10 equaled 9.788 e + 381. Investigating the trends as a linear regression, BF10 equaled 1.137 e + 1486. Both these figures indicate extreme evidence that the change in the divorce rate in this period was different for marriages formed in urban and rural areas.
Divorce rate trend for urban and rural marriages at the 10-year follow-up.
Research Question Four - Age-at-Marriage as a Predictor of Divorce
I first visually analyzed the correlations between mean age-at-marriage and divorce five, 10, and 15 years after marriage and determined that they do not form a conclusive line; see Figure s-4 in the supplemental material. This finding does not support the generally accepted hypothesis that age is a predictor of divorce,42,43 at least not at those time points. However, to further assess if there was strong evidence that age-at-marriage is a predictor for divorce, I applied Bayesian linear regression with age-at-marriage as a predictor and divorce as the outcome variable for divorces that occurred during the first 25 years of marriage using the data from all cohorts from 1981–2018. There was no strong evidence that age-at-marriage is a predictor of divorce. As visualized in Figure 7, at best, there was moderate evidence that age-at-marriage is a predictor of divorce. Indeed, this was only true for a few select years following marriage. There was moderate evidence from the 5th to 8th years for the first-time-married and from the 5th to 10th years for the total sample. For the first-time-married, the prediction for age-at-marriage revealed BF10 = 3.6 and only explained 17.4% of divorce risk, 5 years after the wedding. Ten years into the marriage, it was inconclusive whether age-at-marriage improved our prediction of divorce (BF10 = 1.2 and explained only 11.7% of the change). For some of the time points, there was even moderate evidence for age-at-marriage not being a predictor (BF10 < 1/3).
Evidence for age-at-marriage as a predictor of divorce.
Discussion
This discussion has two aspects. First, I discuss the results from Norwegian marriages in light of data from other countries. Second, I explore the interpretations of current divorce trends and divorce risk as well as issues of data availability and methods of analysis. Commonly known predictors of elevated divorce risk, such as age-at-marriage and remarriage, are discussed as well as the unexpected differences in divorce risk between rural and urban marriages.
Norwegian Marriages in Light of Marriages in Other Countries
The proportion of remarriages in Norway seems to have stabilized at 28% of all marriages. This differs from the growth in remarriage observed in the US. 44 Congruent with the trend in many other countries,45,46 this study found that age-at-first-marriage has shown a steady increase, with the strongest increase in rural areas. However, mean age-at-first-marriage is about four years higher in Norway than in the US. A possible reason for this discrepancy is that more couples in Norway complete their tertiary studies before marriage than in the US and other Western countries. 47
Understanding Divorce Trends and Divorce Risk
It is insufficient to observe trends and important to investigate what drives those trends to understand the past, forecast future events, and provide effective support for families. Three main perspectives were useful when investigating divorce risk and trends: the total sample, first-time versus remarried, and rural versus urban marriages. To observe trends early in the data, the total divorce rate for each cohort was useful, as it clearly visualized the increasing divorce trend from 1950 until about 2000. However, the divorce trends in cohorts closer to today were only identified when investigating the data from another perspective—years from the wedding. From this perspective, the Norwegian sample identified a similar declining trend as identified in the US, 48 a correlation suggesting that divorce trends are similar in Western countries. The forecast for the coming 10 years is positive for those marrying, indicating that almost 90% of marriages will reach their 10th anniversary and as many as 70% will live together “until death do them part.” The gray divorce revolution identified by Brown and Lin, 49 indicating a trend of divorce in people of higher age, was visible in the data; however, the survival analyses showed that the chance of the marriage of an older couple ending in divorce was still marginal. The finding that the risk of divorce increases rapidly in the first years of marriage and peaks between the 4th and 7th years is in line with most studies.50,51
Caution when using crude divorce rate data
The decline in divorce risk from 2000 is inconsistent with that found when investigating crude marriage and divorce data. 52 This inconsistency confirms the importance of using appropriate data for analyzing divorce trends, as exemplified in this study. Several authors have commented that reliable data on divorce are difficult to find,53,54,55 but having reliable and detailed data is crucial to gain a correct understanding of the trends. Although Amato suggests that data on crude marriage and divorce rates are useful, 56 this study exemplifies the concern of Heaton 57 that they cannot be used to investigate trends effectively (note the discrepancy between the observed divorce rate (straight line) and divorce rate divided by the marriage rate (dotted line) in Figure 1).
Since data following individuals or cohorts of married couples are difficult to obtain in most countries, researchers are often left a) inspecting a small sample or b) calculating the divorce rate using the register of married and divorced individuals for each year. The first leads to low power for the analysis and the latter might lead to faulty conclusions, as illustrated in Figure 1, because the crude divorce rate uses the total number of divorces (or separations) to calculate how many in the population experience divorce (total divorce rate) and the number of marriages and divorces fluctuate from one year to another. Further, it is not the people from the same population used in the numerator and denominator in the calculation, since people who marry in one year are not the same people who divorce the same year. In addition, as exemplified in this study, even if crude divorce rate data manage to detect an actual trend, they cannot place that trend in relation to the right cohorts of married couples. I therefore suggest caution when analyzing trends or calculating divorce rates based on those types of data.
Remarriages are less stable early in married life than later
This study confirmed the internationally recognized trend that remarriages are more prone to end in divorce than first marriages. Moreover, it identified that the decline in the divorce rate for the total population stems from a decline in the divorce rate for first marriages and remarriages. Indeed, the decline in the divorce rate was higher for remarried than for first-time-married individuals. Because this study investigated several successive cohorts, it also found that the differences have declined over time, indicating that the relative risk of divorce among remarriages is not as high as it formerly was. Moreover, since this study followed remarried couples for up to 37 successive years it identified that the increased risk of divorce depends on the timespan investigated. The increased divorce risk for the remarried was found to be the highest when investigated only a few years from the wedding and decreased in the long term. Hence, even if there is an elevated risk of divorce for remarriages compared with first marriages, this is related to longevity, which would not be detected with a shorter follow-up period. This might indicate that the elevated risk of divorce for remarriages identified by several studies with a shorter follow-up is misleading. Indeed, the claim that remarriage is a persistent and significant predictor of divorce is debatable, 58 highlighting the need for prospective studies with longer timeframes for follow-up from other countries. From the lifelong perspective, remarriages are not as prone to divorce as previously thought. 59
Three possible interpretations for the elevation in divorce risk, especially in the early years of marriage, are presented. First, it might be that unstable remarriages tend to dissolve faster than unstable first marriages. One reason for this could be that those who have prior experience of divorce more easily give up and not fight as much or long as in their first marriage for it to last. 60 Second, it could be that remarried couples more quickly recognize a relationship “not worth fighting for”—a relationship that they know will end in divorce anyway. Third, as suggested by several authors, 61 remarriages are more complex (e.g., accommodating children from a first marriage, lower adult well-being, suffering the economic consequences from a divorce) and it is reasonable to assume that problematic issues rise earlier than in first marriages. This could be a reason for the elevated risk of divorce in the first years after the wedding. This interpretation is congruent with the divorce patterns identified among complex remarriages. 62 Over time, several first marriages also develop unhealthy relationships with problematic issues and thus eventually dissolve.
Remarriages as an aspect of the total divorce trend puzzle
One explanation for the upward divorce trend observed for marriages formed between the 1950s and mid-1980s is the increase in remarriages in the population of married people. Although this study confirms that remarriages are more prone to end in divorce than first marriages, it also points out that this is not the only explanation for the observed trend. Indeed, among the first marriages examined in the present study, an increase in the divorce rate was observed during that period. Thus, the stability of first marriages was also challenged during that time. Indeed, first marriages have always been the largest population63,64 and thus the most important factor impacting the total divorce rate.
Regarding theories to understand the more recent decline in the divorce trend, the theory of the second demographic transition framework 65 has become popular among scholars in European countries, North America, Australia and New Zealand, and East Asia. 66 The second demographic transition partly builds on Maslow's theory of shifting needs and explains features such as “the baby bust, the systematic postponement of marriage and parenthood, sub-replacement fertility, the rise of alternative forms of partnerships, and parenthood outside marriage.” 67 The increasing popularity of cohabitation is a part of that transition and, as noted earlier, Norway has the highest proportion of cohabiting couples between the ages of 25 and 29 in Europe. 68 This may be a key factor contributing to the lower divorce rates in the later years of this study. However, a similar decreasing divorce trend is also observed in the US, where the proportion who cohabit is lower than that in any other country on the entire North and South American continents and lower than in Western European countries 69 and thus differs from Norway. That these two countries with such divergence on cohabitation show similar divorce trends might indicate that cohabitation does not have as much impact as suggested. Indeed, Zahl-Olsen found no support for the theory that the increased popularity of cohabitation affects the divorce rate. 70
Lesthaeghe writes that the best-educated are more prone to convert their cohabitation union into marriage and suggests that this is a sorting effect. 71 Thus, marriage is viewed as a more popular lifestyle choice for individuals in a higher financial bracket. Since education is free in Norway, the proportion of highly educated people from all social classes is high, though skewed toward those with highly educated parents and those living in urban areas. This interpretation suggests that the data included in this study might be similar to the college-educated population in the US, which has an identified decrease in divorce risk during the last decade. 72 However, the differences in financial resources, labor, and educational opportunities between classes are not as large in Norway as in the US.
Another related reason for the identified drop in the divorce rate is connected to the high-investment parenting style that requires the time, earnings, and energy of two parents to be achieved successfully. 73 This theory could apply to the sample in this study, which investigates marriages in a country with a high educational level, low unemployment, numerous two-income families, and a high standard of living. 74
The trend of a decreased divorce rate over the last two decades could also be understood from the gender equality phase theory. According to the arguments of Esping-Andersen and Billari as well as Goldscheider, Bernhardt, and Lappegård,75,76 a country that has progressed toward an egalitarian society should observe decreased levels of divorce. Norway is one of the world's leading countries in gender equality, which assumes a decrease in relational conflicts. This could thus explain the two-decade-long decline in the divorce rate.
Another aspect to consider to understand the decline is that many younger cohorts have experienced their parents divorcing. This has long been identified as a risk factor for divorce. 77 Even though Amato and DeBoer find that those with experience of parental divorce enter their marriages with lower expectations for a long-lasting relationship, 78 this is not the case for everyone. Indeed, Whitton et al. find that women's (but not men's) parental divorce is associated with lower relationship commitment and lower relationship confidence. 79 This indicates that parental divorce does not affect all people equally negatively. As many have experienced parental divorce, it might even be that the trend has shifted so that people with those experiences, who choose to marry, are more committed to protecting their children from the hardship they suffered. Thus, the children of the divorce generation decrease divorce risk. However, this theory has yet to be empirically verified.
Urban and rural marriages
Lyngstad finds no difference in divorce risk between urban and rural areas in Norway, 80 while Kulu and Boyle show that separation rates are the highest in urban areas in Austria. 81 The present study found a lower divorce rate in rural areas than urban areas in earlier cohorts, but this has changed over time. The risk of divorce is now higher in rural than in urban areas. Living in a rural area in Norway was previously a protective factor against divorce, but that has shifted so that living in a city is now protective. This result supports the suggestion by Sassler and Lichter that living in a rural area is not a protective factor against changes in marriage or divorce patterns. 82 I suggest three explanations for this trend.
First, rural areas lag behind the changes observed in urban areas by only a few years. When the divorce rate started to increase in urban areas, rural areas experienced the same trend a few years later. One explanation is therefore that the divorce risk will eventually fall in rural areas, as it has in urban areas, and thus follow the urban trend. Higher divorce risk observed in rural marriages is, from that perspective, only a matter of lag.
Second, between 1950 and 2018, the world's population was urbanizing rapidly, with Europe and Northern America in the lead. 83 In Norway, this trend is especially strong among women and more women than men have gained higher education. 84 This factor might be related to the divorce trend in rural areas. Women might seek higher education, more liberal values, better job opportunities, and more cultural facilities in cities, leaving men with possibly more traditional values behind in rural areas. 85 Although 90% of the Norwegian population already lives in an urban area, this trend has not slowed, indicating that the rise in divorce risk in urban areas might also continue in the years to come. In addition, in Norway, more women than men relocate from rural to urban areas, 86 which results in an underrepresentation of young women in rural areas. Hence, it is reasonable to argue that men in rural areas face difficulties in finding a suitable marriage partner, in relation to age and social and cultural sameness. This theory gains some support from the fact that age difference was higher in rural areas compared to urban.
Third, the lack of accessibility of family agencies in rural areas might be an explanation. The Norwegian family agency has been restructured 87 to locate a clinic within a two-hour drive of every family, but this has not yet been achieved. Even if it had been, it still takes almost a day to undertake a therapy session in rural areas. Urban Norwegian couples struggling in their relationship can sign up for a couple-enhancement program or contact the free service provided through the family agency and receive couples therapy, while those living in rural areas do not have the same opportunity. This implies that it is important to provide professional help to people in rural areas (e.g., through online services and mobile agencies).
Age as a predictor of divorce
Several authors have identified a strong correlation between age-at-marriage and divorce risk, with couples who marry very young being at the greatest risk.88,89,90 As age-at-marriage increases, we would expect lower divorce rates, 91 but during the period of increase in divorces in Norway (e.g., between 1950 and 1990), age-at-marriage increased each year. This inconsistency may relate to the increased proportion of couples who remarried in that period, a population with a higher age and a higher risk of divorce. However, this study showed that age-at-marriage has increased for the first-time-married as well. Since cohabitation became an alternative in Norway in 1972, the choice to marry has diminished. Thus, one would expect divorce rates to decline as cohabitation has increasingly become popular. However, the divorce rate in Norway continued to rise for another 25 years before it plateaued and began to decline, contradicting that theory. Indeed, the analyses does not seem to support the theory that cohabitation decreases the divorce rate by elevating the age-at-marriage 92,93 because I did not find strong support for age-at-marriage being a predictor of divorce, except for a few select years into a married life. Indeed, other factors might impact divorce directly and indirectly through increased age. One such factor could be what Bystrov defined as emancipative values, which were found to be the strongest predictor of the postponement of marriage in the 45 countries investigated. 94 Kendal and Raymond describe different pathways to change in people's values, observing that people's actions change as values change. 95 People with higher educational levels marry later than those with lower educational levels and have a lower risk of divorce. Level of education and changes in values are found to be strongly correlated. 96 Thus, I suggest that different values are a significant force behind what has been identified as the age effect on divorce risk. This could also partly explain why these results indicate an increased divorce risk and a simultaneous increase in age-at-marriage in rural areas.
Implications
The finding that divorce risk is declining for both first marriages and remarriages brings hope for those who choose to marry and to those working in the field of marriage relations. For stakeholders that finance treatment programs and agencies, the results suggest that help should focus on marriages formed in rural areas, where the highest divorce risk is found. For researchers, this study adds to the knowledge of marriage and divorce trends in Western countries by comparing 133 cohorts of married couples, supported by as much as 60 years of follow-up. Indeed, it expands the knowledge of which countries exhibit a declining divorce trend and how these trends are different in subsamples such as first marriages and remarriages and rural and urban areas. It also adds to our knowledge of what types of data are needed and what analyses can be carried out to identify trends. Finally, this study presents theories useful to understand the trends in the data.
Conclusion
This study confirms the international trend that marriage is becoming a less popular lifestyle choice for adult relationships and that the marriages of those who choose to marry have become more stable. Remarriages are still more prone to divorce than first marriages, but the increased divorce risk is not as influential now as it was 10, 20, and 30 years ago. Furthermore, the study also identifies that the increased divorce risk observed for remarriages compared with first marriages declines when years of follow-up are increased. Further, rural areas were protective to marriages formed early in the time series, but the trend has reversed: marriages formed in rural areas are now more prone to end in divorce than marriages formed in urban areas. The trend is still rising in rural areas.
Importantly, this study questions the validity or strength of two of the persistent predictors of elevated divorce risk: remarriage and age-at-marriage, and explains the underlying factors for these occurrences based on elements from the second demographic transition. It also questions the usefulness of crude divorce rate data when investigating trends.
Supplemental Material
sj-docx-1-jfh-10.1177_03631990221077008 - Supplemental material for Understanding Divorce Trends and Risks: The Case of Norway 1886–2018
Supplemental material, sj-docx-1-jfh-10.1177_03631990221077008 for Understanding Divorce Trends and Risks: The Case of Norway 1886–2018 by Rune Zahl-Olsen in Journal of Family History
Footnotes
Strengths and Possible Limitations
A strength of this study is that it includes data for the total population of married couples in Norway, including 133 cohorts of married different-sex couples, supported by annual individual follow-up for up to 60 years. Further, it differentiates between first marriages and remarriages as well as between marriages formed in rural and urban areas. The fact that the study only includes data from one country and that aggregated data lacking information for confounding factors are used for the analyses are limitations.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
Author Biography
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
