The Correlation between Unemployment and Divorce Rates in the United States - Term Paper Example

?Running head: DIVORCE AND UNEMPLOYMENT The Correlation between Unemployment and Divorce Rates in a Population Affiliation Divorce and unemployment are two phenomena that most people would like to avoid, at least when they are not planning for them. However, bad economic conditions sometimes making unemployment inevitable. The effect of unemployment on one’s home life may be intuitive, but its relationship to one’s married relationships is not always clear. For that reason, the present study examines the correlation between unemployment and divorce rates. Without speculating on the directionality or causal relationships between the two variables, it is important to note that since 1980 higher unemployment has been significantly correlated with higher divorce rates. Conclusions and recommendations are drawn to guide future research in this area. Table of Contents Abstract…………………………………………………………………………..…2 Table of Contents……………………………………………………………….…..3 Introduction………………………………………………………………………...4 Presentation of Raw Data and Discussion…………………………..……………..5 Presentation of Data in Arrayed form with Discussion………………….………...6 Descriptive Statistics with Discussion for Each Field of Data………………….…7 Grouped Data for Each Field and Discussion……………………………………..8 Graphs of Grouped Data and Discussions…………………………………….…...10 Graph of the Two Fields with Discussion…………………………………………11 Correlation of Two Variables with Discussion……………………………………12 Regression Equation with Discussion……………………………………………..13 Conclusions Regarding the Original Thesis……………………………………….14 Critique and Future Directions…………………………………………………….15 Appendix A (References)………………………………………………………….16 The Correlation between Unemployment and Divorce Rates in the United States A tendency to emphasize the social and economic costs of unemployment on society in the aggregate avoids the clear and definite problems that joblessness creates in the home. Some of these household challenges are responsible for lowering levels of subjective well-being in society, which exacerbates the negative overall effects of unemployment. One particularly troubling hypothesis is that divorce tends to increase to some degree in proportion to unemployment rates in developed countries (Jensen & Smith, 1990). The marriage destabilization caused by the loss of a job and the perhaps long-term unemployment that results may explain a great number of divorces. However, especially as one deals with aggregate population data from past years, one is not dealing with causal inferences but rather correlational observations. While it is intuitive to speculate that unemployment increases risk to divorce, one could wonder whether divorce is likely to increase unemployment. It does not seem out of the question that marital instability increases one’s chances of being dismissed or issuing a resignation from his or her work. In fact, Rogers and May (2003) found a significant correlation between increases in marital discord (defined as thoughts or actions supportive of divorce) and declines in job satisfaction. Finding a statistical correlation between unemployment and divorce would signify a number of things, namely that individuals facing long-term unemployment but are happy in their marriages should take steps to ensure the sustainability of that relationship through financial hardships. At a broader level, predictors of unemployment can also be taken as predictors of higher divorces, which give society a chance to plan for increased levels of marital instability in addition to unemployment risk. We hypothesize that there is a correlation between unemployment rates and divorce rates through time. There are 29 values in each category, reflecting the measurement of unemployment and divorce rate over a period of 29 years from 1980 to 2009. The raw divorce rates and unemployment statistics from 1980 to 2009 are presented based on data from the National Vital Statistics Reports of the CDC and the U.S. Bureau of Labor Statistics respectively. Unemployment percentages are calculated based on percentages of the national population (seasonally adjusted) looking for gainful employment. Divorce rates are based on the number of people newly divorced in samples of 1,000 married couples. Year Unemployment Rate Divorce Rate 1980 6.3% 5.2 1981 7.5% 5.3 1982 8.9% 5.1 1983 10.4% 5.0 1984 8.0% 5.0 1985 7.3% 5.0 1986 7.2% 4.9 1987 6.6% 4.8 1988 5.7% 4.8 1989 5.2% 4.7 1990 5.4% 4.7 1991 6.4% 4.7 1992 7.3% 4.8 1993 7.3% 4.6 1994 6.6% 4.6 1995 5.6% 4.4 1996 5.6% 4.3 1997 5.3% 4.3 1998 4.6% 4.2 1999 4.3% 4.1 2000 4.0% 4.2 2001 4.2% 4.0 2002 5.7% 3.9 2003 5.8% 3.8 2004 5.7% 3.7 2005 5.3% 3.6 2006 4.7% 3.7 2007 4.6% 3.6 2008 5.0% 3.5 2009 7.8% 3.4 Table 1. Raw Unemployment (seasonally adjusted) and Divorce Rates, U.S. 1980-2009 Raw data is not explanatorily useful for pointing to clear trends or describing what is happening in reality. All one can really take away from looking at the raw data set is that there is considerably more variation in the unemployment rate than there is in the divorce rate from year to year. The difference in variability may reflect volatility of financial markets and the lack of volatility in how societies treat the institution of marriage through a relatively short period. One in which these data sets can be satisfactorily compared is by taking a “marriage market” perspective, which is introduced by Stevenson and Wolfers (2007). The degree to which laborers in the job market are able to satisfactorily find job openings is a determining factor in how long one stays unemployed. In addition, a great person-job fit makes it less likely that a divorce or separation will occur. Year Unemployment Rate Divorce Rate 2000 4.00% 4.2 2001 4.20% 4 1999 4.30% 4.1 1998 4.60% 4.2 2007 4.60% 3.6 2006 4.70% 3.7 2008 5.00% 3.5 1989 5.20% 4.7 1997 5.30% 4.3 2005 5.30% 3.6 1990 5.40% 4.7 1995 5.60% 4.4 1996 5.60% 4.3 1988 5.70% 4.8 2002 5.70% 3.9 2004 5.70% 3.7 2003 5.80% 3.8 1980 6.30% 5.2 1991 6.40% 4.7 1987 6.60% 4.8 1994 6.60% 4.4 1986 7.20% 4.9 1985 7.30% 5 1992 7.30% 4.8 1993 7.30% 4.6 1981 7.50% 5.3 2009 7.80% 3.4 1984 8.00% 5 1982 8.90% 5.1 1983 10.40% 5 Table 2. Arrayed Unemployment (seasonally adjusted) and Divorce Rates, U.S. 1980-2009 By arraying the raw data in ascending order based on year, one is able to gain insights into trends resulting from divorce rates and unemployment rates respectively. As was previously discussed, the unemployment rate shows considerably more variation than the divorce rate, as well as less of a definite trend. On the basis of this data, one can begin to form statistical analysis to determine whether there is a significant correlation between the two variables from the period of 1980 to 2009. Our hypothesis deals with the existence of a correlation between unemployment and divorce rates over a length of time (in this case, 29 years). Therefore, it is necessary to pursue correlational analysis of the data collected. In this situation, the most appropriate analytic tool is the R-square, which is used to predict future outcomes on the basis of existing information. The existing information, in our case, is the supposedly unrelated data that exists for divorce and unemployment rates. The future outcome is whether a change in unemployment will affect a change in divorce rates, and whether a change in divorce rates will affect a change in unemployment in the aggregate. In other words, once we have developed a model—or the alternative hypothesis—the r-square measure will give an indication of direction (positive or negative) and strength of that model. “Goodness of fit” is another word for how well the model fits the data points being analyzed. If a model is not high in goodness of fit, the predicted values do not align with the values actually obtained. In the present case, if the values expected by the alternative hypothesis proposed above do not align with the values obtained through empirical measurements, then the model is not a satisfactory reflection of reality. The R-square measurement will indicate how much of the variance is explained versus how much of the variance is unexplained. The goal is to explain all (or at least most of) the variance present in a given context, although this is not often possible. In a case such as the relationship between unemployment rates and divorce rates, a satisfactory explanation of variance could exist below 25%, given the complexity of the context in which measurements are being taken. For example, many mediating variables may be at work pushing unemployment down in accordance with divorce rates. Some of these potential problems are highlighted in the critique section on page 14. For now, however, the correlations established between these two sets of data should be sufficient for drawing conclusions about the two variables. Year Range Average Unemployment Rate Average Divorce Rate 1980-1984 8.22% % 5.12 1985-1989 6.40% 4.84 1990-1994 6.60% 4.68 1995-1999 5.08% 4.26 2000-2004 5.08% 3.92 2005-2009 5.48% 3.56 Table 3. Grouped Unemployment (seasonally adjusted) and Divorce Rates, U.S. 1980-2009 When the arrayed data set is grouped together based on yearly ranges, not individual years with their respective unemployment and divorce rates, trends become clearer. The reason for this is purely pragmatic: instead of being confronted with three decades of ordered data, averaged data from six equal data ranges is easier to process to see connections. As one may have noticed by looking at the individual yearly divorce rates in the United States, there is a definite downward trend starting from 1980 and proceeding to 2009. Unemployment, in contrast, was expectedly quite variable throughout that time. However, once these rates are reduced into averages, their means reveal a very general tendency to follow each other. Yet, without statistical analysis of these averages, no firm statistical correlation can be established. Another fact from these time ranges is that they do not distinguish between significant events in American history (particularly economic events) in the past three decades. In the early 2000s, a worldwide recession occurred that increased unemployment rates from 4.0% in 2000 to 5.8% in 2003 (Table 2). Table 3 aggregates this event into the first half of the 2000s, erasing the single-order effect of recessionary events on the labor market. Longer recessions, such as the downturn that occurred from 1975 to 1984, are reflected in their respective time ranges. For instance, the 1980 to 1984 time range has the highest unemployment of any 5-year period since (8.22%). In support of a face-level correlation, this high unemployment in the 1980s corresponds to the highest level of average divorce rates included in the three decades. From the three figures above, we have a graphical representation of the general thrust of Tables 2 and 3, which indicated a general correlation between unemployment rates and divorce rates in the United States from 1980 to 2009. Although there was no statistical significance to this face conclusion, it still seemed clear that both categories were steadily decreasing together through time. On Page 7, however, it became clear that there was indeed a statistical correlation between the two variables such that 32% of the variance is accounted for by the relationship between unemployment and divorce rates from 1980 to 2009. Putting together data in terms of year ranges not only removes year-to-year variation in unemployment rate, but it also provides a clear picture of general trends through extended periods. Comparing the 1980 to 1984 period to the 2005 to 2009 period, for instance, reveals the same approximate distance between unemployment and divorce rates when on the same scale. After performing a regression analysis with the arrayed data given above, the coefficient of determination that was returned was 0.32, meaning that 32% of the variance is explained by the variables being studied here. The r-squared value obtained here approximates how well the regression line follows the real data points on a graph of the data obtained regarding the unemployment and divorce rates at specific years. With a 0.32 r-squared value, there remains 68% of the variance unexplained by the relationship between these two variables, which is not very surprising given the high-level concepts that this study is dealing with. Unemployment and divorce are highly situated phenomena that occur in very complex social arrangements; accordingly, it is inevitable that they will co-vary with other similarly complexly embedded factors determining how unemployment and divorce are related over time. While a 0.32 value may seem small, it does achieve statistical significance. Our tabulated F value is 13.600, but the critical significance F is equal to 9.637. Since the critical significance is less than the calculated F, this enables us to reject the null hypothesis (that unemployment rate and divorce rates are unrelated), and to accept the alternative notion that a relationship exists between the two variables, which based on our r-squared value indicates a positive relationship. As unemployment goes up, divorce rates follow relatively close behind, and as divorce rates go up, unemployment follows. Two significant words of caution, however, need to be said about that result. First, it is not clear which variable is leading the other, if indeed either is leading the other. Second, although unemployment would theoretically follow divorce rates if divorce rates were going upward, what we see in reality is that both numbers have been going down since 1980. That means while both may follow downward, it remains to be seen whether the same relationship would exist upward. The formula for this analysis is: t In this formula, Y is equal to the divorce rate at time t, ? is a constant, and ? is a regression coefficient. This formula is adapted from Amato and Beattie (2011), which examined state-level data for the correlation between unemployment and divorce at the level of states. (The present analysis does not incorporate state-level data, and instead relies on federal statistics.) This formula has the advantage of being simple and reproducible across a number of years, producing a numeric result that expresses trend information throughout the period under review (1980 to 2009). The statistical result of this equation was a value of 0.32, which is the R-square value for how much variance is explained at time t. It should be noted that this analysis relied on the reasonable estimates of its data sources, namely the CDC and the U.S. Bureau of Labor Statistics. Amato and Beattie (2011) calculated their divorce rate with an equation that accounted for 95% of the variance in state divorce rates (R-squared adjusted). The accuracy of ? in this case depends on how much of the variance that the CDC calculation of divorce rate explains, and whether this is similar (or the same as) the Amato and Beattie (2011) calculation. The calculations must be fundamentally similar insofar as the CDC data takes into account state-level reporting of divorce rates, which are then summed into a federal average. However, as the CDC reports in its “Technical Notes” to its preliminary 2009 report on divorce rate data, “NCHS has ceased publishing divorce counts and rates based on provisional data from the combined 50 states because it is no longer statistically feasible to calculate estimates of monthly divorce counts for non-reporting states” (Tejada-Vera & Sutton, 2010, p. 2). Instead, the report publishes only 12-month data on which the present statistical analysis is based. With regard to the original hypothesis, that there is a link between unemployment rates and divorce rates, the statistics seem to indicate a significant correlation between the two data sets. Supportive of the intuitions behind this hypothesis is the intuition that marriage partners desire the other to fulfill his or her role in that institution. This perspective is supported in recent literature that points to a greater likelihood of divorce when the husband is unemployed than when the wife is unemployed, which supports an idea of gender norms as a factor in this relationship (Sayer, England, Allison, & Kangas, 2011). In addition, one might intuitively expect negative personality traits that are counterproductive in a marriage relationship to lead to negative workplace interactions, which may play a role in the relationship between the two variables (Amato & Beattie, 2011). Although intuitive and likely correct in some cases, this perspective does not explain the three-decade trend in the data seen in institutional reports on unemployment and divorce. As was noted previously, both divorce and unemployment rates have fallen steadily from 1980 to 2009. It is not clear if the reverse is true, so additional correlational studies could access data from smaller time ranges where unemployment rose to see if similarly isolated increases in divorce rates are also present. A reliable theory for why unemployment increases might be associated with divorce rate increases (and similarly with decreases) is the so-called “psychological stress perspective,” which highlights the debilitating role that stress has on a marital relationship. The loss of a job is a noted stressor that presumably follows divorce rate as a leading indicator or, in other words, a definite cause for the erosion of marriages. This idea is especially tied to the gender roles observation made previously that low earnings from the male member of a marital arrangement or abject unemployment is likely to be a factor in the eventual divorce of the couple (Lewin, 2005). In terms of limitations to this analysis, it does not provide much proof to speculations about why divorce and unemployment might be related insofar as a correlation only permits claims about associations—not causality. For that reason, the only purpose of this kind of analysis is to refute theories such as the “cost of divorce perspective” that says unemployment and divorce are negatively correlated because during unemployment, divorce is too much of an expense for individuals to deal with (Amato & Beattie, 2011). However, if divorce and unemployment are reliably correlated, then it does not seem to be the case that people will put off divorce for better financial times. A second limitation to this analysis is that the reliability of its results (outputs) is a product of the reliability of the data being analyzed (inputs). Because the data was collected and shared by an agency that did not fully disclose its methodology for calculating the divorce rate at a federal level, it is perhaps too risky to make evaluative conclusions from the results of this study. In addition, this agency has recently changed its methodology for collecting federal divorce data, which poses a challenge to analyzing trends from 1980 to 2009, as this study promises to do. Even if we permit the reliability of the divorce rate data, there is still the unanswered question of whether 32% accounting for variance by the R-squared value is enough to generalize these conclusions across national contexts. The United States may be quite different from countries like Denmark in terms of gender roles in marriage and work-home life interactions. Lastly, a word should be said about future research directions. Before Amato and Beattie (2011), the last study to collect aggregate data to analyze the relationship between divorce and unemployment was published in 1985. The lack of an extensive literature on this subject poses a problem for diagnosing high divorce rates in the aggregate. Future research could develop and refine how we measure divorce at this aggregate level so accurate analyses can be conducted. References Amato, P., & Beattie, B. (2011). Does the unemployment rate affect the divorce rate? An analysis of state data 1960–2005. Social Science Research, 40, 705-715. Jensen, P., & Smith, N. (1990). Unemployment and marital dissolution. Journal of Population Economics, 3, 215-229. Lewin, A. (2005). The effect of economic stability on family stability among welfare recipients. Evaluation Review, 29, 223-240. Rogers, S., & May, D. (2003). Spillover between marital quality and job satisfaction: Long-term patterns and gender differences. Journal of Marriage and Family, 65, 482-495. Sayer, L., England, P., Allison, P., & Kangas, N. (2011). She left, he left: How employment and satisfaction affect women's and men's decisions to leave marriages. American Journal of Sociology, 116, 1982-2018. Stevenson, B., & Wolfers, J. (2007). Marriage and divorce: Changes and their driving forces. Cambridge, MA: National Bureau of Economic Research. Tejada-Vera, B., & Sutton, P. (2010). Births, marriages, divorces, and deaths: Provisional data for 2009. Hyattsville, MD: National Vital Statistics Reports, 58. Read More