Applied Economics Letters, 2014 Vol. 21, No. 8, 528–532, http://dx.doi.org/10.1080/13504851.2013.872753

Marriage, divorce and economic activity in the US: 1960–2008 Hamid Baghestania,* and Michael Malcolmb a

Department of Economics, School of Business and Management, American University of Sharjah, Sharjah, 26666 UAE b Department of Economics, West Chester University of Pennsylvania, West Chester, PA 19383, USA

We utilize a time-series model to examine the interrelationship between marriage and divorce and their connections with macroeconomic conditions for the period 1960 to 2008. Our findings suggest that marriage and divorce are pro-cyclical, although macroeconomic conditions affect divorce only when the economy is underperforming. Marriage is pro-cyclical in all circumstances. Further, bidirectional causation exists, with marriage (divorce) affected by lagged rates of divorce (marriage). Keywords: demographics; marital formation; GDP gap; unemployment gap; causality JEL Classification: J12; C32

I. Introduction Individual choices related to household and family structure are known to be associated with aggregate economic conditions. Becker (1981) observed that voluntary marriages must raise the expected utility of both spouses above what it would have been if they had remained single. Economies of scale, income gains from specialization and the prospect of investing in high-quality children are among the most important quantifiable welfare improvements, and all are affected by macroeconomic conditions. Conversely, divorces occur when one spouse determines that his or her utility is higher outside marriage, and this comparison is also affected by the economy at large. Most existing research estimates the magnitude of these effects by using household-level data. Time-series studies that relate marriage and divorce to macroeconomic aggregates are sparse. We specify a time-series model to better understand the interrelationship between marriage and divorce and their connections with

macroeconomic conditions for the period 1960–2008. We use both GDP and unemployment gaps, which we argue are more informative indicators than the unemployment rate commonly used in the literature. We find that marriage displays pro-cyclical behaviour, with the responsiveness to the unemployment gap being larger when the economy is underperforming. Divorce is procyclical only when the economy is underperforming; widening of the gap inversely affects divorce. Finally, the relationship between marriage and divorce exhibits bidirectional causality. Section II presents a literature review. Section III reports the empirical results. Section IV interprets the results.

II. Related Literature While the economic climate can affect incentives to marry and divorce, the direction of these effects is theoretically ambiguous. Ekert-Jaffe and Solaz (2001) and GuitierrezDomenech (2008) find that the formation of initial

*Corresponding author. E-mail: [email protected]

528

© 2014 Taylor & Francis

Marriage, divorce and economic activity

529 III. Data and Empirical Results

marriages is positively associated with good economic circumstances. Jensen and Smith (1990) and Eliason (2012) find that bad economic circumstances leading to job loss induce divorce. As such, these studies suggest that improvements in economic conditions increase marriages and reduce divorce. However, bad economic circumstances can elevate the importance of economies of scale from cost-sharing in multi-person households. Harknett and Schneider (2012) find that economic distress reduces the probability of divorce. Shore (2010) argues that the risk-sharing benefits of marriage are more important during bad economic times. To the extent that cyclical fluctuations affect men and women differently, a change in economic conditions can disrupt bargaining power within marriages, which has implications for the divorce rate (Kesselring and Bremmer, 2006). But the direction is uncertain. Clark and Summers (1981) argue that cyclical fluctuations affect women more strongly than men in all age groups. Elsby et al. (2010) find the opposite with respect to the recent recession. Further, in newer theoretical models, the marriage– divorce relationship features bidirectional causality. Evaluation of the utility from marrying versus remaining single depends on the probability of divorce, which in turn depends on how fluid the remarriage market is. Each feeds back to the other, possibly with a time lag (Chiappori et al., 2008). Among previous time-series studies, South (1985) finds that increases in unemployment are associated with higher divorce rates, while Schaller (2013) finds that both marriage and divorce are pro-cyclical with respect to employment outcomes. There are a few important innovations in our article. First, we find that the level of responsiveness is different when the economy is overperforming versus underperforming. Second, our model allows for bidirectional causality between marriage and divorce. Third, we find that the GDP gap and the unemployment gap are more informative indicators than the unemployment rate.

The annual data on the total marriages and divorces per 1000 population come from various issues of the National Vital Statistics Report, published by the National Center for Health Statistics up to 2008. The unemployment gap is the actual (civilian) unemployment rate minus the natural unemployment rate. The real GDP gap is 100 times the difference between the logarithm of actual GDP and the logarithm of potential GDP. The data on economic variables are available on the Federal Reserve Bank of St. Louis website. Both marriage and divorce rates (plotted in Fig. 1 for the period 1960 to 2008) display strong quadratic trends. According to our unit root test results, the de-trended marriage and divorce series, the unemployment rate, unemployment gap and real GDP gap are all stationary around a constant. As such, we specify the following general model: M t ¼ a1 þ

7 X

b1i Mti þ

i¼1

7 X

b2j Vtj

j¼0

þ b3 D1 jUtgap j þ b4 ð1  D1 ÞjUtgap j þ b5 D2 jYtgap j þ b6 ð1  D2 ÞjYtgap j þ u1t V t ¼ a2 þ

7 X

c1i Vti þ

i¼1

7 X

c2j Mtj

j¼0

þ c3 D1 jUtgap j þ c4 ð1  D1 ÞjUtgap j

where Mt and Vt are, respectively, the de-trended marriage and divorce rates. jUtgap j is the unemployment gap (in absolute value), and D1 is a dummy (=1 when Utgap > 0, and = 0 when Utgap < 0). jYtgap j is the GDP gap (in absolute value), and D2 is a dummy (=1 when Ytgap < 0, and = 0 when Ytgap > 0). Note that D1 jUtgap j and D2 jYtgap j correspond to the period when the economy is underperforming, and ð1  D1 ÞjUtgap j and ð1  D2 ÞjYtgap j correspond to the period when the economy is overperforming.

Marriage rate 5.0 10

4.5 4.0

9

3.5 8

Divorce rate

3.0 2.5

7

–2.0 6 1960

–1.5 1965

1970

1975

1980

1985

Fig. 1.

1990

1995

2000

2005

(2)

þ c5 D2 jYtgap j þ c6 ð1  D2 ÞjYtgap j þ u2t

5.5

11

(1)

1960

1965

1970

1975

1980

1985

Time plots of marriage and divorce rates: 1960–2008

1990

1995

2000

2005

H. Baghestani and M. Malcolm

530 Equation 1 is initially estimated using OLS for 1960– 2008. Excluding the highly insignificant variables, the equation is then re-estimated, with the results reported in column 1 of Table 1. As seen, the parameter estimates on both D2 jYtgap j and ð1  D2 ÞjYtgap j are not significantly different from zero. Excluding these variables, we have re-estimated the equation with the results reported in column 2. These estimates pass a series of diagnostic tests. For instance, the Ljung–Box test and the White Table 1. Marriage rate equation estimates (dependent variable = Mt) OLS Independent variables

(1)

(2)

Constant Mt – 1 Mt – 2 Mt – 6 Vt – 3 Vt – 6 D1|Utgap| (1 – D1)|Utgap| D2|Ytgap| (1 – D2)|Ytgap| Adjusted R2 Jung–Box test p-value White test p-value

0.107 (1.94) 0.436 (2.81) 0.401 (2.47) –0.483 (4.17) 0.731 (3.90) –0.154 (1.15) –0.236 (3.50) 0.175 (1.33) 0.028 (0.51) 0.026 (0.46) 0.725 0.218 0.471

0.112 (2.09) 0.470 (3.25) 0.380 (2.44) –0.469 (4.22) 0.714 (2.70) –0.155 (1.21) –0.187 (3.75) 0.118 (1.98) 0.736 0.206 0.279

Notes: Mt and Vt are, respectively, the de-trended marriage rate and divorce rate. jUtgap j is the unemployment gap (in absolute value), and D1 is a dummy (=1 when Utgap > 0, and = 0 when Utgap < 0). |Ygapt| is the GDP gap (in absolute value), and D2 is a dummy (=1 when Ytgap < 0, and = 0 when Ytgap > 0). Absolute t-ratios are in parentheses. The Ljung–Box test examines the null hypothesis of no autocorrelation problem up to the 12th order.

1.4 1.2 1.0

test results point to the absence of autocorrelation and heteroscedasticity in the error term, and the cusum of squares test results in Fig. 2 confirm that the equation is stable in terms of parameters. The parameter estimates on both D1 jUtgap j and ð1  D1 ÞjUtgap j are significant, with the signs suggesting that marriage is pro-cyclical. However, the parameter estimate on D1 jUtgap j, 0.187, is larger than the parameter estimate on ð1  D1 ÞjUtgap j, 0.118, indicating that marriage is more responsive to the gap when the economy is underperforming.1 Further, the marriage rate, while contemporaneously independent of divorce, is affected by the lagged divorce rates. Next, Equation 2 is initially estimated using OLS. Excluding the highly insignificant variables, the equation is then re-estimated, with the results reported in column 1 of Table 2. As seen, the parameter estimates on both ð1  D1 ÞjUtgap j and ð1  D2 ÞjYtgap j are not significantly different from zero. Excluding these variables, we have re-estimated the equation with the results reported in column 2. The corresponding two-stage least squares (TSLS) estimates, which purge the simultaneous bias, are reported in column 3.2 As seen, the TSLS estimates are very similar to the OLS estimates in column 2 and pass a series of diagnostic tests. For instance, the LjungBox test and the White test results point to the absence of autocorrelation and heteroscedasticity in the error term, and the cusum of squares test results in Fig. 2 confirm that the model is stable in terms of parameters. The parameter estimates on both D1 jUtgap j and D2 jYtgap j are significant, with the signs suggesting that divorce is procyclical when the economy is underperforming; that is, widening of the gap inversely affects divorce. However, divorce does not respond to the gap when the economy is overperforming. Finally, divorce is contemporaneously

1.4 The marriage rate equation Cusum of squares;

5% significance

1.2 1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.0

0.0

–0.2

–0.2

–0.4 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08

Fig. 2.

The divorce rate equation 5% significance

Cusum of squares;

–0.4 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08

Cusum of squares parameter stability test results

We have re-estimated the marriage equation in column 2 with the unemployment rate included along with D1 jUtgap j and ð1  D1 ÞjUtgap j. Our results indicate that the parameter estimate on the unemployment rate (=0.222, with a p-value of 0.345) is insignificant. The parameter estimates on both D1 jUtgap j and ð1  D1 ÞjUtgap j, however, remain significant. 2 The instrumental variables include the exogenous variables plus the lagged marriage and lagged divorce rates in the marriage equation in column 2 of Table 1 and in the divorce equation in column 2 of Table 2. 1

Marriage, divorce and economic activity

531

Table 2. Divorce rate equation estimates (dependent variable = Vt) OLS

TSLS

Independent variables

(1)

(2)

(3)

Constant Vt – 1 Vt – 7 Mt Mt – 7 D1|Utgap| (1 – D1)|Utgap| D2|Ytgap| (1 – D2)|Ytgap| Adjusted R2 Jung–Box test p-value White test p-value

0.013 (0.59) 0.838 (10.4) –0.231 (4.88) 0.187 (3.75) 0.154 (2.60) 0.148 (3.02) 0.035 (0.60) -–0.085 (3.45) –0.017 (0.63) 0.930 0.236 0.550

0.017 (0.88) 0.831 (11.6) –0.228 (4.96) 0.192 (4.54) 0.158 (2.77) 0.144 (3.03)

0.014 (0.72) 0.819 (11.0) –0.232 (4.98) 0.208 (4.11) 0.169 (2.80) 0.151 (3.08)

–0.083 (3.49)

–0.085 (3.53)

0.933 0.327 0.528

0.933 0.390 0.538

Notes: See the notes in Table 1.

determined by marriage and also is affected by the lagged marriage rates.

IV. Interpretation and Conclusions Consistent with theory, both marriage and divorce are procyclical when the economy is underperforming. For marriage, this effect is captured by a widening of the unemployment gap; for divorce, the unemployment gap and the GDP gap contain independent information. An important factor in the decision to marry is the potential for economic gains. A bad economy reduces the potential for these gains. However, for couples who are already married, an improvement in economic conditions increases the potential utility from separation. This effect is especially strong for women, who initiate most divorces (Brinig and Allen, 2000). The effect of the economic climate on divorce is active only when the economy is underperforming. This amplifies earlier findings that economic distress stabilizes existing marriages. A worsening economy makes it difficult to divorce, because divorce dissolves the economies of scale and the risk sharing that characterize marriage. By contrast, the formation of new marriages is pro-cyclical regardless of whether the economy is overperforming or underperforming. This follows the literature that the decision to marry is a calculated consideration of expected gains from marriage,3 while the relationship between divorce and the economy is specific to the economic stresses associated with an underperforming economy. As for the bidirectional causality between marriage and divorce, the cumulative effect of lagged increases in one is an increase in the other. This follows newer theoretical

models which treat the process as dynamic (Chiappori et al., 2008). An increase in the divorce rate feeds back to higher marriage rates by increasing remarriage. This improvement in remarriage prospects increases utility outside of marriage, which in turn endogenously increases the divorce rate. Consistent with this approach, we find the effect of divorce rates on subsequent marriage rates is lagged, while the effect of changes in marriage rates on divorce rates is both lagged and contemporaneous.

Acknowledgement We thank Todd Sandler for helpful feedback.

References Alm, J. and Whittington, L. A. (1995) Does the income tax affect marital decisions?, National Tax Journal, 48, 565–72. Becker, G. S. (1981) A Treatise on the Family, Harvard University Press, Cambridge. Brinig, M. F. and Allen, D. W. (2000) ‘These boots are made for walking’: why most divorce filers are women, American Law and Economics Review, 2, 126–69. Chiappori, P. A., Iyigun, M. and Weiss, Y. (2008) An assignment model with divorce and remarriage. IZA Discussion Paper No. 3892. Available at http://ssrn.com/abstract=1318851 (accessed 23 December 2013). Clark, K. B. and Summers, L. H. (1981) Demographic differences in cyclical employment variation, Journal of Human Resources, 16, 61–79. Ekert-Jaffe, O. and Solaz, A. (2001) Unemployment, marriage, and cohabitation in France, Journal of Socio-Economics, 30, 75–98. Eliason, M. (2012) Lost jobs, broken marriages, Journal of Population Economics, 25, 1–33.

3 A man’s future labour market potential is important (Oreffice and Quintana-Domeque, 2010). Alm and Whittington (1995) show that the potential tax benefits are a significant determinant of the decision to marry.

532 Elsby, M. W., Hobijn, B. and Sahin, A. (2010) The labor market in the Great Recession. NBER Working Paper No. 15979. Available at http://www.nber.org/papers/w15979 (accessed 23 December 2013). Gutiérrez-Domènech, M. (2008) The impact of the labour market on the timing of marriage and births in Spain, Journal of Population Economics, 21, 83–110. Harknett, K. and Schneider, D. (2012) Is a bad economy good for marriage? The relationship between macroeconomic conditions and marital stability from 1998–2009, National Poverty Center Working Paper Series #12-06. Available at http://npc.umich.edu/publications/u/2012-06%20NPC% 20Working%20Paper.pdf (accessed 23 December 2013). Jensen, P. and Smith, N. (1990) Unemployment and marital dissolution, Journal of Population Economics, 3, 215–29.

H. Baghestani and M. Malcolm Kesselring, R. G. and Bremmer, D. (2006) Female income and the divorce decision: evidence from micro data, Applied Economics, 38, 1605–16. Oreffice, S. and Quintana-Domeque, C. (2010) Anthropometry and socioeconomics among couples: evidence in the United States, Economics & Human Biology, 8, 373–84. Schaller, J. (2013) For richer, if not for poorer? Marriage and divorce over the business cycle, Journal of Population Economics, 26, 1007–33. Shore, S. H. (2010) For better, for worse: intrahousehold risksharing over the business cycle, The Review of Economics and Statistics, 92, 536–48. South, S. J. (1985) Economic conditions and the divorce rate: a time-series analysis of the postwar United States, Journal of Marriage and the Family, 47, 31–41.