The Economics of Hypergamy

Ingvild Almås; Andreas Kotsadam; Espen R. Moen; Knut Røed

doi:10.3368/jhr.58.3.1219-10604R1

Article Figures & Data

Figures

Tables
Additional Files

Download figure
Open in new tab
Download powerpoint
Figure 1
Probability of Having Found a Partner by 2015, by Own, Predicted, or Parental Earnings Rank
Notes: The filled diamonds refer to women, and the hollow circles refer to men in the left part of the figure. The graphs cover the 1952–1975 birth cohorts and show the fractions who have been married and/or had at least one child by 2015. The right-hand-side panels include 95 percent confidence intervals. Panels A and B are based on ranks within own birth cohort’s distribution of earnings (including self-employment income) during age 28–40. Panels C and D are based on ranks in the predicted age 28-40 distribution. Panels E and F are based on ranks within the age 52–58 earnings distribution of all parents belonging to each offspring birth cohort (inflated to a common calendar year value). Panels B, D, and F include 95 percent confidence inervals. Number of observations is 757,868 for men and 723,317 for women.
Download figure
Open in new tab
Download powerpoint
Figure 2
Probability of Having Had Multiple Partners by 2015, by Own and Parental Earnings Rank
Notes: The filled diamonds refer to women, and the hollow circles refer to men in the left part of the figure. The graphs cover the 1952–1959 birth cohorts and show the fractions who have been married and/or had a child with atleast two different persons by 2015. See note to Figure 1 for the definition of the different rank measures. The right-hand-side panels include 95 percent confidence intervals. Number of observations is 200,074 for men and 202,449 for women.
Download figure
Open in new tab
Download powerpoint
Figure 3
Average Partner Rank by Own Rank, Based on Offspring’s Own or Parental Earnings
Notes: The filled diamonds refer to women, and the hollow circles refer to men in the left part of the figure. The graphs cover all couples formed between men and women in the 1952–1975 birth cohorts and show the average percentile rank of the partner in own and parental earnings distributions, respectively. The two lower panels include 95 percent confidence intervals. For the rank measure based on maximum earnings two to five years before the match (Panels E and F), we use a reduced sample of actual matches, namely those occuring when both partners have completed their education and had positive earnings before the match (N = 95,233 men and 95,800 women).
Download figure
Open in new tab
Download powerpoint
Figure 4
Trends in Hypergamy
Notes: Panel A shows for each birth cohort (1952–1975) the estimated difference between women and men in the relationship between the probability of being matched by age 40 and earnings rank potential. Each data point corresponds to the number reported for all cohorts in Table 2, Column 3. Panel B shows, for each birth cohort (1952–1959), the estimated difference between women and men in the relationship between the probability of being matched multiple times and earnings rank potential and corresponds to the number reported in Table 3, Column 3. The lower number of cohorts in Panel B than in Panel A reflects that multiple matches typically occur at higher ages, such that we need to observe individuals at mature ages. Panel C reports the average parental rank difference for all couples established 1982–1995. The limited time period is chosen to avoid selectivity with respect to the age composition of couples for which we are able to identify parental earnings rank (see text). All data points are reported with 95 percent confidence intervals.

View popup

Table 1

Overview of the Data Sets and Descriptive Statistics

	Total Sample		Partner Sample
	Men	Women	Men	Women
Number of observations	757,868	723,317	533,711	524,981
Average own earnings percentile rank (age 28–40)	50.6	50.6	54.6	50.9
Average parental earnings percentile rank (age 52–58)	50.6	50.6	51.6	51.2
Average percentile rank based on predicted earnings	50.5	50.5	51.9	51.0
Average percentile rank based on maximum earnings 2–5 years before match	NA	NA	54.9	54.4
Correlation between parental and own earnings rank	0.19	0.15	0.19	0.16
Correlation between predicted and own earnings rank	0.39	0.40	0.38	0.41
Correlation between rank 2–5 years before match and own earnings rank	NA	NA	0.70	0.58

Notes Predicted earnings are based on the following regression model estimated by gender: , where is individaual i’s age 28-40 labor earnings, is a set of indicator variables for each decile in the parental earnings distribution and E_y is a set of indicator variables for highest education attained. Number of observations with predicted earnings are, for men, 753,173 in the total sample and 529,376 in the partner sample. For women, the corresponding numbers are 719,675 and 520,545. Number of observations based on maximum earnings 2–5 years before match are 95,233 for men and 95,800 for women. This sample is restricted to individuals who have completed their education and had positive earnings before the match.

View popup

Table 2

Gender Difference in the Probability of Partnering—IV Estimates

	Linear Model			Quadratic Model
	Men (1)	Women (2)	Gender Diff. (2) (1) (3)	Men (4)	Women (5)	Gender Diff. (5)–(4) (6)
Own rank	0.374^***	0.160^***	−0.214^***	0.688^***	0.462^***	−0.217^**
	(0.007)	(0.008)	(0.011)	(0.061)	(0.070)	(0.093)
Own rank squared				−0.003^***	−0.003^***	−0.0006
				(0.0006)	(0.0007)	(0.0005)
Mean outcome	0.84	0.90		0.84	0.90
N	757,868	723,317		757,868	723,317

Notes: Own earnings rank is instrumented with parental earnings rank. Estimates and standard errors in Panel B are multiplied by 100, such that they are measured in percentage points. The gender differences in Columns 3 and 6 are evaluated within a joint model with gender interactions on all variables. Robust standard errors in parentheses. Significance:
* p < 0.10,
↵** p < 0.05,
↵*** p < 0.01.

View popup

Table 3

Gender Difference in Multiple Partnerships—IV Estimates

	Linear Model			Quadratic Model
	Men (1)	Women (2)	Gender Diff. (2)–(1) (3)	Men (4)	Women (5)	Gender Diff. (5)–(4) (6)
Own rank	0.036^**	−0.232^***	−0.268^***	−2.424^***	−3.973^***	−1.549^***
	(0.017)	(0.022)	(0.028)	(0.214)	(0.307)	(0.374)
Own rank squared				0.022^***	0.035^***	0.013^***
				(0.002)	(0.003)	(0.004)
Mean outcome	0.13	0.11		0.13	0.11
N	200,074	202,449		200,074	202,449

Notes: Own earnings rank is instrumented with parental earnings rank. Estimates and standard errors are multiplied by 100, such that they are measured in percentage points. The gender differences in Columns 3 and 6 are evaluated within a joint model with gender interactions on all variables. Robust standard errors in parentheses. Significance:
* p < 0.10,
↵** p < 0.05,
↵*** p < 0.01.

View popup

Table 4

Gender Difference in Partner’s Parental Ranks, Ordinary Least Squares Estimates

	Ranks Based on Own Earnings		Ranks Based on Parental Earnings
	Partner Rank (1)	Partner with Higher Rank (2)	Partner Rank (3)	Partner with Higher Rank (4)
Own rank	0.158^***	−0.499^***	0.092^***	−0.736^***
	(0.001)	(0.002)	(0.001)	(0.002)
Female (lowest rank)	6.179^***	25.387^***	0.739^***	9.355^***
	(0.120)	(0.172)	(0.115)	(0.161)
Female × own rank	−0.006^***	−0.307^***	−0.000	−0.088^***
	(0.002)	(0.003)	(0.002)	(0.002)
N	1,065,534	1,242,148	1,058,692	1,237,577

Notes: For the dichotomous outcome in Columns 2 and 4, the estimates and standard errors are multiplied by 100, such that they are measured in percentage points. The regressions are based on the 1952–1975 birth cohorts. All regressions control for year of birth fixed effects. Robust standard errors in parentheses. Significance:
* p < 0.10,
** p < 0.05,
↵*** p < 0.01.