[cf. Eaves et al 1999] Recent research has found a surprising degree of homogeneity in the personal political communication network of individuals but this work has focused largely on the tendency to sort into like-minded social, workplace, and residential political contexts. We extend this line of research into one of the most fundamental and consequential of political interactions—that between sexual mates.

Using data on thousands of spouse pairs in the United States, we investigate the degree of concordance among mates on a variety of traits.

Our findings show that physical and personality traits display only weakly positive and frequently insignificant correlations across spouses. Conversely, political attitudes display interspousal correlations that are among the strongest of all social and biometric traits.

Further, it appears the political similarity of spouses derives in part from initial mate choice rather than persuasion and accommodation over the life of the relationship.

…Here, with the aid of data from the “Virginia 30,000” study of kinships and their relatives (“VA30K”), these initial findings are extended more concertedly into the political realm. …individuals included in the original VA30K came from a population registry originated in the late 1970s as a result of collaboration between Virginia Commonwealth University and the Virginia Vital Records Office in which all birth records in Virginia were accessed to identify twins. This Virginia twin sample was supplemented with additional twins drawn from a national mailing to American Association of Retired Persons (AARP) members. First-degree relatives and spouses of the twins in the registry were then surveyed. Response rates for the “Health and Lifestyles” survey used here, which was conducted in the mid-1980s, were 70% for the twins and 45% for the first-degree relatives and spouses, far better than the response rate for typical surveys. Obviously, this sample is in no respects intended to be random given that a particular component of the population—twins and their relatives/spouses—was targeted. Nonetheless, core demographics indicate a reasonably typical group: mean age = 49; 36% Republican, 32% Democrat, 32% moderate (or don’t know); 59% female; 32% with college degrees, 25% with some college but not a 4-year degree, 29% with only a high-school degree, and 11% not finishing high school. Additional details on the sample are available in Truett et al 1994 (pg224–25) and Lake et al 2000_24ya.

…The use of a population based study built around twins is somewhat unique. However, for the purposes of this study, we know of no reason to think that the match between a twin and that twin’s spouse should be any different from the match between a nontwin and that non-twin’s spouse, and none of the conclusions we are about to draw changes appreciably when the analysis is confined to the 773 cases in which neither spouse is a twin (see Column 3 of Table 1 below).

**Table 1**: Spousal Concordance on 16 Traits—Pearson’s r (n).
*Source*: VA30K survey data (as described in text). *Note*: The reported correlations are Pearson’s r’s followed by the number of spouse pairs in parentheses. All of the correlations are statistically-significant at the 0.001 level except for Stunkard silhouette for parents of twins (which reaches the 0.05 level) and all of those for extraversion and impulsivity (which fail to reach statistical-significance at even the 0.1 level).
Trait	All Pairs	Twins and Spouses	Parents of Twins
Church attendance	0.714 (4,950)	0.727 (4,250)	0.631 (700)
W-P Index (28 items)	0.647 (3,984)	0.658 (3,443)	0.534 (541)
Drinking frequency	0.599 (4,984)	0.593 (4,244)	0.625 (740)
Political party support	0.596 (4,547)	0.595 (3,924)	0.598 (623)
Education	0.498 (4,957)	0.462 (4,261)	0.583 (696)
Height	0.227 (4,964)	0.239 (4,257)	0.175 (707)
EPQ lie scale	0.217 (4,475)	0.203 (3,847)	0.306 (628)
Smoking frequency	0.211 (4,266)	0.203 (3,417)	0.276 (484)
Weight	0.154 (4,985)	0.154 (4,286)	0.108 (699)
Sleep length	0.127 (5,086)	0.111 (4,360)	0.206 (726)
EPQ psychoticism	0.122 (4,545)	0.118 (3,918)	0.142 (627)
Stunkard Silhouette ideal	0.121 (4,894)	0.120 (4,068)	0.139 (671)
Stunkard Silhouette	0.119 (5,019)	0.121 (4,316)	0.086 (703)
EPQ neuroticism	0.082 (4,991)	0.074 (4,273)	0.118 (718)
EPQ extraversion	0.005 (4,739)	0.006 (4,059)	−0.010 (680)
EPQ impulsivity	0.002 (4,875)	−0.006 (4,181)	0.044 (694)
Mean correlation	0.278	0.274	0.285

…The first column of Table 1 is the Pearson’s correlation for each of the variables for the full sample of spouse pairs. The first feature of note is that the sign for all variables is positive, casting further doubt, at least when it comes to mate choice, on the notion that opposites attract.

Still, there is considerable variation in the size of the correlation coefficients. Many, notably those for some of the main dimensions of personality, are quite weak. Physical measures such as height, weight, and Stunkard silhouette (both actual and desired) are positively correlated across mates, but only mildly, with correlations running from barely 0.1 to a little over 0.2. Correlations in personality traits tend to be similar or even smaller than those for physique. In fact, the only personality index with a correlation over 0.2 is the social desirability (“lie”) scale with a correlation of 0.217 (similar results can be found in Eaves et al 1989_35ya; Feng & Baker1994). Neither sleeping nor smoking patterns is strongly correlated between spouses, but correlations for alcohol consumption and church attendance are large (one might speculate about the impact of two very common, if socially divergent, locations in which prospecting for mates often occurs).

Generally, the largest correlations are found for those measures that might be expected to have greater social impact, notably church attendance, educational attainment, and political affiliation. Support for one political party or the other is definitely concordant, with a correlation between spouses of nearly 0.6 (see also Stoker & Jennings2005). The W-P index, which contains numerous political items in addition to items designed to assess other attitude domains is explored more closely below.

…Overall, the spousal correlations for the W-P items are positive and large. The correlations for the EPQ personality items, on the other hand, are much smaller with many hovering around 0.0 and some even registering as slightly negative. With regard to a few isolated personality items, such as being talkative, opposites may attract—though the size of the correlation coefficients indicates that this pattern is hardly a strong one. The slightly different item formats and distributions discourage precise comparisons, but it is clear that in terms of spousal concordance, social and political attitudes function differently from personality traits.

…The general finding that political attitudes of spouses are about as similar when they are first married as when they have been married decades will be surprising to some readers.

…But why does assortative mating occur at all? It is certainly possible to posit biological reasons for assortative mating,⁴ but it is also possible that mate selection in the modern era is at least in part a specialized case of friend selection. Thanks to substantial mental capabilities and extended periods of time during which survival and reproduction are not at issue, much of what humans do and think (though not nearly as much as humans imagine) is orthogonal to canonical conceptions of biological selection. Survival and reproduction aside, people appear to prefer to be around those who share their sociopolitical orientation but not necessarily those who share their personality. This is true of selection of friends and selection of mates and, as such, may indicate the need for new thinking regarding the workings of evolution as it applies to traits remote from reproductive fitness.