The effect of personal attributes and mate preference on marriage and relationship an experimental investigation

The Effect of Personal Attributes and Mate
Preference on Relationship and Marriage:
An Experimental Investigation

ZHANG MAN
(Bachelor of Economics, Nankai University)

A THESIS SUBMITTED
FOR THE DEGREE OF MASTER OF SOCIAL SCIENCES
(BY RESEARCH)
DEPARTMENT OF ECONOMICS
NATIONAL UNIVERSITY OF SINGAPORE
2013


Declaration

This thesis work is the original work of Zhang Man, Performed independently under
the guidance of Prof. Zhang Jie (supervisor), Economics Department, FASS, National
University of Singapore, from May 2012 to Aug 2013.

Zhang Man

Zhang Man

07/08/2013

Full Name

Signature

Date

1


Acknowledgments

This thesis is like a souvenir signifying the end of my school life, and I am very
grateful to many people during my study journey at NUS.
First of all, I would like to deeply appreciate my supervisor, Prof. Jie Zhang, for his
invaluable guidance throughout my research process, if without his continuous
encouragement this thesis might not be completed.
And then, sincerely thanks the co-chairs and other committee members for their
presence and helpful comments to my thesis.
Additionally, I gratefully acknowledge to all the teachers and classmates who taught
me and helped me in these years, and have to say I like all of you from my heart!
Moreover, I should give many thanks to the administrative managers and staffs in our
Economics Department of FASS for their kind assistances.
Last but not least, thank you, mother and father. Thanks for emotional and financial
support all the time. I will try to make me better and better because I am definitely
sure that you always deserve a better daughter. Love you, ma and pa.
At this moment, any words fail to express my heartfelt gratitude to ‘the big family and
families’ of our Economics Department in FASS of NUS, but I still want to say that
thank you for gifting into my school life such a precious memory.

2


Table of Contents
Pages

Declaration..................................................................................................................... 1
Acknowledgements........................................................................................................ 2
Table of Contents........................................................................................................... 3
Summary........................................................................................................................ 3
List of Tables..................................................................................................................5
List of Figures................................................................................................................ 5
Sections:
1. Introduction.................................................................................................................6
1.1 Justification of Research....................................................................................................... 6
1.2 Objectives of Research......................................................................................................... 8
1.3 Research Background......................................................................................................... 10

2. Precondition of Research......................................................................................... 14
3. Modeling and Algorithm..........................................................................................17
3.1 Modeling............................................................................................................................. 17
3.2 Matching Algorithm............................................................................................................21

4. Experimental Investigation...................................................................................... 24
4.1 Individual Information Gathering.......................................................................................24
4.2 Results and Analysis........................................................................................................... 27

5. Conclusion................................................................................................................33
Reference......................................................................................................................34
3


Summary

Using data from the individual investigation and Speed Dating Experiment via

Gale-Shapley Algorithm, we study the relation between personal attributes and mate
preference, and shed light on their effect on mate selection and relationship
satisfaction. The design of our investigation and experiment can allow us to observe
both the selection decision and matching results. The relationships and marriages of
the couples who were assorted for age, appearance and character were shown more
stable and happier. Furthermore, the most stable and happiest matches are more
probably based on the similarity of personal character.
Thus, personal attributes and mate preference interact with each other by the three
main similarities during the process of mate selection. The effect of personal attributes
and mate preference on romantic and marital relationship is significant.

4


List of Tables

Table 1.1 Comparison of Matching Model Simulations……………………………..11
Table 2.1 The Relationship and Marriage Investigation in Five Cities* of China…...13
Table 2.2 The Relationship and Marriage Investigation in Rural Area of China…….14
Table 2.3 The Relationship and Marriage Investigation during 2000-2004 in China..14
Table 4.1 Cross-individual Correlations for Personal Attributes
Sample Group: Married & Dating………………………………………....26
Table 4.2 Cross-individual Correlations for Personal Attributes
Sample Group: Experimental Matching…………………………………...28
Table 4.3 Rating on Matching Stability and Happiness in Average………………….30

List of Figures

Figure 3.1 An Unstable Matching Assignment………………………………………20
Figure 3.2 A Stable Matching Reassignment (Optimal Solution)……………………20

Figure 3.3 Three Rounds Stable Solution of Matching Assignment…………………22

5


1. Introduction
This research attempts to shed light on the relation between mate selection and two
main factors: one is personal attributes and another is mate preference, and to further
our understanding about why individuals accept and reject romantic or marital
relationships and how to ensure the stability and happiness in relationships.
The whole thesis is composed of six sections. Section 1 introduces the justification,
objective and background of this research. Sections 2 and 3 provide the preconditions
and algorithm to support the following experimental investigation in Section 4.
Additionally, Section 4 lists and analyzes the experimental results for establishing
evident basis for the explanation and guidance on individual relationships through
correlation analysis in the following Section 5. At last, Section 6 presents the
concluding remarks.

1.1 Justification of Research
At any point of time in human society, most of individuals will experience or are
experiencing or experienced their own marriages in their limited lifetime. It has really
long been the goal of social scientists to better know why individuals accept and reject
a romantic or marital proposal. Simply through this statement, we can be aware of the
consequence of marriage and its impact on both individuals and society at large. Thus,
the first reason of this research is directly and briefly embodied in the marriage
importance which is indicated on its individual effect and society effect.
6


Additionally, by a common-sense view about the marriage that is most of individuals

do not want to just be married but desire a relatively stable and happy marital
relationship with their partner in a long-term, it would be reasonable to assume that
individuals like making their marriage decisions by happiness and stability most at
least. In other words, they must be keeping the best interests of their mates and
themselves during making the marriage decision. To some extent, it cannot be too
much saying that a happy and stable marriage is far more significant than anything
else in determining personal well-being (David Brooks, 2010). Thus, exploring how
to make sure of the marital happiness and stability would benefit most individuals; it
is also the key point of the research.
In the view of Economics and Law, the points of marital happiness can be based on
three points. The first one is the compatibility of supply and demand with the
emotional feelings between wife and husband; the second point is the satisfaction of
rights and duties in their current marriage; the last one is the expectation for their
future marriage. In this thesis work, the further experimental design is based on these
three points to evaluate the marital happiness and stability.
However, as supposed that the happiness of marriage can be evaluated, an optimal
marriage may not be only decided by the maximization of marital happiness in fact,
because marriage is multi-angle and multi-dimensional. For instance, parents and
friends tend to be involved in one’s mate choice even marriage, so that they must be
willing to make some suggestions on one’s relationship and even marriage in their
view. In a way, it can cause a kind of information asymmetry from multi-angle views.

7


And, the other factors, such as individual preferences for the mate selection, will pose
different emphases of a relationship or marriage. To simplify the complexity and
reduce the multi-angle and multi-dimensional interferences, we are going to introduce
several rating questionnaires and speed dating experiments for gathering individual
information in this study, and then use the results to do the analogy analysis on the

individual relationship and marriage. It would be an essential part of this research.

1.2 Objectives of Research
As the historical records of anthropology, the complexity of marital relationship
subject was established as early as the word ‘mariage’ in Middle English first
appeared. In The Future of Marriage in Western Civilization (1936), Edvard
Westermarck defined Marriage as a relation of one or more men to one or more
women that is recognized by custom or law. According to the marriage definition in
Notes and Queries, Kathleen Gough (1959) modified the definition as a union
between a woman and one or more other persons such that children corn to the
woman are the recognized legitimate offspring of both partners. However, the other
anthropologists such as Duran Bell (1997) then criticized this legitimacy-based
definition of marriage. Definitely, marriage is a very complicated study subject. The
enormous argument and criticism over marriage definition is still circulated in
academic circles all these years. And, it seems like no one can tell what exactly the
marriage is because of its complicacy, though lots of famous scientists indeed do great

8


contribution to marriage research.
Thus, here for a reasonable simplification in analyzing the complicated subject, this
paper intends to observe the notable factors in a marriage and detect significant
internal and external effects on the marriage market through experimental
investigation and correlation analysis, and then on this basis examine the causal effect
of personal attributes and mate preference on the mate selection behavior by
simulated experiments (Speed Dating Experiments). In the view of economics, the
experiments would be focused on spanning a variety of main aspects of marital and
romantic relationship and estimate their effect on the matching in a dynamic duration.
Thereby, we are able to provide a relatively comprehensive evidence and guidance on

how to use personal attributes and mate preference in combination to get more
happiness, stability and satisfaction from a marital or romantic relationship.
The main purpose of this research is to collect relatively accurate and sufficient
personal information from 31 pairs of married couples and 48 dating couples and 50
young singles, and investigate the impact of personal attributes and mate preference
on their current relationships. It would be expected that the experimental data and the
tractable dating and matching results, to some extent, can accurately illustrate the
corresponding relations among personal attributes, mate preference and the quality of
matching which is evaluated by stability and happiness. Moreover, the research result
is expected to be significant and persuasive to clearly explain the effect of attributes
and preference on mate selection, and provide a good guidance to make more
satisfaction on a romantic or marital relationship for more individuals.

9


1.3 Research Background
Gary S. Becker (1973 and 1974, summarized in 1991) applied the economic approach
to analyze the most sensitive and fateful individual decisions, such as choosing a
spouse in matching process, and exposited the transferable utility model of the
marriage market, a benchmark model at present. Since Becker’s path breaking
research on marriage, many economists have estimated this benchmark model of
marriage and systematically studied the clear condition of marriage markets. However,
there are relatively fewer studies on individual mate selection and marriage decision
making in the well-known documents of marriage economics.
The most relevant literatures among the rare records would be Gillian Hamilton and
Aloysius Siow (2001) that estimated the contributions of social heterogeneity and
assortative matching, examined the frictions to aggregate marriage behavior by
analyzing the data of Quebec in the 18th century. They found that a simple random
matching model of the marriage market, in which there are gains to assortative

matching and women may leave the marriage market at a higher rate than men, can
explain the data of Quebec. Moreover, their estimates suggest that women fared
slightly better than men in the impact of differential fecundity on the welfare. Eugene
Choo and Aloysius Siow (2003) estimated the transferable utility model which
rationalizes the standard interpretation of marriage rate regressions in marriage market.
They resolved two problems early in preliminary study: One was an observation
problem in equilibrium transfers which are seldom observed in modern marriages.
The other was an individual identification problem in the wide marriage market,
10


which would be further discussed in detail later. However, unlike in the Choo and
Siow (2003), the types of women and men should be identified respectively by their
personal attributes and mate preferences in this research work. For a more accurate
individual identification, the Personal Code Test referring to the Taylor Hartman
Theory (2007), and the Multiple Speed Dating Experiments referring to the Five
Factors in Mate Selection and Marital Satisfaction (M. D. Botwin, 1997), should be
introduced into our research, because these two investigations would be used to
evaluate both impacts of personal attributes and mate preferences on the mate choices.
Additionally, the personal code can be adequately shown by individual characteristics
to some extent. The personal characteristics figure prominently in unstructured
nomination of what individuals want in a mate (Langhorne and Secord, 1955).
Furthermore, personal traits would be able to make individuals structuring and
designing instruments to assess mate preferences (Buss and Barnes, 1986; McGinnis,
1958). For instance, both women and men state clearly that they want a mate in their
rest of life, who are more likely to be dependable, sociable, kind, understanding,
stable, and intelligent enough to realize their own mate preference, such as physical
appearance, economic resources, education background, etc. (Botwin et al., 1997).
Preparing for a good experimental simulation of the relationship interrelation, we
viewed the DYNASIM model, which originally developed by Urban Institute

(www.urbaninstitude.org) and then successively renovated by Zedlewski in 1990 and
Favreault & Smith in 2004. The comparison of results about several matching model
simulations is shown in Table 1.1, from which we can extract some useful information

11


for our following experimental investigation and trackable simulation.

Table 1.1 - Comparison of Matching Model Simulations
Model

Year and Reference

Matching Variables

DYNASIM

Zedlewski,

Age Difference, Education Difference

1990
Matching Process: Singles randomized. First pair evaluated using exponential
probability function. Paired if random number less than probability, otherwise repeat
for up to 10 potential partners. If no matches, then pair those with highest
probability.
SAGE

Cheesbrough & Scott,

Age, Education, Marital Status, Pregnant

2003
Matching Process: Monte Carlo simulation comparing transition probabilities based
on logistic regression models using retrospective partnership histories from survey.
APPSIM

Bacon & Pennec,

Age Difference, Education Difference

2007
Matching Process: Similar to DYNASIM. Singles randomized. First pair evaluated
using exponential probability function. Paired if random number less than
probability, otherwise repeat for up to 10 potential partners. If no matches are made
and the total number of couples has not been met, then pair those with highest
probability.
Source: jasss.soc.surrey.ac.uk/16/1/6.html
12


As illustrated in this table, the matching algorithm of the DYNASIM model is mainly
based on two individual differences, the age difference and education difference.
Unlike in the DYNASIM model (Zedlewski 1990), we intend to put more personal
attributes which can be identified into our further investigation and simulation.
In short, prior studies indeed provide valuable evidence on the matching model of
marriage, the behavioral difference with different gender in marriage market, personal
traits, and mate preference. However, most of prior studies emphases on the
mathematical modeling and computing simulation. No prior study has undertaken a

systematic research on how to provide individuals with a clear and practical guidance
to make a successful marital or romantic relationship, through individual investigation
and the vector analysis of personal attributes and mate preference. Therefore, this
empirical study attempts to fill the gap among the previous literature.

13


2. Precondition of Research
Love is an eternal theme, especially when the topic discussed is the Marriage. For
most people, it has to be admitted that love can lead to a marriage. And, the initial
cause of love must be a kind of chemical reaction motivated between female and male,
called attraction. Certainly, there are some other special cases, such as arranged
marriage, also leading to a marital relationship.
However, in the modern age, more and more people can intuitively find that there are
fewer and fewer parents select the person whom their children should marry. Here, it
is necessary to provide some evidence for this intuition. According to the data from
several relationship and marriage sample investigations, it is obvious that a dramatic
decline exists in the number of arranged marriage while the number of free-love
marriage is continuously increasing in both urban area and rural area. The details are
clearly shown in Table 2.1, 2.2 and 2.3.

Table 2.1 - The Relationship and Marriage Investigation in Five Cities* of China
Sample Size: 4858 Couples
Proportion
Types
Free-love Marriage
(100%)
Arranged marriage
(100%)

Before
1937

1937 – 1967

1967 – 1997

After 1997

4.99%

32.98%

62.03%

null

54.72%

44.34%

0.94%

null

Source: Family Survey in Main Five Cities, Liu Ying, Chinese Family and Marriage
Study [M]. Beijing: Social Science Reference Press, 1987.

14

Table 2.2 - The Relationship and Marriage Investigation in Rural Area of China
Sample Size: 2866 Couples
Proportion
Types
Free-love Marriage
(Rural Area)
Arranged marriage
(Rural Area)

Before
1966

1967 – 1976

1977 – 1986

1987 – 1996

32.8%

37.9%

48%

56.8%

67.2%

62.1%

52%

43.2%

Source: Xu Anqi, Family Investigation in Urban and Rural Area of China, Social
Science Press of China, 1997, September.

Table 2.3 - The Relationship and Marriage Investigation during 2000-2004 in China
Types
Proportion
(100%)

Free-love Marriage
(Average)

Arranged marriage
(Average)

85.6%

14.4%

Source: Zhang Shouzhi, Marriage Database of China [N], Health and Lives Report,
Shanghai Social Science Research Institution, 2004.

Logically, we may easily reach a prevailing view that free-love marriage has already
almost replaced the arranged marriage at present because of love, the primordial
nature of marriage. Therefore, ‘Free-love Marriage’, which can imply the freedom to

choose one’s spouse, is obviously the first precondition in this experimental research.
Next, the second precondition should be that some people can be realistic and have
clear awareness of themselves and others. In accordance with this premise,
participants in the individual experiment would be divided into different groups in the
15


same processing period. That would be also referring to individual mate preference
and marital decision-making. Actually, in the process of mate selection both women
and men can show their personal attributes to the others. In the meanwhile, they
would express affection to the ideal candidate with considering their mate preference,
so-called ‘Rational Behavior’, or without considering their mate preference, so-called
‘Irrational Behavior’ here, due to their own personality. Since the process of mate
selection is a two-way choice under the free-love condition, the second premise
regarding awareness of own attributes and mate preference is an important element in
the experiment.
However, a good research result may not be achieved through the experimental
investigation just with the two preconditions yet. How to ensure the marital possibility
and feasibility is still a difficult problem to the experimental design and control at
present, especially when previous theory or hypothesis condition is not sufficient.
Thus, the Marriage Possibility Index should be innovative to introduce as the third
precondition into the following study.

16


3. Modeling and Algorithm
3.1 Modeling
According to the purpose of this research, it is needed to establish a model to simulate
the interaction process in individual relationships, consequently refining the data

analysis of experimental investigations. It should be two kinds of simulations because
of the experimental subjects. The first one is counterfactual simulation, which will be
applied to observe the married subjects; the second one is tractable simulation, which
will be applied to observe both dating and single subjects.
Moreover, it has to be performed under the hypotheses for the simulations, and the
hypotheses should be able to shed light on the following questions initially for this
research work. (1) How does it work between individuals and marriage market? (2)
To understand the individual identification problem, the statistics of the female and
male types would be imperative to identify the individuals in the marriage market. (3)
How many personal attributes could be observed and compared in the process of mate
selection? (4) What is the correlation between personal attributes and mate preference.
(5) How much of distinguished features in mate preference can be put down to the
individual heterogeneity or experience heterogeneity? (6) Is there any utility matching
algorithm can be used to simulate the interaction during the mate selection process? (7)
How can we assess the happiness in both marital relationship and romantic
relationship? (8) How can we make marriage and relationship more effective, happy
and health? With all the eight highlighted points, it would be a right time to start
17


setting up hypotheses and modeling systematically.
In accordance with the Choo and Siow’s marriage-matching function (2006) and the
other related studies by Dagsvik, et al. (2001) and Siow (2008), we can suppose that:
(1) Let F and M be the population vector of female and male respectively, which is
investigated as experimental subjects in this research. There are I types of female and
J types of male in our sample. Let f i be the number of eligible type i female and

m j be the number of eligible type j male. Additionally, let N i 0 , N 0 j and N ij be
the number of unmarried type i female, the number of unmarried type j male ,and
the number of type i female married to type j male, respectively.

Thus, the accounting constraints should be:
J

Ni 0 +

N

N0 j +

N

j 1

ij

I

i 1

ij

= fi ,  i

= mj ,  j

N i 0 , N 0 j , N ij  0 , i , j .
(2) In a {i, j | i  I , j  J } marriage, a systematic marriage surplus,  ij , that depends
on the type of marital match is generated as a total surplus of the marriage. Let  ij
be the share of marriage surplus that is obtained by the wife ( Fij ), and then ij  ij
is the share of marriage surplus that is obtained by the husband ( M ij ). Meanwhile,

each type i female gets an idiosyncratic payoff, eFij , that depends on her specific
identity and the type of spouse that she marries, not on his specific identity. Her
idiosyncratic payoff does not depend on her marriage surplus (  ij ). In the same way,
each type j male gets an idiosyncratic payoff, eM ij , that depends on his specific

18


identity and the type of spouse that he marries, not on her specific identity. His
idiosyncratic payoff does not depend on his marriage surplus ( ij  ij ).
(3) Similarly, in an {i, j | i  I , j  J } romantic relationship, a systematic relationship
surplus, ij , that depends on the type of relationship match is generated as a total
surplus in the romantic relationship. Let  ij be the share of romantic relationship
surplus that is obtained by the girlfriend ( Fij ), and then ij  ij

is the share of

romantic relationship surplus that is obtained by the boyfriend ( M ij ).
Meanwhile, each type i female gets an idiosyncratic payoff, eF ij , that depends on
her specific identity and the type of her boyfriend, not on his specific identity; her
idiosyncratic payoff also does not depend on her romantic relationship surplus (  ij ).
Equally, each type j male gets an idiosyncratic payoff, eM ij , that depends on his
specific identity and the type of his girlfriend, not on her specific identity; His
idiosyncratic payoff also does not depend on his marriage surplus ( ij  ij ).
(4) Each individual who chooses to remain single also gets a systematic unmarried
surplus,  i 0 for the female or  0 j for the male, which depends on their own type,
as well as an idiosyncratic payoff, ei 0 for the female or e0 j for the male, which
depends on their own identity.
(5) In a two-individual marriage group mk , where m is denoted as the marriage
group, k is an integer and k  1 , both wife and husband utilities of female of type

i who marries a male of type j should be:
Fij : uijmk   ij  eFij
M ij : uijmk  ij  ij  eM ij

19


Similarly, in each two-individual romantic relationship group rk , where r is denoted

as a romantic relationship, k is an integer and k  1 , both girlfriend and boyfriend
utilities of female of type i who match a male of type j should be:

Fij : uijrk   ij  eF ij
M ij : uijrk  ij  ij  eM ij
If an unmarried individual chooses to remain single or break-up to become single and
available, then the utility of this single individual sk should be:

Fi 0 : ui 0sk   i 0  ei 0
M 0 j : u0 jsk   0 j  e0 j
(6) Given the systematic surplus and idiosyncratic payoff in three different situations,
by considering every potential spouse choice including keeping unmarried, we can
assume on the one hand that the utility of a married individual has already established
without any consideration of the individual who has been divorced, separated and
widowed in this research; on the other hand, each unmarried individual would like to
make an initial mate selection which can maximize their own utility rather than the
utilities of both mate candidate and themselves. Unlike a commodity selection, the
mate selection is not only two-way selection but also taking on both emotional nature
and physical nature. Consequently, it should be introduced an efficient algorithm to
analyze the process of individual mate selection in the meantime.

20


3.2 Matching Algorithm
The Gale and Shapley’s algorithm is the pivotal theoretical foundation for our further
experimental investigation. According to their paper work (Gale and Shapley, 1962),
it is not difficult to get that: (1) The matching assignment is unstable if there are two
male assigned to two female respectively, although a male prefers the female to whom
the other male is assigned while the female he prefers also likes him better. An
unstable matching assignment can be shown as the Figure 3.1.

M 1 ( F1 , F2 )  F1 ( M 2 , M 1 )
M 2 ( F1 , F2 )  F2 ( M 1 , M 2 )
Figure 3.1 - An Unstable Matching Assignment

(2) The matching assignment is optimal if there are a male and a female who prefer
each other to their assigned mates, then the male and the female will be reassigned to
each other, and the rest male and female will be so also. In the same way, a stable
matching reassignment can be illustrated in the Figure 3.2.

M 2 ( F1 , F2 )  F1 ( M 2 , M1 )

M 1 ( F1 , F2 )  F2 ( M 1 , M 2 )
Figure 3.2 – A Stable Matching Reassignment (Optimal Solution)

21


(3) For the matching assignment with more than two participants, the two theorems of

Gale-Shapley algorithm should be reaffirmed in this section.
Theorem 1: There always exists a stable set of marriages.
Based on the first theorem, however, Gale and Shapley (2012) found that male
participants would be strictly better off if they were the ones proposing. If female
participants proposed, it would lead to the second theorem.
Theorem 2: Every proposer is at least as well off under the deferred acceptance
procedure as he would be under any other stable assignment.
Through examples and solutions comparison, Gale and Shapley concluded that the
male participants would be better off only when they proposed to female participants.
The reason is that every male will be assigned to their first choice under the deferred
acceptance procedure when these male participants have different first choices at the
end of the matching assignment. It means each male participant can reach to his
highest-attainable choice which is impossible for him to be better off under any other
stable solutions. To indicate this more directly and clearly, it would be better to take
an example of matching assignment with more than two participants as below.

22


The First Round Assignment

The Second Round Assignment

The Third Round Assignment

Figure 3.3 – Three Rounds Stable Solution of Matching Assignment

In order to get these highest-attainable choices, the optimal solution of matching
assignment, the male will be always as the ones proposing in this research, which can
simplify the following experimental investigation.

23


4. Experimental Investigation
Based on all above preparations, the experimental investigation would be launched
among 208 valid young participants, who were divided into three different groups: the
group of married couples, the group of dating couples, and the group of single and
available individuals. And, the three sample groups had to complete the personal
information gathering separately. Moreover, this sample classification could make our
further observation and analysis of how the personal attributes and mate preference
affect the individual marriage and relationships more clearly and conveniently to
some extent.
According to the different marital statuses in the three groups, the married and dating
couples were engaged in the information investigation only, while the available
singles were additionally invited to our speed dating experiment on One’s Day.

4.1 Individual Information Gathering
This section is mainly introducing the collection method of individual information
and the test method of information validity.

4.1.1 Sample Groups
As previously mentioned, three samples were used in the experimental investigation.
The group of married couples. The number of participants in the first sample group
was 62 young people comprising 31 pairs of married couples. All of these couples
24