Genes and Politics: A New Explanation and Evaluation of Twin Study
Results and Association Studies in Political Science
Doron Shultziner
[Forthcoming in Political Analysis Vol. 21 Issue 3]
Abstract
This paper offers a new explanation for the results of twin studies in political science that
supposedly disclose a genetic basis for political traits. I argue that identical twins tend to
be more alike than non-identical twins because the former are more similarly affected by
the same environmental conditions, but the content of those greater trait similarities is
nevertheless completely malleable and determined by particular environments. The twin
studies method thus can neither prove nor refute the argument for a genetic basis of
political traits such as liberal and conservative preferences or voting turnout. The
meaning of heritability estimates results in twin studies are discussed, as well as the
definition and function of the environment in the political science twin studies. The
premature attempts to associate political traits with specific genes despite counter trends
in genetics are also examined. I conclude by proposing that the alternative explanation of
this paper may explain certain puzzles in behavioral genetics, particularly why social and
political traits have higher heritability estimates than common physical and medical traits.
I map the main point of disagreements with the methodology and the interpretation of its
results, and delineate the main operative implications for future research.
Funding
This research was supported by the Israel Science Foundation [grant number 297-10].
Acknowledgements:
I thank Hagar Weinberger, Elizabeth Suhay, Martin Stevens, Duc Quang Nguyen, Orly
Fuhrman, and Melvin Konner who made useful comments on the paper. My special
thanks to Diane Lavett and Yechiel Barilan who read and helped improve several drafts
1
of this paper. I also want to thank the anonymous reviewers who contributed constructive
points.
2
1. Introduction
Recent studies have suggested there is a connection between genetic factors on
the one hand, and political preferences and behavior on the other. A small but very
influential group of scholars in political science, aided by behavioral geneticists, have
termed their work the “new empirical biopolitics” (Alford and Hibbing 2008) and
“genopolitics” (Biuso 2008). Compared to earlier biopolitics scholarship1, these scholars
are now “confident that this recent, decidedly empirical variant of biopolitics will flourish
as political scientists become increasingly aware that biology does not equate with either
universalism or determinism” (Alford and Hibbing 2008: 185; see also Fowler and
Schreiber 2008; Hatemi et al. 2009a).
A major argument concerning genes and politics was first raised in relation to the
question “why do people think and act politically the way they do?” Alford, Funk, and
Hibbing (2005) suggested that left and right political orientations and ideologies are
grounded in genetic predispositions. Fowler, Baker, and Dawes (2008) made a similar
argument regarding the question “why do people vote?” and Hatemi, Medland, and Eaves
(2009) suggested that genes contribute to the ‘gender gap’ in political preferences.
Hatemi et al. (2009a) found that genes determine the strength of one’s party identification
but not the party identification itself. The primary criticisms of these studies have focused
on the tenability of the equal environment assumption (EEA), namely, that identical and
1
Application of biological theories to politics is not new (Somit 1976; Easton 1976). For
insiders’ reflections on the biopolitics subfield and why it has not been well received see
Hines (1982), Losco (1982), Masters (1990), Somit and Peterson (1998), Alford and
Hibbing (2004); see also Tooby and Cosmides (1992) and Pinker (2003).
3
fraternal twins are affected by equal environments (i.e., rather than by more similar
environments for identical twins; Charney 2008a, b; Beckwith and Morris 2008; Joseph
2010; Suhay, Kalmoe, and McDermott 2007). Beyond this, the political science literature
remained silent about this scholarship and especially about its biological arguments.
The main contribution of this paper is to offer an alternative explanation for the
results of twin studies in political science. I argue that the genetic factor, namely, the
greater genetic similarity in identical twins, can only explain why identical twins are
more similar to each other in various traits compared to non-identical twins, but this
genetic factor cannot explain the specific political content or direction of these traits. In
fact, the twin research design employed in political science does not and cannot actually
test the hypothesis that twins and people in general are genetically predisposed to have
certain political traits rather than completely opposite ones for the exact same genotype.
This explanation has not yet been considered in interpreting the results of the twin studies
in political science, but it may require the rethinking of some of the assumptions,
interpretations, and methodologies that are used in this field of inquiry. The analysis
offered here can also contribute to current discussions and considerations on how to
proceed in funding research on the possible connection between genes and politics in
political science (Lupia 2011).2
2
It is important to stress that this paper focuses on the classic twin studies models and
genetic association studies that have been published in political science. It is beyond the
scope of this paper to discuss other genetic approaches such as adoption studies and
extended twin family studies (Hatemi et al. 2010). Those approaches have their own
complications and their results can at times contradict the results of twin studies (e.g.,
4
It should be stressed that the alternative explanation of the twin study results I
propose in this paper does not posit that there are no biological dispositions which are
relevant to social life and politics. Human universals and dispositions have been
identified as well as their relevance to social science (e.g., Brown 1991, 1999, 2004;
Pinker 2003; Shultziner 2010, Shultziner et al. 2010). My main point is that the method
of comparing twins does not and cannot reveal a genetic basis of political preferences.
The assertion that the comparison of traits in twin pairs could disclose a disposition for
having certain political traits and not completely different ones is thus based on an a
priori and untested assumption about what twin studies actually prove. Accordingly, this
paper addresses twin studies as a general method of comparing pair types of twins; it is
not a critique against a particular twin study method or experiment. None of the twin
study methods in political science has dealt with or tested the alternative explanation that
I propose in this paper.
This paper proceeds as follows. In section 2, I present an alternative biological
explanation to the common results of the standard twin research design that show greater
correlations of trait similarities in identical twins compared to non-identical twins. In
section 3, I discuss the problematic ways in which the concepts of ‘heritability estimates’,
‘environment’, and the ‘genetic factor’ are defined and used in twin studies. In section 4,
I shortly address the attempt that has been made to associate specific genes and
Feldman and Otto 1997; Keller, Medland, and Duncan 2009). For similar reasons, I do
not inquire into the recent literature on personality and politics that tends to cite and be
cited in twin studies (Gerber et al. 2010; Mondak et al. 2010; compare to Verhulst, Eaves,
and Hatemi 2012; and to Greenstein 1992; and Winter 2003).
5
chromosome regions with political traits. In section 5, I examine the wider relevance of
this analysis to results and puzzles in behavioral genetics. In section 6, I map the
operative implications of this paper to future research.
2. An Alternative Explanation to the Twin Studies
The twin studies method relies on comparisons of traits in monozygotic (MZ, or
identical) twins and dizygotic (DZ, or fraternal) twins. Each pair of twins is assumed to
be affected by an equal, yet special, environment and life history from the time of
inception. It is important to stress that the comparison of twins is not a direct genetic
comparison. We know nothing about the specific genetic makeup of the twins. We do
know that MZ twins are born from the same fertilized egg that split and produced two
embryos, and hence they are taken to be genetically identical. DZ twins are born from
two separate eggs that were fertilized by two different sperms and developed in the same
womb, and they share an average of fifty percent of their genes (identical by descent),
like any other pair of siblings (the proportion of genetic makeup of full siblings was
found to range between 0.374 to 0.617 percent; Visscher et al. 2006).
The results of greater trait resemblance in MZ compared to DZ twins have
produced heritability estimates for various traits (for reviews see Bouchard and McGue
2003; Johnson et al. 2009; Turkheimer 2000, 2004). Heritability is the proportion of
phenotypic variance in the population that can be attributed to genetic differences in that
population (see the discussion on heritability in section 3). Political scientists who used
twin studies have also presented heritability estimates for left and right wing ideological
preferences, and for political behavior such as voting turnout (Alford, Funk, and Hibbing
2005; Fowler, Baker, and Dawes 2008; Hatemi, Medland, and Eaves 2009). In this paper
6
I refer to those political preferences and behavior as ‘political traits’ for short. Authors of
these studies contend that the estimates indicate a genetic effect or cause in the form of
genetic structures (“political genotypes”) that produce distinct liberal or conservative
traits or “political phenotypes” (Alford, Funk, and Hibbing 2008). They argue that the
genetic effect is not deterministic but probabilistic. Yet this does not change the basic
argument that the hypothesized genetic effect is for a certain political preference or
behavior. This assertion is crucial to everything that follows from twin research (see
Alford, Funk, and Hibbing 2005: 155). I argue that there is indeed a biological (or
genetic) explanation for the twin study results but it is not the type of genetic effect which
is suggested in the literature.
The central question that must be raised in this context is why MZ twin pairs tend
to be more similar than DZ pairs in some of their political traits. I set aside non-genetic or
environment-only explanations and focus on the biological side of the argument. The
most logical explanation of the results of twin studies is that individuals with identical
genomes (e.g., MZ twins) react more similarly to the same particular environmental
effects compared to individuals with non-identical genomes (e.g., DZ twins) under
equivalent conditions, but the genomes do not narrowly determine the traits themselves
and equally enable a wide range of possible reactions and phenotypic expressions for the
exact same genome. In the context of the twin studies this would mean that MZ twins
tend to be more alike on average compared to DZ twins simply because the former are
identical (or at least much more genetically alike than DZ twins) and hence they are more
similarly affected by the same environmental effects but the content of those greater
similarities is nevertheless malleable and predominantly determined by particular
7
environments. When the environment changes, as is often the case, the expression and
content (e.g., political preference) of the trait in question could easily change as well. Put
differently, changing the environment would not necessarily affect the greater degree of
trait similarity in MZ compared to DZ twins (though it could certainly do that as well) but
it could have a significant impact on the content of the trait itself. This explanation makes
a distinction between the greater trait similarities in MZ twins, which is indeed due to a
simple genetic factor, and the separate issue of the trait content which is shaped and
altered by the environment. While this explanation is new in the twin studies context, the
basic notion underpinning it has been part and parcel of biological theory for decades.
As Richard Lewontin noted early in response to the debate about the heritability
of social traits, “A trait can have a heritability of 1.0 in a population at some time, yet
could be completely altered in the future by a simple environmental change” (1974: 400).
To illustrate this point in relation to the twin studies let us first take a simple example of
two pairs of plant seeds. The one pair is genetically identical, and the other pair is only
fifty percent alike. No matter which (equal) environment we plant these two pairs of
seeds, the identical pair will tend to appear more similar in the expression of their traits
(phenotypes) compared to the non-identical pair for the very simple fact that one pair is
identical and the other is not. For instance, the identical seeds will tend to exhibit greater
resemblance in terms of their eventual height compared to a greater variation in the
height of the non-identical seeds when each pair is influenced by equal environments. But
the height itself could vary significantly in different climates or altitudes or due to other
such unique conditions. The greater resemblance is indeed solely accounted for by the
greater genetic similarity of the identical seeds but the particular phenotype is not.
8
Lewontin’s paradigmatic concept of heritability and the example above rest on an
extensive body of research in branches of biology, primarily in developmental behavioral
genetics, developmental evolution, and ecological developmental biology, which have
established beyond any doubt that “[t]he same genotype can produce astonishingly
different phenotypes, depending on individual experience” and various environmental
effects (Bateson 1998: 161). Abundant examples from ecological developmental biology
show that the manifested traits (the phenotypes) of various species are not predictable
from their specific genetic makeup (their genotype) but crucially depend on the
ecological environment in which the organism is developing (i.e., “context dependency”;
Gilbert 2001). This dependency of varied trait expression is due to a wide array of
environmental cues, which are often unpredictable, during various stages of development.
“At one temperature, the snapping turtle embryo becomes male; at another temperature it
becomes female. Fed one diet, a female ant larva becomes a sterile worker; fed another
diet, the same larva becomes an enormous fertile queen” (Ibid: 4). Such phenotypic
expression of organisms due to environmental stimuli is also known as ‘developmental
plasticity’ and ‘phenotypic plasticity’ (Bateson et al. 2004; Gilbert 2001; Sultan 2003;
Price, Qvarnstrom, and Irwin 2003).
The argument that the same genotype can be expressed in very different
phenotypes is not a contentious issue in biology; it is more of a bedrock truth that we
must consider when interpreting the twin study results. Precisely given this accepted
biological premise we must ask: is it not in the genetic repertoire of the twins (and any
individual for that matter) to equally and just as easily have an opposite political trait?
After all, and unlike morphological traits, many political traits and preferences are clearly
9
malleable, are not costly to change, and could easily be modified in order to meet
changing circumstances (Price, Qvarnstrom, and Irwin 2003).3 An Answer to this
question is already suggested from within the twin study results. It should be recalled that
members of MZ pairs also have different traits under the same environmental effects in
all the twin studies. Their trait resemblance is only relative to members of DZ pairs on
average; it is not absolute. Different traits often do result from the same shared genotype
of any given pair of MZ twins. Only on statistical average do the traits of members of MZ
pairs tend to be more alike relative to members of DZ pairs. Furthermore, unlike the
studies in developmental biology that were noted above, twin studies do not examine the
adaptive political repertoires, the possible range of political reactions, or the degree of
behavioral plasticity of the individuals studied. Twin studies cannot solve these cardinal
issues and are unsuitable for this purpose because this design does not test the same
genotype in a variety of conditions (Gilbert 2001; Gottlieb 2003; Ross 2006; Sultan
2003). The untested assumption is that, on average, these political preferences will
remain the same in different environments due to a genetic effect (Lewontin 1974: 409).
It should be emphasized that the problem of the underlying causality assumptions
of twin studies cannot be resolved by improving statistical methods or by other twin
study designs. The heart of the problem is that comparing pairs of twins could not, in any
statistical form, support the hypothesis that there are genetic dispositions for having
certain political traits and not others.4 Arguably, this research design also cannot refute
3
On the other hand, stable political preferences, like one’s party identification or one’s
religion, are purely environmental (see also in Hatemi et al. 2009).
4
Compare also to the most advanced decomposition and multivariate model that does not
10
the hypothesis because even if in certain environments there were no co-twin differences
between MZ and DZ pairs on average, this would not mean that different genotypes
could not give rise to similar traits in one set of environments and to completely disparate
traits and correlations in others. At this point it should also become obvious that the real
null-hypothesis for the twin studies is to test whether twins (and other people) are
genetically disposed to having the traits they exhibit and not completely opposite ones for
the exact same genotype just as easily.
Yet twin studies involve an implicit assumption that the greater trait resemblance
among MZ co-twins compared to DZ co-twins is indicative that certain political traits are
hereditary in the population at large. In fact, the only possible way to generalize the twin
study results is by interpreting them to mean that the greater resemblance of political
traits among MZ twins is due to specific “political genotypes” (Alford, Funk, and
Hibbing 2005, 2008) or, put less crudely, due to some “genetic effects” (Hatemi et al.
2009a). If the greater similarity between MZ twins is simply due to a very broad and
malleable range of reactions of similar genotypes to random and special environments, no
conclusion can be drawn about innate dispositions for any political trait. Only if the traits
in the twin studies are interpreted and framed as resulting from “political genotypes” that
tend to shape similar political preferences in different environments (or that these genetic
effects have a narrow ‘norm of reaction’, Lewontin 1974: 404; Gilbert 2001; Sarkar
1999; Sultan 2003) does it make sense to argue that twins’ political preferences are
tests or solve this cardinal problem (Verhulst, Eaves, and Hatemi 2012). See also the
discussion in the next section about reared-apart twins as well as the problem of
calculating environmental effects into the genetic factor measurement.
11
genetically affected. And if twins’ specific political traits are genetically shaped, so
presumably are political traits in the population at large. This, again, is not what the twin
studies actually test. Thus the most obvious alternative explanation and null-hypothesis
are untested in these studies.
3. Heritability and Environment in the Twin Studies
Given the basic problems I identified above, this immediately raises the question
of what the heritability estimates in twin studies really mean and whether significant
estimates are at all indicators of a genetic basis of political traits. Heritability is a much
misunderstood and often misapplied concept (Visscher, Hill, and Wray 2008). In
common language (though not in formal definition) it denotes a trait that is passed down
from parents to offspring, such as eye or hair color, and is stable and not highly
susceptible to the environment. The terms inherited or hereditary are closer to this latter
meaning. But in genetics heritability is simply a statistical estimate that is technically
defined as “the proportion of phenotypic variance among individuals attributable to
genetic differences in that population” (e.g., Alford and Hibbing 2008: 184). As I will
show in this section, this definition of heritability, while useful for some purposes (e.g.,
plant and livestock breeding), is least appropriate, if relevant at all, to answer the question
of whether political genotypes exist. It should be stressed in advance that the limitations
pertaining to heritability that I will mention are common knowledge in biology. Some of
these limitations are also stated in the twin studies in political science, but not all of them
(notably in the seminal paper by Alford, Funk and Hibbing 2005). Therefore many
political scientists (and the media) often do not understand or misinterpret the results. But
12
beside these common problems, I will observe far less obvious and more crucial points
about the definition of the environment, which may help explain the twin study results.
The first thing to note about heritability is that it refers to variation of a trait
across individuals in a certain population (e.g., that of MZ relative to DZ twins). The
meaning of this technical definition is that heritability is a statistical estimate that can and
does change both between populations, and, more importantly, in the same population
under different environmental effects and time periods.5 Heritability estimates of this sort
give only a snapshot of trait variation between MZ and DZ twins, at a specific point of
time and therefore due to particular environmental effects. In Lewontin’s words, “the
result of the analysis has a historical (i.e., spatiotemporal) limitation and is not in general
a statement about functional [or causal] relations” (1974: 403). Such heritability estimates
can be easily modified to the point of statistical insignificance by environmental effects.
5
For example, one study has shown that DZ twins in the age cohort of 18-20 have greater
resemblance in their political attitudes relative to MZ twins of the same cohort. After this
point, the similarity correlations for MZ twins did not become stronger. Rather, the
similarities among DZ co-twins dropped (Hatemi et al. 2009b). What Hatemi and his
colleagues actually show is that heritability estimates can be produced not because MZ
become more similar in time (as they argue) but simply due to the fact that DZ twins
show less similarity in a new environment or later in life relative to MZ twins (see also
Suhay, Kalmoe, and McDermott 2007; Smith et al. 2012: 24, 28; see also discussion in
section 5). Based on this technical definition, the scholars argue that their result
demonstrates a genetic influence on the transmission of political attitudes in adult life or
that heritability increases with age (Ibid: 1152).
13
Thus, “high heritability [estimate] says little about how difficult it is to change a trait in
circumstances other than those in which it was measured” (Mameli and Bateson 2006:
165). High heritability estimates do not mean that a certain trait will remain similar or
stable in a different environment and at a later time, as opposed to what tends to be
popularly assumed and even argued from these studies (see Alford, Funk, and Hibbing
2005: 164-165; 2008; and contrast with Hatemi et al. 2009b).
The second point to note about the technical definition of heritability is that it
refers to variation of a trait between populations (inter-individual) at a certain point of
time and environment and not to the degree of variation of that same trait within an
individual (intra-individual) in all one’s developmental periods and in all different
environments (Molenaar and Campbell 2009; compare Hatemi et al. 2009b: 1143).
Namely, the heritability estimates do not say anything about an individual’s level of “trait
plasticity” so to speak. Twin study results do not say whether an individual was likely to
hold a political preference for only the environment in which it was measured and less or
more rigidly in later environments. Indeed, this is precisely why heritability estimates can
and often do change in a population sample in a later point of time depending on various
factors ranging from age, through the size of the group and methods of measurement, to
socio-economic status and ecological conditions (see also Turkheimer et al. 2003;
Visscher, Hill, and Wray 2008: 261). If the environment changes, the heritability measure
changes with it (Rose 2006: 526; Lewontin 1974). Yet the published literature often
interprets the results to say that individuals are genetically disposed to have certain traits
and not others. Alford, Funk, and Hibbing interpret their results in this way and also add
that this genetic factor explains why individuals are unable to persuade one another with
14
rational arguments (2005: 164-166; Alford and Hibbing 2008: 197). Similarly misleading
is to write in the abstract of a paper that “genes appear to play a pivotal role in shaping
the strength of an individual’s party identification” (Hatemi et al. 2009a: 584 – emphases
added).
The thorny issue is how to differentiate between the environmental causes of
variation of a trait on the one hand and the genetic factor of that trait on the other. The
scholars working in this field in political science acknowledge the importance of genes
and the environment, as well as the interaction between the two, in determining political
traits. A typical statement is as follows: “The underlying construct that is responsible for
a specific trait value (phenotype) is due to some combination of genetic and
environmental influences” (Hatemi et al. 2009: 585). Yet terms such as “some
combination” and “interaction” of genes and environment rest on certain crucial
assumptions which have also not been properly discussed in the political science
literature so far.
The first aspect to note is the definition of ‘environment’ in the twin models:
“‘Environment’ is all of the nongenetic external factors that influence trait variation
across a population. These influences range broadly from the earliest biological
environment of the womb, to the physical environment of a childhood house, to the social
environment of the adult workplace” (Alford, Funk and Hibbing 2005: 156, emphases
added; Alford and Hibbing 2008: 187; see also Plomin, Owen and McGuffin 1994:
1735). With such an all-encompassing and residual delineation of the environment, the
model, by definition, cannot test any substantial interaction with environmental factors.
What this model and definition do is to hold everything but the genetic difference
15
between MZ and DZ twins as one general constant factor (as recently admitted by Smith
et al. 2012: 18). This is done in order to isolate the genetic difference between MZ and
DZ twins as the genetic factor and thereby to calculate the percentage of the total trait
variance that can be associated with this factor in a specific spatiotemporal environment.
The causal importance of the genetic factor is then defined as that percentage, leaving
interaction with the environment unspecified and, in fact, non-existent in the model. The
‘environment’ only explains the variance part that cannot be statistically associated with
the genetic variable; it does not interact with the genetic component. The environment is
defined in such a way that it cannot explain any part of the variance that is already
associated with (or “explained by”) the genetic factor. The classic twin studies model is
thus based on independent variables, not interacting variables, each with its correlations
and separate explanatory power (see also Suhay, Kalmoe, and McDermott 2007).
This separation will of course result in heritability measures that are higher than
what really exists or does not exist. As Visscher, Hill, and Wray explain, “Heritabilities
can be manipulated by changing the variance contributed by the environment. This can be
as simple as changing the method of measurement” (2008: 261). One such way to obtain
statistically significant heritability measures between MZ and DZ twins is to define the
environment in this broad, residual way. This method is very prone to produce falsepositive results of heritability estimates for all sorts of traits which are not hereditary. For
example, Schonemann (1997) had shown using the ‘National Merit Scholarship twin
data’, that the question ‘did you sell a used textbook during the last year?’ resulted in
23% heritability for men and 32% for women. The question ‘did you have your back
rubbed?’ was 92% heritable for men and 21% for women; and ‘adult religious
16
preference’ was 85% heritable for men and only 21% heritable for women (Schonemann
1997: 104).
An additional consequence of the broad definition of the environment is that most
of these studies do not define, specify or test distinct environments that may have
affected the twins as having a certain political trait. Namely, we do not know if the twins
exhibit their political traits because of or despite being affected by a particular
environment. We also do not know if any two MZ twins who hold a democratic or
republication political preference do so because or despite of their environment or
because of their genotype (in most of the twin studies we know nothing about their
genotype). This point also raises the crucial question of what exactly is an environment
for having a certain political trait. It is a tricky issue to disentangle the relative weight of
various possible factors – such as family, friends, schooling, media, framing, culture,
identity, domains of self-esteem, social class and income, and random events at various
points of life – in defining an environment for having a certain political trait. For
example, being brought up in Texas rather than in New York is not enough in itself for
holding conservative views. Yet had we “switched” MZ and DZ twins growing up in
these two states, MZ would still likely be more similar than DZ twins but could
nonetheless have opposite political traits because their in-group identities and attitudes
are held in contrast to positions of various perceived out-groups (Brewer and Pierce
2005; Tajfel and Turner 1979; Turner 1982). The twin studies do not and cannot test this
possibility for the simple fact that it is impossible to switch or test the same pair of twins
(or any individual for that matter) between political environments in which there are at
least assumed opposite effects (see also Lewontin 1974: 409).
17
The closest possible method for testing the same genotype under different
environments is investigating the case of twins who have been separated at an early stage
of childhood and have been reared apart (see review in Bouchard et al. 2003). A greater
degree of trait similarity under these circumstances may suggest that the same genotype
results in similar traits despite developing under unequal environments, assuming that the
upbringing environments are indeed sufficiently dissimilar. There are several problems
with this argument, however. First, it is important to stress that the twin studies in
political science have not been empirically conducted on reared-apart twins but on a
comparison of MZ and DZ twins who have been reared together. None of these studies
compare political traits in twins who were reared apart in environments that are as
dissimilar as any two randomly chosen environments, let alone interaction of any
particular genotype with defined environments.
Second, a greater similarity of reared-apart MZ twins relative to DZ twins still
does not specify what is an environment for having a certain political trait. The studies of
twins reared apart from behavioral genetics (Bouchard and McGue 2003; Bouchard et al.
1990), which are mentioned and cited in political science in order to make the case for
the validity of reared together twin studies, do not resolve this complex issue because
they also do not attempt to define such trait-shaping environments. Third, the results of
reared-apart twins showed great difference compared to regular twin studies, and these
studies are hard if not impossible to replicate in political science (Medland and Hatemi
2009: 193). The limited number of studies of twins reared apart and, no less important,
the very small sample of twins participating in them (only several dozens MZ and DZ
twins) greatly limit their explanatory power and make them highly prone to false positive
18
and other confounding effects because these types of studies require samples in the
thousands to have any statistical significance. As pioneers of these studies acknowledge,
“Sample sizes were… small, and associated confidence intervals were wide, suggesting
the need for caution in interpreting individual correlation coefficients” (Bouchard and
McGue 2003: 28; see also Benjamin 2011: 72; Kendler 2005: 342; Martin, Boomsma,
and Machin 1997: 389; Medland and Hatemi 2009: 199; Waller et al. 1990). These
limitations are not a case against studying reared-apart twins. These limitations, however,
do show why relying on a few small samples of reared-apart twin studies in behavioral
genetics in order to validate results of reared-together twin studies in political science
(e.g., Alford, Funk, and Hibbing 2005: 155) is a problematic argument (see also Charney
2008: 303-304; Suhay, Kalmoe, and McDermott 2007).
On top of the points above there is another cardinal issue and assumption
involving the separation of causes of trait variation into genetic and non-genetic causes.
The assumption is that the causes of trait variation can be broken into these independent
factors, each contributing or adding separately to the outcome. The assumption that
genetic and environmental causes are independent from each other and that their relative
causal weight can be measured separately is known as the ‘additivity assumption’. And
indeed the models employed in the twin studies are additive models in which the effects
of genetic and environmental causes are measured separately from each other (e.g.,
Alford, Funk and Hibbing 2005: 158; Fowler, Baker and Dawes 2008: 234-235). It is
assumed that if the environment is defined and held as a constant factor, the heritability
estimate discloses a genetic cause for the trait.
19
Yet this additivity assumption greatly simplifies and distorts the meaning of
genes-environment interaction, as it is currently understood in the relevant branches of
biology, including behavioral genetics. This point was first made by Lewontin who
critically noted, “If an event is the result of the joint operation of a number of causative
chains and if these causes ‘interact’ in any generally accepted meaning of the word, it
becomes conceptually impossible to assign quantitative values to the causes of that
individual event. Only if the causes are utterly independent could we do so” (Lewontin
1974: 402). Indeed, it is widely acknowledged in biology that it is impractical to quantify
the relative contribution of genes versus the environment in the development of a specific
trait because “it is biologically impossible for gene and environment to operate
independently of one another” (Meaney 2001: 57). “Biological systems are complex,
non-linear, and nonadditive. Heritability estimates are attempts to impose a simplistic and
reified dichotomy (nature/nurture) on nondichotomous processes” (Rose 2006: 527).
The scholars working in this field are of course aware that genetic and
environmental causes are not separate. For example, they note, “Providing statistical
estimates of the effects of genetics, shared environment, and unshared environment is not
tantamount to asserting that these effects work in isolation from each other” (Alford,
Funk, and Hibbing 2008: 323). Fowler, Baker and Dawes similarly admit that “[s]ince the
variance components are not directly observable, the ACE model’s assumption of
additivity cannot be tested and more complicated relationships are possible. For example,
it is possible that genes interact with the environment (GxE) or with other genes (GxG) to
yield variation in behavior...” (2008: 235; see also Hatemi, Byrne, and McDermott 2012).
Yet the way that statistical associations of additive models are presented and over-
20
claimed in political science has given the false impression that the genetic and
environmental factors have been analytically separated when in reality they were not.
The limitations of heritability estimates and the additive model are more known
outside political science. What is more important is a derivative fatal flaw of this method
which is that the genetic estimate itself actually consists of a total and complete
conflation of an environmental effect on random genomes for which we know nothing
about. This is so because the genetic factor is defined and calculated as the degrees of
trait (or phenotypic) differences relative to the ratio of the twins’ genetic similarities, but
the trait differences themselves are already affected by environmental factors that have
not been separated. The genetic factor therefore includes both the genetic proportion
between the twin types and various random environmental effects on the expression of
their traits. The method concurrently inflates the heritability estimate and creates a false
nature-nurture partition. Put simply, environmental effects are calculated as genetic
effects. Leading scholars in this field relate to this point and explain that in “a univariate
model, additive genetic variance will include all the genetic influence from all covariates,
some part of gene-environment covariation if it exists, and some part of gene environment
interaction, if it exists” (Hatemi, Dawes, et al. 2011: 74 – emphases added). They also
admit that this model simplifies a far more complex interaction and reality (Ibid). What
they apparently fail to see is that this ‘covariation’ and ‘interaction’ are not minor
possibilities (“if”) or mere noise in the measurement of the genetic factor. This
covariation, or rather conflation, exists by definition and is built into the model because
the genetic factor already includes and measures environmental effects. This point is
crucial because it is at the heart of what the defined genetic factor actually measures in
21
the twin studies. At minimum, it produces higher heritability estimates than real by
calculating environmental effects as the genetic factor; at worse, it produces false
positive results of a genetic effect where it does not actually exists. 6
Heritability estimates might therefore convey the wrong notion that the
heritability estimate is the amount of causal weight of the genetic factor and that this
factor then interacts with the environment in whatever residual percentage of the
unexplained variance. As the observation above suggests, the genetic factor is already an
expression and confusion of the genetic and environmental factors. Furthermore, it is well
known that a heritability estimate of 0.8 does not mean that the trait is eighty percent
genetic and twenty percent environmental (Charney 2008: 300; Gottlieb 2003: 338; see
also Hatemi, Byrne, and McDermott 2012). It merely suggests that a genetic effect may
exist but it says nothing about the relative importance of the genetic factor and the nature
of its interaction with the environment in causing the actual trait: to what degree the
genetic-factor estimate of X percent for a given political trait is important, a negligible
1% or a crucial 99%; how exactly the genetic factor interacts with the environment; and
6
Although they stress the importance of gene and environment interaction, scholars
working in this field are aware that the twin study model does not actually capture this
interaction on the one hand, and does not account for the possibility of such interaction
and conflation in the definition and calculation of the genetic factor on the other. They
admit that more advanced multivariate models are needed for this purpose and that these
studies are forthcoming (Smith et al. 2012; Hatemi et al. 2011: 72). As I noted above,
however, the problem and limitation of deriving conclusions on genetic effects from the
method of comparing MZ and DZ twins applies to all twin study models.
22
how the environment affects it. This limitation is well known but the twin study scholars
in political science do not resolve these issues of relative weight and types of interaction,
other than reiterating that the topic is indeed complex and requires further research.
Instead, they argue that if statistical associations in the additive model disclose any role
of possible genetic involvement, “then genes are also likely to play a role in more
complex specifications” (Fowler, Baker and Dawes 2008: 235; see also Hatemi, Byrne,
and McDermott 2012).
This type of argument vastly over-claims the findings. Similar to their limits in
social science (Mahoney 2001; Rustow 1970: 342), statistical associations (or
correlations) in behavioral genetics are essentially statistical claims and initial and limited
tools to test hypotheses. They are not biological proofs of causation and certainly do not
confirm a causal path from any set of genes to a political trait (in the context of
psychiatric genetics see Kendler 2005: 1245). Namely, even a relatively high heritability
estimate does not mean that the environment does not predominantly or entirely shape
any given political preference. More importantly, as I observed above, the problem runs
deeper than the obvious limitations of correlations; it goes to the heart of what the genetic
factor actually measures in twin studies due to the way it fuses and confuses the ratio of
genetic differences between MZ and DZ twins and their respective degrees of trait
similarities (which are already affected by the environment) as a single genetic factor.
Finally, the inference and explanatory limitations of association studies and
heritability estimates in twin studies have also been recently acknowledged by leading
behavioral geneticists who extensively employed these studies over the years. We are at a
stage where scholars in behavioral genetics admit that twin studies are no longer needed
23
precisely because no inferences can be drawn from them about the actual degree,
direction and type of genetic and environmental interactions. As Johnson, Turkheimer,
Gottesman, and Bouchard have recently noted, “To understand why heritability estimates
are no longer important, it is necessary to understand that they are completely dependent
on the specifics of the samples and environmental conditions from which they are taken”
(2009: 217). Those pioneers of twin studies of course do not deny the role of genes in
behavior but rather that “little can be gleaned from any particular heritability estimate and
there is little need for further twin studies investigating the presence and magnitude of
genetic influences on behavior” (Ibid: 218). In this sense, twin studies and heritability
estimates cannot provide the sort of explanations that are sought by political scientists
interested in the complex causal relations that underlie political preferences and behavior.
4. Genes for Specific Political Traits?
A number of studies have taken further steps since 2008 to identify physiological
mechanisms and specific genes that may determine political traits. Oxley et al. (2008)
argue that political attitudes are affected by a genetic tendency to fear; Fowler and Dawes
(2008) maintain that they have identified two genes that code for voting turnout, and one
gene that affects partisanship (Dawes and Fowler 2009); and Settle et al. (2010) report
that a certain dopamine gene variant (DRD4-7R) seems to make a person more liberal in
early adulthood if the person who carries the variant had more than six friends in
adolescence. The political science scholarship has remained rather silent about such
studies until now while none of them has been replicated or validated.
Oxley et al. (2008) argue that sensitivity to threat, and conservative and liberal
political attitudes involve neural activity patterns around the amygdala. Because genes
24
affect the amygdala they contend that political attitudes are also affected by genetic
factors that affect fear. Yet, the distinction between those who hold different political
attitudes is not whether they are more or less inclined to fear, or have greater concern for
the social order “but rather, that they perceive different things as constituting threats to
that order” (Charney 2009: 5). The question is to what extent a person’s cognition
triggers, or is triggered by, certain physical processes. Genes only code for and regulate
protein production, they do not directly shape political preferences. Genes that regulate
the production of testosterone, for example, may lead one to be aggressive toward
conservatives or democrats, religious or secular people, men or women, one ethnic group
or another, depending on one’s social identity complexity and the perceived out-group in
a specific context (Brewer and Pierce 2005; Tajfel and Turner 1979). The environment
and cognition are major causes both in activating genes and in channeling physical
processes to various and possibly opposite political ends, for the exact same genotype.
Furthermore, physiologic manifestations of cognitive correlates, like electrical
and chemical activity in the brain, sweaty hands or eye blinks, do not constitute proof that
a specific political trait is genetic. These physiological processes are enabled and
constrained by the human genome in a broad sense, as all human traits are, but this
obviously does not mean that any social or political attribute is genetically affected or
heritable. The fact that the amygdala is associated with fear does not mean that it, and the
genes that affect it, constitute the source of political attitudes in the brain. It has been
well-established that any given brain structure, including the amygdala, may have a role
in various types of behaviors; and even apparent basic types of behavior and cognition
depend on interactions between different brain structures. “Just as there is no single
25
language “organ” in the brain, neither will there be a single political organ in the brain”
(Lieberman, Schreiber, and Ochsner 2003: 695-696).
The argument for versions of genes that affect political behavior such as voting
turnout (Fowler and Dawes 2008) and partisanship (Dawes and Fowler 2009) is based on
the association of a version of a gene (allele) and a certain political trait, a method known
as ‘genetic association.’ The logic behind this method is that a statistical correlation
between gene variants and traits may disclose some sort of relationship between the two,
although genetic association does not in itself entail a causal pathway from any gene to
shaping the political trait (see also, Fowler and Dawes 2008: 580; Settle et al. 2010:
1195). Given the current understanding in genetics, and in light of the evidence brought
forward so far in political science, the results of these studies are highly suspect of being
cases of false positives and must be evaluated with great caution, as will be explained.
It has been widely accepted in the last decade that the crucial majority of traits are
affected by multiple-genes (i.e., polygenic traits), as opposed to monogenic traits
whereby a single gene affects a single trait, which is quite rare in the population (e.g.,
(Tay-Sachs, Cystic fibrosis). Basic traits, such as skin color, height, body weight, and
disorders such as autism, diabetes, cancer, and psychiatric disorders are known to involve
many genes and genomic regions in complicated constellations, and such phenotypes are
also susceptible to environmental effects (Kendler 2005). Furthermore, “[s]ince the
proteins produced by genes generally appear to have multiple effects, most of which we
have barely begun to understand, it seems unlikely that we can be confident about all of
the consequences of any particular genotype in the foreseeable future” (Benjamin 2011:
68). The surprisingly low degree of genetic factors that have so far been associated with
26
common traits and diseases continues to puzzle geneticists (Maher 2008, 2010; Manolio
et al. 2009). Social behavior is even further complicated with a mix of various
interactions and causes. Fowler and Dawes (2008: 581) also acknowledge that behavioral
traits in politics are not coded in the genome and are shaped by a multitude of forces.
Thus in the context of existing theoretical and empirical research on common
traits and social traits, arguments that one gene affects partisanship, two genes affect
highly cognitive and environmentally sensitive political behavior such as voting turnout,
and a variant of dopamine gene shapes liberal ideology are very unusual, notwithstanding
the supposed controls and moderating affects of limited environmental factors. Even if
genes did affect political traits we would expect to find far more complicated
relationships. More important, beyond the fact that the studies have not been replicated or
validated, some of these candidate genes have received rigorous review and evaluation.
For example, the impact of both of the candidate genes (5-HTT and MAOA) that Fowler
and Dawes chose to associate with political behavior, in light of earlier studies that
associated them with social behavior, have been reevaluated. With regard to 5-HTT, a
meta-review study revealed that research results so far “are still compatible with chance
findings” and that even in depression, previously assumed to be directly affected by 5HTT, the effect “is very small or negligible” and “the possibility that the polymorphism
has no association with the disorder at all has not been excluded” (Munafo et al. 2009:
216). Other research similarly concluded that the “widespread acceptance of these
findings is likely to have been in part attributable to the acclaim the original article [Caspi
et al. 2003] received, as well as a field that was eager for a new approach due to the
frustrating lack of progress in gene identification for mental disorders despite intensive
27
efforts for more than a decade” (Risch et al. 2009: 2469). Investigation of the association
of the MAOA gene and anti-social behavior has shown mixed and conflicting results
(Prichard et al., 2008; Weder et al. 2009; compare to Caspi et al. 2002 and Fergusson et
al. 2011; and see also MAOA and depression, Huang et al. 2009).
It appears that Fowler’s and Dawes’ results about the assumed connection
between candidate genes and political traits are additional cases of random variations or
false-positive associations, especially in light of the fact that their choice of these genes
was due to those earlier studies that now lack validation (see also Hatemi et al., 2011:
272). Similar non-reproducible ‘false positives’ and chance findings have been reported
for dozens of traits and disorders, many of which are far more basic (e.g., physiologic) or
common than political traits (Joseph 2010; Munafo and Flint 2009; Munafo et al. 2009;
Risch et al. 2009). In light of the inability so far to find and isolate the causal and
predictive role of genes in physiologic traits, “the challenge is likely to be at least as large
for behavioral [political] traits where the causal mechanisms are arguably more complex”
(Benjamin 2011: 69). The neurological evidence about multiple roles of brain structures
and various interacting neurological networks in the brain (Lieberman, Schreiber, and
Ochsner 2003), which are clearly affected by many genes, also suggest why it is unlikely
that political traits are associated with specific genes in any meaningful analytic way.
Fowler’s and Dawes’ (2008) study was critically examined by Charney and
English (2012) in paper just published in the American Political Science Review. Charney
and English illustrate that Fowler’s and Dawes’ findings are indeed non-reproducible and
in some cases even yield opposite results than those reported by Fowler and Dawes.
Furthermore, they found several problems in the design of the study including the
28
specification of the phenotype (i.e., political traits); population stratification of ethnic
groups; the classification of the genotype and the choice of alleles; and the lack of
independence of the cases and their controls whereby, among other things, a larger
proportion of MZ and DZ twins (than that which exists in the general population) was
included in the sample and therefore confounded the results.
It is quite uncommon at the present stage of scholarship in genetics to argue for a
connection between single genes and social traits, even when supposedly controlling for
moderating environmental factors. Behavioral genetics has moved away from seeking
such a simple connection between genes and traits (Rose 2006; Todd 2006). Several
behavioral geneticists publishing in political science have also admitted that such studies
have not proven useful in terms of explanatory power and replication and that “there may
be no gene for a specific issue preference or ideological orientation” (Smith et al. 2012:
18). Instead they have recently offered a genome-wide approach that seeks to identify
regions on the genome which could be associated with political traits (Hatemi, Gillespie,
et al. 2011). Given the very early stage of this research, the fact that it has not been
replicated and its silence about interactions with the environment, it too lacks explanatory
power about political traits and behavior (see also Charney and English 2012: 11).
Furthermore, in this latter type of research “it is not generally recognized that
QTL [genome-wide] analysis itself relies on a prior assumption of significant
heritability” (Rose 2006: 526). For example, Hatemi, Gillespie, et al. (2011) seem to
assume that the political positions and lifestyles obtained in questionnaires from subjects
between 1988-1990 could not be different or opposite at some earlier or later point in the
subjects’ life or at the same time they were taken but under different environmental
29
conditions. Regardless of the plausibility of this assumption, it is not what this study
tests. The mapping of the genome relative to political preferences and life styles at one
point (or points) in one’s life cannot prove or even refute the hypothesis. If environmental
effects changed these traits (as they often do), the associations with the chromosome
regions would disappear or could even be in an opposite relationship. The contrary is also
true: The lack of correlation of a trait with a specific region does not mean that such a
correlation could not be established in other conditions or at a later point in time. Similar
to single candidate gene associations, though to a lesser degree, the genome-wide
associations are also prone to false positives, especially regarding highly malleable
characteristics such as political traits (on the genome-wide technique see Manolio 2010).
The mapping of the genome in this way also does not say how easy it is to change the
political traits in question relative to the many chromosomal regions that are identified,
nor does it show any interaction or moderation effects of the environment on these traits
and on gene expression, effects that are known to be highly important.
5. Wider Implications
Interestingly, reports of heritability measures in twin studies are often higher for
personal characteristics and political preferences than for medical disorders and other
physiological traits. For example, Happiness (self-reported wellbeing, 0.8), General
Intelligence and IQ (0.52 and 0.65~), Divorce (0.52), Extraversion (0.51), neuroticism
(0.46), vocational interests in adolescence (0.42) and scholastic achievements in
adolescence (0.38) have higher heritability estimates than memory (0.22), and processing
speed (0.22) in twin studies (Plomin, Owen and Mcuffin 1994: 1734-35; McGuffin,
Riley, and Plomin 2001: 1232; McGue and Lykken 1992; Lykken and Tellegen 1996).
30
Political preferences and traits show similar oddities. Martin et al. (1986) were first to
measure heritability estimates for positions on ‘death penalty’ (0.51), ‘Apartheid’ (0.43),
‘censorship’ (0.41), ‘white superiority’ (0.40), ‘disarmament’ (0.38), and ‘women judges’
(0.27); but oddly no less significant estimates for positions on ‘Jazz’ (0.45), ‘military
drill’ (0.40) ‘conventional clothes’ and ‘white lies’ (0.35), ‘fluoridation’ (0.34), ‘selfdenial’ (0.28), ‘computer music’, ‘teenage drivers’ and ‘learning Latin’ (0.26). Similarly,
Alford, Funk and Hibbing (2005: 159) reported heritability estimates of 0.41 for political
positions on ‘school prayer’ and ‘property tax’; 0.39 heritability for ‘astrology’; and 0.26
heritability for positions on ‘busing’, ‘nuclear power’, ‘democrats’ and ‘divorce’; and
0.25 for positions on ‘abortion’ and ‘modern art’. In striking comparison, medical
disorders such as hypertension, ischemic heart disease, cancer and Parkinson’s disease
have modest to negligible estimates in twin studies and none above 0.2 heritability
estimates (Plomin, Owen, and McGuffin 1994). Clearly something is wrong about these
higher heritability results for political preferences over common medical conditions. If
the genes supposedly involved in the expression of political (polygenic) traits are indeed
not independent of the environment, as twin study scholars admit (e.g., Fowler and
Dawes 2008: 581), heritability estimates for political traits must therefore be higher
rather than significantly lower compared to the expression of biological conditions.
In fact, the heritability estimates of traits in twin studies are greatly at odds with
actual studies of the genome that reveal very modest to tiny effects of genes on those
same traits. Whereas twin studies found heritability estimates of more than ninety percent
for autism and more than eighty percent for schizophrenia, the identification of various
gene variants and their associations with those traits did not yield anything close to these
31
results, a gap which is also known as the “missing heritability” puzzle (Maher 2008,
2010; Manolio 2010; Manolio et al. 2009). Accordingly, doubts have been raised about
the actual accuracy and utility of twin studies’ heritability estimates (Maher 2008: 21).
The contrasts between the heritability estimates of common medical conditions and
political positions strongly suggest that the latter are either inflated or simply false
positives. The alternative explanation that I have offered to the results of the twin studies
in political science, thus, may have broader implications to explaining these puzzles.
First, it could now be seen why comparison of twins exaggerates estimates for
certain conditions. For example, a certain medical condition could appear in a MZ pair
because they have a unique genotype which is responsible for it. Alternatively, the
medical condition could appear in the MZ pair because they were adversely affected by
the mother’s condition during pregnancy or due to other environmental effects after
pregnancy, despite not having that genotype (i.e., a non-genetic congenital condition).
The comparison of twins cannot distinguish between the causes and it will inflate the
heritability estimate by confounding and conflating the environment effect as a genetic
impact because MZ twins will show greater resemblance than DZ twins regardless of
whether the source of the condition is genetic or due to random (non-genetic) birth
problems. This may explain the gap that was noted above between the high heritability
estimates of twin studies and actual genetic research.
Second, twin studies can produce heritability estimates that are lower than they
actually are in common traits. I will illustrate this through an examination of the
Falconer’s formula of heritability (Falconer and MacKay 1996), which is to subtract the
trait correlation of DZ from that of MZ pairs and multiply by two the resulting difference
32
[h = (MZ – DZ) * 2]. Falconer’s formula is employed, for example, in the seminal paper
of Alford, Funk, and Hibbing (2005) in political science, and well beyond.7 A genetic
component for having a common physiological trait such as acne, hair loss, or one nose,
is likely to have very similar trait correlations in MZ and DZ twins. In such cases when
correlations are similar (MZ = DZ) or almost similar, the result is heritability estimates of
zero (MZ – DZ = 0, and 0 * 2 = 0) or low and insignificant estimates. The result implies
that there is no genetic basis for these common traits when this is clearly not the case
(Jerry Hirsch made a similar argument, see in Schonemann 1997: 100).
Third, and most relevant to the social sciences, the alternative explanation I
proposed in this paper also suggests why certain social and political traits can have higher
heritability estimates than certain common traits and conditions in twin studies. This is
due to the simple fact that DZ twins react more differently to the same environmental
effects compared to MZ twins. In reality, the twins can equally express opposite political
preferences under different environments. Yet when we compare MZ and DZ twins under
equal environmental impacts, this would result in emphasizing and accentuating their
relative trait differences. This explanation is consistent with, and strengthened by, the
findings that equal environments often make DZ twins growing up in the same family
different rather than more similar to each other; and when MZ twins growing up in the
same family also show difference in their traits the reason is normally nonshared
7
Whereas there are twin studies in political science that employ different and more
sophisticated methods than the Falconer formula (Medland and Hatemi 2009; Fowler,
Baker, and Dawes 2008), this analysis equally applies to them because they are based on
the same general method of comparing traits of MZ and DZ twins (see Smith et al. 2012).
33
environmental factors (Plomin 1989; Plomin, Owen and Mcuffin 1994: 1735-36).
Scholars who are using twin studies in political science explicitly admit this in a paper
just published, where they note that MZ twins “share precisely the same susceptibility to
environmental influence” while the reaction DZ twins to similar environments
“produce[s] variations in the phenotype” (Smith et al. 2012: 28 – original emphasis).
What they fail to conclude, however, is that the crucial issue is not the “susceptibility” of
any gene or genotype, but simply the obvious point that MZ will always tend to respond
more alike to similar environments, no matter what they are, compared to a similar
experiment with DZ twins. This explains both why political traits have such inflated (and
hence false) heritability estimates and why these estimates are unreasonably higher than
those of certain common physiological traits and conditions.
While many aspects of human behavior are certainly affected and shaped by
genetic factors, the question remains whether twin studies actually reveal a hereditary
basis of political traits. If the proposed explanation of this paper is correct, it will require
rethinking the broader meaning of the twin study results not in political science alone.
6. Conclusion and Operative Implications for Future Research
It is likely that the scholars who are working with twin studies will not agree with
my alternative explanation to twin study results and that they may want to address this
paper in some way. In this section I summarize the key arguments and note where they
disagree with the published literature. The purpose of this section is also to focus the
debate on the central issues and reduce the risk of sidetracking it into minor points. I also
specify what are the operative implications for future research, at least in my view.
34
First, the most important argument of this paper is that there is indeed a biological
explanation to the twin study results but it is not the explanation that has been proposed
in the literature so far. I argued that MZ twins are often more alike than DZ twins because
similar genomes tend to react more similarly to similar effects, no matter what these
effects are and what the traits are. I made a distinction between the explanation of trait
similarity (which is due to a simple genetic factor) and the completely different issue of
explaining trait content (e.g., a political preference) which is shaped only by the
environment. Given that it is a bedrock truth in biology (accepted also by those working
in political science) that the exact same genotype can give rise to diverse and even
opposite traits, the twin studies do not and cannot test the hypothesis that they set to
prove, namely, that there is a genetic effect on political attitudes. What these studies
show is that trait similarity is indeed accounted for because identical genomes tend to
react similarly to a similar environment. However, scholars have been wrongly
interpreting this obvious result as an indication of a genetic effect on shaping the content
of political traits. The operative implication for future research is to show that in twin
studies the environment is not actually the sole determinant of the trait content. Future
studies need to find a way to separate the explanation of trait similarity from that of trait
content and refute this null-hypothesis. As I explain above, this will not be an easy
challenge but it has implications beyond political science in reference to the missing
heritability puzzle that was noted above.
Second, I argued that the measurements in twin studies both exaggerate the
genetic effect and confound between the genetic and environmental effects. The
definition of the environment in the classic twin studies leaves everything but the ratio of
35
genetic similarity between MZ and DZ twins as a constant factor (Smith et al. 2012: 18).
This research design cannot test substantial interactions with environmental factors in the
degree of complexity that is known to exist in political science. The ‘environment’
component only accounts for the variance part that cannot be statistically associated with
the genetic variable. Furthermore, a major problem is that the genetic estimate itself
actually consists of a total and complete conflation of an environmental effect on random
genomes for which we know nothing about. This is so because the genetic factor in twin
studies is in fact the degrees of trait differences relative to a ratio of genetic similarities,
but the trait differences themselves are already affected by environmental factors that
have not been separated. This creates a false nature-nurture partition. The combination of
the two problems yields inflated heritability estimates and false positive results. In order
to demonstrate a genetic effect on political traits, future studies must find a way to define
both the genetic factor and the environment in a more sophisticated manner and to
disentangle the two separate effects. Twin studies do not appear to be the right tool for
this purpose, in any statistical form, as I explained in section 2.
Third, I indicated that in twin studies the heritability measurements for political
preferences (many higher than 0.4) are often more than twice than those of many medical
disorders and other physiological traits (many lower than 0.2). Political traits are clearly
affected by multiple genes (polygenic traits). There is a wide consensus about this fact,
including by scholars working to integrate biology and political science. There is clearly
something suspect about the fact that heritability measures for political traits are greater
than the expression of biological conditions. The analysis of the confluence of the genetic
and environmental factors in twin studies may explain why comparison of twins
36
exaggerates estimates for certain conditions where the source may be either genetic or
environmental. I suggested (using Falconer’s formulae) that twin studies are prone to
produce heritability estimates that are lower than they actually are in common traits
where the differences between twins are not substantial yet have a genetic source.
Finally, the alternative explanation of this paper explains why the twin study model is
prone to yield high heritability estimates and false positive results for many political
preferences that most likely do not have a genetic basis. The implication of this goes
beyond political science. Solving these puzzles may also help solve the missing
heritability mystery of many basic physiologic traits in biology.
37
References:
Alford, John, Carolyn Funk, and John Hibbing. 2005. Are Political Orientations Genetically
Transmitted? American Political Science Review 99 (2):153-67.
Alford, John, and John Hibbing. 2004. The Origin of Politics: An Evolutionary Theory of
Political Behavior. Perspectives on Politics 2 (4):707-23.
Alford, John R., Carolyn L. Funk, and John R. Hibbing. 2008. Beyond Liberals and
Conservatives to Political Genotypes and Phenotypes. Perspectives on Politics 6 (2):3218.
Bateson, Patrick. 1998. Genes, environment and the development of behaviour. In The Limits of
Reductionism in Biology. Chichester: Novartis Foundation.
Bateson, P., D. Barker, T. Clutton-Brock, D. Deb, B. D'Udine, R. A. Foley, P. Gluckman, K.
Godfrey, T. Kirkwood, M. M. Lahr, J. McNamara, N. B. Metcalfe, P. Monaghan, H. G.
Spencer, and S. E. Sultan. 2004. Developmental plasticity and human health. Nature 430
(6998):419-21.
Beckwith, Jon, and Corey A. Morris. 2008. Twin Studies of Political Behavior: Untenable
Assumptions? Perspectives on Politics 6 (4):785-91.
Benjamin, Daniel. 2011. A White Paper by Daniel Benjamin. In Genes, Cognition, and Social
Behavior: Next Steps for Foundations and Researchers, ed. A. Lupia. Arlington, VA.:
National Science Foundation.
Biuso, Emily. 2008. Genopolitics. The Times Magazine, December 12.
http://www.nytimes.com/2008/12/14/magazine/14Ideas-Section2-B-t-007.html.
38
Bouchard, T. J., D. T. Lykken, M. Mcgue, N. L. Segal, and A. Tellegen. 1990. Sources of
Human Psychological Differences - the Minnesota Study of Twins Reared Apart. Science
250 (4978):223-8.
Bouchard, T. J., and M. McGue. 2003. Genetic and Environmental Influences on Human
Psychological Differences. Journal of Neurobiology 54 (1):4-45.
Bouchard, T. J., N. L. Segal, A. Tellegen, M. McGue, M. Keyes, and R. Krueger. 2003.
Evidence for the construct validity and heritability of the Wilson-Patterson conservatism
scale: a reared-apart twins study of social attitudes. Personality and Individual
Differences 34 (6):959-69.
Brewer, M. B., and K. P. Pierce. 2005. Social Identity Complexity and Outgroup Tolerance.
Personality and Social Psychology Bulletin 31 (3):428-37.
Brown, Donald E. 2004. Human Universals, Human Nature & Human Culture. Daedalus 4:4754.
Brown, Donald E. 1999. Human Universals. In The MIT Encyclopedia of the Cognitive Sciences,
ed. R. A. Wilson and F. C. Keil. Cambridge, Mass: MIT Press.
Brown, Donald E. 1991. Human Universals. New York; London: McGraw-Hill.
Caspi, A., J. McClay, T. E. Moffitt, J. Mill, J. Martin, I. W. Craig, A. Taylor, and R. Poulton.
2002. Role of genotype in the cycle of violence in maltreated children. Science 297
(5582):851-4.
Caspi, A., K. Sugden, T. E. Moffitt, A. Taylor, I. W. Craig, H. Harrington, J. McClay, J. Mill, J.
Martin, A. Braithwaite, and R. Poulton. 2003. Influence of life stress on depression:
Moderation by a polymorphism in the 5-HTT gene. Science 301 (5631):386-9.
Charney, Evan. 2008a. Genes and Ideologies. Perspectives on Politics 6(2):299-319.
39
Charney, Evan. 2008b. Politics, Genetics, and "Greedy Reductionism". Perspectives on Politics
6(2):337-43.
Charney, Evan, and William English. 2012. Candidate Genes and Political Behavior. American
Political Science Review 106(1): 1-34.
Charney, Evan. 2009. Physiology may not be (political) destiny. PsyCrit, February 25, 2009.
Dawes, Christopher T., and James H. Fowler. 2009. Partisanship, Voting, and the Dopamine D2
Receptor Gene. Journal of Politics 71 (3):1157-71.
Easton, David. 1976. The Relevance of Biopolitics to Political Theory In Biology and Politics,
ed. A. Somit. Paris: Mouton.
Falconer, Douglas S., and Trudy F.C. Mackay. 1996. Introduction to Quantitative Genetics (4th
ed.). Essex, UK: Longmans Green, Harlow.
Feldman, M. W., and S. P. Otto. 1997. Twin studies, heritability, and intelligence. Science 278
(5342):1383-4.
Fergusson, D. M., J. M. Boden, L. J. Horwood, A. L. Miller, and M. A. Kennedy. 2011. MAOA,
abuse exposure and antisocial behaviour: 30-year longitudinal study. British Journal of
Psychiatry 198 (6):457-63.
Fowler, James H., Laura A. Baker, and Christopher T. Dawes. 2008. Genetic Variation in
Political Participation. American Political Science Review 102 (2):233-48.
Fowler, James H., and Christopher T. Dawes. 2008. Two Genes Predict Voter Turnout. Journal
of Politics 70 (3):579-94.
Fowler, James H., and Darren Schreiber. 2008. Biology, Politics, and the Emerging Science of
Human Nature. Science 322 (5903):912-4.
40
Gerber, A. S., G. A. Huber, D. Doherty, C. M. Dowling, and S. E. Ha. 2010. Personality and
Political Attitudes: Relationships across Issue Domains and Political Contexts. American
Political Science Review 104 (1):111-33.
Gilbert, S. F. 2001. Ecological developmental biology: Developmental biology meets the real
world. Developmental Biology 233 (1):1-12.
Gottlieb, G. 2003. On Making Behavioral Genetics Truly Developmental. Human Development
46 (6):337-55.
Hatemi, Peter K., Enda Byrne, and Rose McDermott. 2012. What is a ‘Gene’ and Why Does it
Matter for Political Science? Journal of Theoretical Politics [forthcoming].
Hatemi, P. K., C. T. Dawes, A. Frost-Keller, J. E. Settle, and B. Verhulst. 2011. Integrating
Social Science and Genetics: News from the Political Front. Biodemography and Social
Biology 57 (1):67-87.
Hatemi, P. K., N. A. Gillespie, L. J. Eaves, B. S. Maher, B. T. Webb, A. C. Heath, S. E.
Medland, D. C. Smyth, H. N. Beeby, S. D. Gordon, G. W. Montgomery, G. Zhu, E. M.
Byrne, and N. G. Martin. 2011. A Genome-Wide Analysis of Liberal and Conservative
Political Attitudes. Journal of Politics 73 (1):271-85.
Hatemi, P. K., J. R. Hibbing, S. E. Medland, M. C. Keller, J. R. Alford, K. B. Smith, N. G.
Martin, and L. J. Eaves. 2010. Not by Twins Alone: Using the Extended Family Design
to Investigate Genetic Influence on Political Beliefs. American Journal of Political
Science 54 (3):798-814.
Hatemi, P. K., J. R. Alford, J. R. Hibbing, N. G. Martin, and L. J. Eaves. 2009a. Is There a
"Party" in Your Genes? Political Research Quarterly 62 (3):584-600.
41
Hatemi, P. K., C. L. Funk, S. E. Medland, H. M. Maes, J. L. Silberg, N. G. Martin, and L. J.
Eaves. 2009b. "Genetic and Environmental Transmission of Political Attitudes Over a
Life Time." Journal of Politics 71 (3):1141-56.
Hatemi, P. K., S. E. Medland, and L. J. Eaves. 2009. Do Genes Contribute to the "Gender Gap"?
Journal of Politics 71 (1):262-76.
Hines, Samuel M. 1982 Politics and the Evolution of Inquiry in Political Science. Politics and
the Life Sciences 1 (1):5-37.
Huang, S. Y., M. T. Lin, W. W. Lin, C. C. Huang, M. J. Shy, and R. B. Lu. 2009. Association of
monoamine oxidase A (MAOA) polymorphisms and clinical subgroups of major
depressive disorders in the Han Chinese population. World Journal of Biological
Psychiatry 10 (4):544-51.
Johnson, W., E. Turkheimer, I. I. Gottesman, and T. J. Bouchard. 2009. Beyond Heritability:
Twin Studies in Behavioral Research. Current Directions in Psychological Science 18
(4):217-20.
Joseph, Jay. 2010. The Genetics of Political Attitudes and Behavior: Claims and Refutations.
Ethical Human Psychology and Psychiatry 12 (3):200-17.
Keller, M. C., S. E. Medland, and L. E. Duncan. 2010. Are Extended Twin Family Designs
Worth the Trouble? A Comparison of the Bias, Precision, and Accuracy of Parameters
Estimated in Four Twin Family Models. Behavior Genetics 40 (3):377-93.
Kendler, Kenneth S. 2005. "A Gene for...": The Nature of Gene Action in Psychiatric Disorders.
American Journal of Psychiatry 162 (7):1243-52.
Lewontin, R. C. 1974. Analysis of Variance and Analysis of Causes. American Journal of
Human Genetics 26 (3):400-11.
42
Lieberman, M. D., D. Schreiber, and K. N. Ochsner. 2003. Is Political Cognition Like Riding a
Bicycle? How Cognitive Neuroscience Can Inform Research on Political Thinking.
Political Psychology 24 (4):681-704.
Losco, Joseph. 1982. Biopolitics: On Evolution in the Inquiry into Political Science - a response
to Hines 1982. Politics and the Life Sciences 1 (1):18-9.
Lupia, Arthur. 2011. Genes, Cognition, and Social Behavior: Next Steps for Foundations and
Researchers. In National Science Foundation’s Political Science Program. Arlington,
VA.: National Science Foundation.
Lykken, D., and A. Tellegen. 1996. Happiness is a stochastic phenomenon. Psychological
Science 7 (3):186-9.
Maher, B. 2010. Hiding place for missing heritability uncovered. Nature (26 January 2010).
Maher, B. 2008. Personal genomes: The case of the missing heritability. Nature 456 (7218):1821.
Mahoney, J. 2001. Beyond correlational analysis: Recent innovations in theory and method.
Sociological Forum 16 (3):575-93.
Mameli, M., and P. Bateson. 2006. Innateness and the sciences. Biology & Philosophy 21
(2):155-88.
Manolio, T. A. 2010. Genomewide association studies and assessment of the risk of disease. New
England Journal of Medicine 363 (2):166-76.
Manolio, T. A., F. S. Collins, N. J. Cox, D. B. Goldstein, L. A. et al. 2009. Finding the missing
heritability of complex diseases. Nature 461 (7265):747-53.
Martin, N., D. Boomsma, and G. Machin. 1997. A twin-pronged attack on complex traits. Nature
Genetics 17 (4):387-92.
43
Martin, N. G., L. J. Eaves, A. C. Heath, R. Jardine, L. M. Feingold, and H. J. Eysenck. 1986.
Transmission of Social-Attitudes. Proceedings of the National Academy of Sciences of
the United States of America 83 (12):4364-8.
Masters, Roger D. 1990. Evolutionary Biology and Political Theory. American Political Science
Review 84 (1):195-210.
Mcgue, M., and D. T. Lykken. 1992. Genetic Influence on Risk of Divorce. Psychological
Science 3 (6):368-73.
Meaney, Michael J. 2001. Nature, Nurture, and the Disunity of Knowledge. Annals of the New
York Academy of Sciences 935:50-61.
Medland, S. E., and P. K. Hatemi. 2009. Political Science, Biometric Theory, and Twin Studies:
A Methodological Introduction. Political Analysis 17:191–214.
Molenaar, P. C. M., and C. G. Campbell. 2009. The New Person-Specific Paradigm in
Psychology. Current Directions in Psychological Science 18 (2):112-7.
Mondak, J. J., M. V. Hibbing, D. Canache, M. A. Seligson, and M. R. Anderson. 2010.
Personality and Civic Engagement: An Integrative Framework for the Study of Trait
Effects on Political Behavior. American Political Science Review 104 (1):85-110.
Munafo, M. R., C. Durrant, G. Lewis, and J. Flint. 2009. Gene x Environment Interactions at the
Serotonin Transporter Locus. Biological Psychiatry 65 (3):211-9.
Munafo, M. R., and J. Flint. 2009. Replication and heterogeneity in gene x environment
interaction studies. International Journal of Neuropsychopharmacology 12 (6):727-9.
Oxley, D. R., K. B. Smith, J. R. Alford, M. V. Hibbing, J. L. Miller, M. Scalora, P. K. Hatemi,
and J. R. Hibbing. 2008. Political Attitudes Vary with Physiological Traits. Science 321
(5896):1667-70.
44
Pinker, Steven. 2003. The Blank Slate: The Modern Denial of Human Nature. London: Penguin.
Plomin, R. 1989. Environment and Genes - Determinants of Behavior. American Psychologist 44
(2):105-11.
Plomin, R., M. J. Owen, and P. McGuffin. 1994. The Genetic-Basis of Complex Human
Behaviors. Science 264 (5166):1733-9.
Price, T. D., A. Qvarnstrom, and D. E. Irwin. 2003. The role of phenotypic plasticity in driving
genetic evolution. Proceedings of the Royal Society of London Series B-Biological
Sciences 270 (1523):1433-40.
Prichard, Z., A. Mackinnon, A. F. Jorm, and S. Easteal. 2008. No evidence for interaction
between MAOA and childhood adversity for antisocial behavior. American Journal of
Medical Genetics Part B-Neuropsychiatric Genetics 147B (2):228-32.
Risch, N., R. Herrell, T. Lehner, K. Y. Liang, L. Eaves, J. Hoh, A. Griem, M. Kovacs, J. Ott, and
K. R. Merikangas. 2009. Interaction Between the Serotonin Transporter Gene (5HTTLPR), Stressful Life Events, and Risk of Depression A Meta-analysis. Jama-Journal
of the American Medical Association 301 (23):2462-71.
Rose, Steven P. R. 2006. Commentary: Heritability estimates - long past their sell-by date.
International Journal of Epidemiology 35 (3):525-7.
Rustow, Dankwark. A. 1970. Transitions to Democracy - toward a Dynamic Model.
Comparative Politics 2 (3):337-63.
Sarkar, Sahotra. 1999. From the Reaktionsnorm to the Adaptive Norm: The Norm of Reaction,
1909-1960. Biology & Philosophy 14 (2):235-52.
Schonemann, P. H. 1997. On models and muddles of heritability. Genetica 99 (2-3):97-108.
45
Settle, J. E., C. T. Dawes, N. A. Christakis, and J. H. Fowler. 2010. Friendships Moderate an
Association between a Dopamine Gene Variant and Political Ideology. Journal of
Politics 72 (4):1189-98.
Shultziner, Doron. 2010. Struggling for Recognition: The Psychological Impetus for Democratic
Progress. New York: Continuum Press.
Shultziner, Doron, Thomas Stevens, Martin Stevens, Brian A. Stewart, Rebbeca J. Hannagan,
and Giulia Saltini-Semerari. 2010. The causes and scope of political egalitarianism
during the Last Glacial: a multi-disciplinary perspective. Biology & Philosophy 25
(3):319-46.
Smith, K. B., J. R. Alford, P. K. Hatemi, L. Eaves, C. L. Funk, and J. R. Hibbing. 2012. Biology,
Ideology, and Epistemology: How Do We Know Political Attitudes Are Inherited and
Why Should We Care? American Journal of Political Science 56 (1):17-33.
Somit, Albert. 1976. Biology and Politics: Recent Explorations. The Hague: Mouton.
Somit, Albert, and Steven Peterson. 1998. Biopolitics after Three Decades - A Balanced Sheet.
British Journal of Political Science 28 (3):559-71.
Suhay, Elizabeth, Nathan Kalmoe, and Christa McDermott. 2007. Why Twin Studies Are
Problematic for the Study of Political Ideology: Rethinking ‘Are Political Orientations
Genetically Transmitted?’ Presented at the International Society of Political Psychology
2007 annual meeting, Portland, OR.
Sultan, S. E. 2003. Commentary: The promise of ecological developmental biology. Journal of
Experimental Zoology 296 (1):1-7.
46
Tajfel, Henri, and John C. Turner. 1979. An Integrative Theory of Intergroup Conflict. In The
Social Psychology of Intergroup Relations, ed. W. G. Austin and S. Worchel. Monterey,
CA: Brooks-Cole.
Todd, J. A. 2006. Statistical false positive or true disease pathway? Nature Genetics 38 (7):7313.
Tooby, John, and Leda Cosmides. 1992. The Psychological Foundations of Culture. In The
Adapted Mind: Evolutionary Psychology and the Generation of Culture, ed. J. H.
Barkow, L. Cosmides and J. Tooby. New York: Oxford University Press.
Turkheimer, Eric. 2004. Spinach and Ice Cream: Why Social Science Is So Difficult. In
Behavioral Gentics Principles: Perspectives in Development, Personality, and
Psychopathology, ed. L. DiLalla. Washington DC: American Psychological Association.
Turkheimer, Eric. 2000. Three Laws of Behavior Genetics and What They Mean. Current
Directions in Psychological Science 9 (5):160-4.
Turkheimer, E., A. Haley, M. Waldron, B. D'Onofrio, and I. I. Gottesman. 2003. Socioeconomic
status modifies heritability of IQ in young children. Psychological Science 14 (6):623-8.
Turner, John C. 1982. Toward a cognitive definition of the social group. In Social Identity and
Intergroup relations, ed. H. Tajfel. Cambridge: Cambridge University Press.
Verhulst, Brad, Lindon J. Eaves, and Peter K. Hatemi. 2012. Correlation not Causation: The
Relationship between Personality Traits and Political Ideologies. American Journal of
Political Science 56 (1):34-51.
Visscher, P. M., W. G. Hill, and N. R. Wray. 2008. Heritability in the genomics era - concepts
and misconceptions. Nature Reviews Genetics 9 (4):255-66.
47
Visscher, P. M., S. E. Medland, M. A. R. Ferreira, K. I. Morley, G. Zhu, B. K. Cornes, G. W.
Montgomery, and N. G. Martin. 2006. Assumption-free estimation of heritability from
genome-wide identity-by-descent sharing between full siblings. Plos Genetics 2 (3):31625.
Waller, N. G., B. A. Kojetin, T. J. Bouchard, D. T. Lykken, and A. Tellegen. 1990. Genetic and
Environmental-Influences on Religious Interests, Attitudes, and Values - a Study of
Twins Reared Apart and Together. Psychological Science 1 (2):138-42.
Weder, N., B. Z. Yang, H. Douglas-Palumberi, J. Massey, J. H. Krystal, J. Gelernter, and J.
Kaufman. 2009. MAOA Genotype, Maltreatment, and Aggressive Behavior: The
Changing Impact of Genotype at Varying Levels of Trauma. Biological Psychiatry 65
(5):417-24.
48