Fixed or Fluid? Sexual Identity Fluidity in a Large National Panel Study of New Zealand Adults

ABSTRACT

Sexual orientation has been defined as an enduring aspect of the self, but emerging evidence reveals that people’s sexual attractions, behaviors, and identities can shift over time. To examine this possibility, we present a large longitudinal analysis of sexual orientation identity fluidity among New Zealand adults (N_total = 45,856; age = 18–99; lesbian, gay, bisexual, and other sexual minority (LGB+) ns = 746–3,387). Over seven years, 5.7% of participants changed sexual identities at least once. Change was bi-directional (i.e. toward and away from LGB+ identities) and most common in people who initially reported a plurisexual identity. Although women reported higher rates of plurisexuality than men, they were not more fluid in their identities, contradicting the notion of male fixedness and female plasticity in sexuality. Moreover, openness to experience was associated with increased odds of changing from a heterosexual to a plurisexual identity, while political liberalism and lower conscientiousness were associated with increased odds of changing from a heterosexual to a plurisexual identity and more identity changes over time. Overall, our study shows that sexual identity can be fluid into adulthood and has implications for how we understand contemporary human sexuality.

Sexual orientation is popularly defined as an enduring aspect of one’s self based on a pattern of attraction to men, women, or people regardless of gender (American Psychological Association, 2019). Indeed, the “master narrative” of sexual orientation posits static categories of sexual orientation that develop early and remain stable over time (Diamond, 2006). However, a growing body of literature reveals that sexual behavior, attractions, and identity do change over time (Diamond, 2016; Katz-Wise & Todd, 2022). Rather than categorizing sexual orientation as static categories, this alternative model of sexual fluidity argues that these dimensions can change in response to situational, interpersonal, and societal factors (Diamond, 2008b; Diamond et al., 2017). Since Diamond’s seminal work, sexual fluidity research has exponentially increased, with research across multiple populations and dimensions of fluidity (Katz-Wise & Todd, 2022).

Indeed, as sexual fluidity research has grown, so has the operationalization of sexual fluidity (for discussion, see Diamond et al., 2020). The definition of sexual fluidity has changed over time and embodies discrete meanings for different individuals. For example, while some forms of sexual fluidity resemble bisexuality (Ross et al., 2012), others relate to more situationally variable forms of sexuality, and these discrete forms of sexual fluidity are largely uncorrelated (Diamond et al., 2020). The distinctions between different forms of sexual fluidity have significant implications for understanding how different people’s identities may change and develop over time and how one’s identity and social context interact. Accordingly, research examining how and for whom sexual fluidity manifests across time is necessary for understanding contemporary human sexuality.

More broadly, examining sexual fluidity allows those who identify outside of more restrictive “categories” of sexuality to be counted in sexuality research. Indeed, the construction of sexual identity as static, discrete categories has led to the erasure of many lesbian, gay, bisexual, and other non-heterosexual (hereafter LGB+) people (among others) in national studies and censuses (see Ruberg & Ruelos, 2020, for discussion). Accordingly, there are calls for nationwide research to consider constructions of sexuality beyond those that “make sense” to the majority (Guyan, 2021; Lyons et al., 2021; Ruberg & Ruelos, 2020). Doing so would address the misrepresentation of LGB+ populations in research, which has ramifications for the representativeness of data as well as LGB+ resource allocation and policy (Browne, 2010; Velte, 2020). Research examining the extent to which sexual identity is fluid (versus enduring) over time is one way to address these concerns.

However, just how fluid (or enduring) sexual identity is in adulthood remains in question. One reason for this is pragmatic; sexual fluidity is best investigated longitudinally to allow for observation of stability or fluidity over weeks, months, or, ideally, when it comes to identity, years (Diamond, 2008b). Several studies do just this (see Table S1), finding meaningful patterns of fluidity for a minority of people. That said, most past studies are limited in the number of LGB+ people sampled or the number of waves (thus possible changes) assessed. Conversely, those that do use large samples typically focus on adolescents and young adults (Ott et al., 2011). As a result, our understanding of sexual fluidity is (primarily) limited to restricted samples over relatively short periods of time.

Another reason is sociopolitical; until recently, there were very few places where LGB+ people had adequate legal and social protections to report their sexual orientation openly and freely over time. Indeed, same-gender sexual behavior remains illegal across much of the world, and in nations that have decriminalized LGB+ orientations, other barriers to true social equality remain (see Mendos, 2019). This is an important consideration for the current research, as the greater legal and social oppression LGB+ people face, the more likely they are to suppress their identities (Charlton et al., 2016). Additionally, even if people do “come out,” societal stigma can affect how these identities are expressed and the extent to which people engage in same-gender sexual behaviors (Pachankis et al., 2017). Thus, an environment whereby people can freely express their sexual identity is needed to appropriately examine sexual fluidity, as more oppressive contexts may underestimate the prevalence of (a) stable LGB+ identities and (b) fluidity within and across LGB+ and heterosexual identities.

Here, we overcome these issues by assessing women’s and men’s self-defined sexual orientation/identity (abbreviated hereafter as “sexual identity”) in a large national longitudinal study (i.e., over eight annual waves). To maximize sample size, we pooled participants across waves to examined the one-year rate of change within our entire sample (see our Modeling strategy below), though we also provide descriptive data on the prevalence of sexual fluidity (or lack thereof) in our sample across seven years. Critically, we examine the relative linearity and fluidity of sexual identity across heterosexual, plurisexual (e.g., those who identify as bisexual, pansexual, bicurious, and so on), and gay/lesbian participants, as well as potential demographic differences between those more (versus less) fluid over time. Overall, our study aims to provide a more nuanced understanding of sexual fluidity by identifying (a) the prevalence and direction of sexual fluidity across different sexual identities and (b) who within these different groups is most likely to be sexually fluid (see Table 1 for an overview of our research questions and hypotheses). By doing so, we contribute to a growing body of literature that acknowledges the potential complexities and fluidity in sexual identity across time.

Table 1. Summary of research questions and hypotheses.

Research Question (RQ)	Hypotheses
(1) Are people directionally sexually fluid?	The extant literature suggests that those who are sexually fluid often experience multiple changes in sexual identity and behavior over the lifespan, both to and away from LGB+ identities (e.g., Campbell et al., 2021; Kaestle, 2019; Savin-Williams et al., 2012). Thus, we expected linear, bidirectional, and “reverting” patterns of sexual fluidity in our study, with bidirectional change as the most “common” pattern over time.
(2) Is plurisexuality stable, transitional, or fluid?	Research suggests that people who adopt plurisexual identities may be more open to lifelong change in identity and, thus, fluidity (Diamond, 2008a; Ross et al., 2012). Accordingly, we expected that those with plurisexual identities may be more likely to be sexually fluid, although we also expect many to hold stable plurisexual identities over time (see Klein, 2014).
(3) Does fluidity change with age?	We anticipate changes in identity to be more common among young adults (compared to older adults), particularly as younger people will have experienced their initial sexual identity development during a relatively stable period of societal acceptance of LGB+ identities.
(4) Are women more sexually fluid than men?	We arbitrate between whether women are more likely to be sexually fluid than men or whether women simply have a higher base rate of plurisexuality than men.
(5) What are the ideological, health, and personality correlates of sexual fluidity?	We assessed several demographic factors to answer this question: 5a: Given that political conservative ideology correlates with sexual prejudice (Herek, 2009), we expect political conservatism and religiosity to be associated with lower levels of sexual fluidity among heterosexuals but that religiosity may increase the number of identity changes among LGB+ people. 5b: Given that previous research identifies a relationship between LGB+ identities, minority stress, and poor physical and mental health (Hoy-Ellis, 2023), we expect similar associations between mental/physical health and sexual fluidity. 5c: Prior work suggests that some LGB+ people are higher in openness to experience and lower in conscientiousness (e.g., Allen & Robson, 2020). As such, we expect similar associations for sexual fluidity.

Moreover, we use data from a place and time where people should feel relatively free to report changes in sexual identity. Indeed, New Zealand is argued to be one of the world’s top 10 “gay-friendly” countries (Spartacus, 2020), and data collection began after legal marriage equality. Notably, the timing of our data collection (from 2013–2021) falls in a time with relatively stable LGB+ policy and protections, with the only significant changes involving the expunging of convictions for historical homosexual offenses in 2018 (Ministry of Justice, 2018) and changes to blood donor restrictions for men who have sex with men in 2020 (Saxton, 2020). This relative stability gives us reasonable confidence to examine true sexual fluidity rather than increases in individuals “coming out” and identifying as LGB+ due to legal protections or policy changes. Thus, the relative legal equality of LGB+ people in New Zealand, while in no way indicating the absence of sexual prejudice (e.g., Daalder, 2022), allows us to get closer to the ideal conditions for sexual fluidity research.

Before outlining the relevant literature and our hypotheses below, we note that the idea that sexual orientation can change away from LGB+ identities has been controversial. So-called “conversion therapy” (often coercive efforts aimed at making LGB+ people “become” heterosexual) is not effective and instead can lead to psychological trauma and suicide (e.g., Alempijevic et al., 2020; Jones et al., 2022; Przeworski et al., 2021). Thus, it is essential to distinguish sexual identity mobility (or fluidity) from any externally directed “conversion” (Diamond, 2008b). Rather than a “conversion,” sexual fluidity refers to natural changes that happen in attractions and behaviors, and subsequently identity, in response to different situations, experiences, relationships, and social contexts (Diamond et al., 2017). The present study focused on one aspect of sexual identity – changes in how participants label and identify themselves. By doing so, we contribute to a growing body of literature on sexual identity fluidity that serves to understand how adult sexual orientation develops and changes over time.

Likewise, it is crucial to recognize that sexual fluidity is distinct from gender fluidity, albeit both exist outside of normative, binary identities (Monro, 2019; Sumerau et al., 2020). Gender fluidity can be understood in several ways; gender fluidity may entail changes in a person’s gender expression or gender identity (or both) over time but can also signal how individuals’ identities have multiple dimensions simultaneously. For example, some people identify as gender fluid and demonstrate fluctuating congruence with feminine and masculine identities (for further discussion, see Bornstein, 1995). Others, however, consider their gender expression or identity as a combination of binary gender categories, neither category or somewhere in between (Callander et al., 2021; Suen et al., 2020). While the complexities of gender fluidity mirror that of sexual fluidity, sexuality and gender are not invariably related; one can be sexually but not gender fluid (or vice versa), and these identities are worthy of both individual and intersectional attention. In the present study, we focused on sexual fluidity due to the very few transgender and gender-diverse people in our sample, though we encourage future researchers to consider the intersections between gender and sexual fluidity over time (see Ruberg & Ruelos, 2020).

Theories of Sexual Fluidity

A popular theory of sexual fluidity proposes a linear pathway: people begin heterosexual and then either remain so (i.e., are stable) or travel (i.e., are fluid) unidirectionally to either an LGB+ identity or through a bisexual identity to a gay/lesbian identity (Cass, 1979; Eliason, 1996). Linear theorists argue that, in a world where heterosexuality is portrayed as the only acceptable outlet for sexual expression (see Cass, 1979), sexual identity formation must start with some form of incongruence with heterosexuality and “end” with congruent identity formation (e.g., Chapman & Brannock, 1987; Moses, 1978; Plummer, 1975). While popular, linear theories of sexual development are critiqued for implying categories of sexuality whereby the “coming out” process inevitably leads to a binary, static identity (Klein et al., 2015). Likewise, the impact of the sociopolitical context is largely overlooked despite considerable evidence suggesting that one’s context plays a significant role in sexual identity development (Diamond, 2006; Kitzinger, 1995; Kitzinger & Wilkinson, 1995). Thus, linear theories discount the dynamic, context-dependent variability in LGB+ identity development (Hall et al., 2021; Kenneady & Oswalt, 2014; Langdridge, 2014).

An alternate idea posits that much fluidity reflects “going through a phase” (Platt & Lenzen, 2013) characterized by (typically young) people exploring same-gender attractions and labels before reverting to heterosexuality. While some people do indeed explore same-gender attractions before “reverting” to heterosexuality, heterosexuality is a continuum (Savin-Williams & Vrangalova, 2013), and fluidity itself can be an identity (Savin-Williams, 2022, 2023). Moreover, many individuals who do demonstrate “phases” of non-heterosexual identity may do so because the sociopolitical environment allows for same-gender exploration or because particular contexts are safer to be “out” as LGB+ than others (e.g., Legate et al., 2012). As such, the assumption of a “base” heterosexual orientation – and fluidity as a “phase” – is contested (e.g., Diamond, 2003; Katz, 2020), and “phases” of LGB+ identity may reflect those with stable LGB+ identities reverting to heterosexual labels for safety.

Another proposed pattern is one of true fluidity, with a minority of people showing bidirectional and often multiple changes in sexual attraction, behavior, and identity over the lifespan (Diamond, 2008b), typically at the confluence of changing individual, interpersonal, or societal contexts that allow for sexual exploration (Katz-Wise & Hyde, 2017). Indeed, the extant literature supports this possibility (Charlton et al., 2016; Dickson et al., 2013; Kaestle, 2019; Mock & Eibach, 2012; Savin-Williams & Ream, 2007; Savin-Williams et al., 2012; Scheitle & Wolf, 2018; Spittlehouse et al., 2020). Research suggests fluidity both to and away from LGB+ identities over time (e.g., Dickson et al., 2003; Kaestle, 2019), sometimes more than once (Scheitle & Wolf, 2018), contradicting theories that suggest a linear trajectory or “phase.” Although dependent on both the measure and sample composition, past studies have found that 1.4–6.2% of men and 2–19.3% of women have changed identities over time, with the highest rates of change representing non-linear trajectories (e.g., 3.5–13.5%; see Table S1). Thus, we expect bidirectional change alongside linear and “reverting” patterns in the present study (see RQ1 in Table 1). To test this thesis, we quantified the percentage of change that fits within each possibility (i.e., linear, “reverting,” and “true” fluidity) to directly examine which of these extant theories “fits” our data best.

Stability and Fluidity in Plurisexuality

The debate alluded to above also has real implications for how we understand plurisexual identities, whose legitimacy and nature have long been debated (e.g., Rieger et al., 2005). Notably, while plurisexual identities (e.g., bisexuality) may share traits of openness and “flexibility” with sexual fluidity (Ross et al., 2012; Weinberg et al., 1995), bisexual orientations are often enduring (and thus, relatively stable) attractions to more than one gender, much like heterosexual, lesbian, and gay identities. Therefore, research understands bisexuality as a “third” sexual orientation of mixed (rather than changing) attraction (Diamond, 2021; Kitzinger, 1995; Klein, 2014; MacDonald, 1981) and bisexual people may only demonstrate specific forms of sexual fluidity, if at all (Diamond et al., 2020).

An alternative explanation sees plurisexual identities as a fleeting “stop-point” before naturally transitioning into a stable monosexual attraction (MacDonald, 1981). A third perspective considers bisexual and plurisexual identities as transitory or, at times, an expression of a denial of same-gender attraction (Gooß, 2008). While similar to transitional perspectives, transitory perspectives argue that people may explore plurisexuality before reverting to their “true” sexual orientation. These perspectives acknowledge the potential fluidity of sexual identity but are critiqued for positioning plurisexual identities as purely “experimental” and, thus, not a true sexual identity (Israel, 2018).

A final perspective (somewhat consistent with the first) sees plurisexual identities as those adopted by people who are naturally more fluid and open to lifelong change in sexual expression (Blumstein & Schwartz, 1977; Ross et al., 2012; Weinberg et al., 1995). Indeed, most longitudinal work suggests that those with plurisexual identities show the most fluidity (see Diamond, 2008a, 2016), although these samples are small and mostly comprised of women. Accordingly, we expected those with plurisexual identities are likely to demonstrate the most fluidity in our sample (RQ2; see Table 1). That said, plurisexuality is a vastly understudied orientation, and thus, we examined and arbitrated discrete perspectives of plurisexuality in the present study.

Demographic Differences in Sexual Fluidity

Beyond establishing the rates and direction of self-described sexual identity change, we also examined the associations between the direction and number of changes that people report and their demographics, politics, health, and personality in the present study. For example, we examine age-related differences in sexual fluidity. As indicated above, many assume that sexual fluidity is developmental, whereby sexual exploration in young adulthood terminates in a stable identity in middle or older adulthood. The present study was ideally suited to test this assumption. Although some smaller studies find that age is unrelated to adult sexual identity change (e.g., Mock & Eibach, 2012), younger people are more likely to report plurisexual identities (Greaves et al., 2017), and plurisexual people show the most sexual fluidity (Diamond, 2016). Additionally, younger people experienced their initial sexual identity development (and thereafter) during increased legal and societal acceptance of LGB+ identities. As such, younger adults may be free to explore their sexual identities more so than their older counterparts, and thus, we proposed that change should be more common among young adults (RQ3; Table 1).

We also examined potential gender differences in sexual fluidity (see RQ4; Table 1). Previous research suggests that women routinely report more plurisexual behavior and attractions relative to men: a pattern that has led many to frame men’s sexuality as fixed and biological and women’s as erotically plastic, changing in response to environmental pressures (Baumeister, 2000). In line with this perspective, some research suggests that more women than men change their sexual identity over 6–10 years (Charlton et al., 2016; Dickson et al., 2013; Mock & Eibach, 2012). Our work will test this proposition insofar as it concerns self-reported sexual orientation identity, modeling gender as a predictor of the volume and direction of any change. From an erotic plasticity perspective (Baumeister, 2000), we might expect women to display more change than men. Recent work, however, highlights contextual shifts in men’s sexuality (Savin-Williams, 2017; Ward, 2015). Thus, it is unclear whether women are more fluid than men year-to-year or simply have a higher base rate of plurisexuality (Diamond, 2016; Diamond et al., 2017). We examined both possibilities in the present study.

Additionally, we explored whether political ideology and religiosity might be associated with sexual fluidity (see RQ5a; Table 1), given political attitudes’ effects on attitudes toward LGB+ people (Herek, 2009). Politically, people often identify as right-wing/conservative (valuing tradition, safety, and security), left-wing/liberal (valuing change, equality, and fairness), or somewhere in between (Jost, 2006). Given their preference for stability, tradition, and existing power structures, more politically conservative people typically display more prejudice toward LGB+ people (Herek, 2000). Consequently, political conservatism may be associated with lower levels of change (or reporting of change) from a heterosexual to an LGB+ identity. Likewise, recent research suggests that religiosity is associated with a lower likelihood of identifying as LGB+ (PRRI, 2023), and one study showed that those who initially identified as LGB+ and more religious had more identity changes (Scheitle & Wolf, 2018). Thus, religiosity may increase the number of identity changes among LGB+ people or may be associated with less sexual fluidity among heterosexuals. While the associations religiosity and conservatism have with sexual prejudice are well-established (see Herek, 2009), whether these ideologies have similar effects on how individuals identify is less known. As such, the associations political orientation and religiosity have with sexual fluidity (or lack thereof) are worthy of exploration.

Finally, plurisexual people are often stereotyped as emotionally unstable (see Hectors, 2023; Mohr et al., 2001; Zivony & Lobel, 2014), and meta-analyses reveal that LGB+ people report slightly higher levels of neuroticism than heterosexual people (Allen & Robson, 2020). Here, we test whether subjective health, psychological distress, and emotionality (i.e., neuroticism) are related to sexual fluidity (i.e., identity changes) to discern whether similar associations emerge for sexually fluid individuals (RQ5b; Table 1). We note, however, that our intention is not to continue the pathologization of non-heterosexual identities (see Hectors, 2023, for discussion). The relationship between sexual identity and (mental) health is complex and requires more research attention that considers the barriers to physical and mental healthcare (Alencar Albuquerque et al., 2016; Rees et al., 2020) and sexual stigma (Herek, 2009; Hoy-Ellis, 2023) as factors (among others) that shape the well-being of LGB+ people. In the present study, we focus solely on the relationships between mental health and sexual fluidity, although we encourage future work to interrogate the factors that shape these associations.

More broadly, we explored how openness to experience, conscientiousness, extraversion, and agreeableness correlate with sexual fluidity (RQ5c; Table 1). Social and environmental factors play an important role in forming personality and sexual identity (see Allen & Robson, 2020, for further discussion). Notably, research suggests aggregate genetic overlaps between personality traits and traits underlying same-gender sexual behavior (see Ganna et al., 2019). Thus, sexual fluidity may be associated with personality differences (see Diamond, 2021, for discussion). Indeed, personality traits have been studied as correlates of sexual orientation in the past (Greaves et al., 2017; Hertzog, 1996; Lippa, 2005, 2010; Perrella et al., 2012; Worthen, 2019; Zheng et al., 2011), albeit largely cross-sectionally. Openness, in particular, is proposed to be central to sexual fluidity (Diamond, 2016), and LGB+ people reliably report greater openness to experience than heterosexuals (Allen & Robson, 2020). Therefore, we expected openness to be positively associated with fluidity. Similarly, bisexual people also reliably report lower levels of conscientiousness relative to both heterosexual and gay and lesbian people (Allen & Robson, 2020), and thus sexual fluidity may also be associated with lower levels of conscientiousness.

It is, however, important to note that our examination of personality correlates is not intended to suggest that sexual fluidity is inherent or exclusively “caused” by differences in personality (see Diamond, 2021, for a discussion). As mentioned, research suggests several genetic traits may influence same-gender sexual behavior (Ganna et al., 2019). However, it is the interaction between genetics and the environment that influences behavior; in other words, whether someone engages in same-gender sexual behavior – or identifies as LGB+— is constrained by the acceptance and legality of same-gender sexuality. Thus, our aim was not to examine whether personality causes sexual fluidity (or lack thereof) but to provide greater insight into the relationships between sexual identity and personality and who may be most likely to be sexually fluid.

Method

Sampling Procedure and Participants

Data came from eight waves of the national probability New Zealand Attitudes and Values Study (NZAVS). We refer to these as Time 1 through Time 8 throughout, although they correspond to Time 5 (data collected in 2013/14), Time 6 (2014/15), Time 7 (2015/16), Time 8 (2016/17), Time 9 (2017/18), Time 10 (2018/2019), Time 11 (2019/2020) and Time 12 (2020/2021) of the NZAVS (sexual orientation was only included from Time 5 of the NZAVS). The NZAVS, which began in 2009, is an ongoing 20-year national longitudinal panel study of the social attitudes, personality, and health of New Zealanders sampled from the New Zealand electoral roll. Full sampling details for the study at each wave are reported in Sibley (2023) and on the NZAVS OSF page (https://osf.io/75snb/). All procedures performed in studies involving human participants were in accordance with the ethical standards of the University of Auckland Human Participants Ethics Committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Among the Time 1 sample, 85.5% of participants reported ethnicities as European, 12.7% as Māori (Indigenous peoples of New Zealand), 3.4% as Pacific, 4.5% as Asian, and 5.1% as some other ethnicity (participants could report more than one ethnicity). Regarding gender, 62.8% of the sample were women, while 39.4% were religious, 73.7% were parents, and 72.4% were in a relationship. The mean age was 48.14 (SD = 14.07).

The main set of analyses focused on two specific samples of respondents, described in the Modeling Strategy section below. Among the respondents who responded to every wave (Time 1 – Time 8) with a codable sexual orientation, 95.6% reported their ethnicity as European, 7.6% as Māori, 1.4% as Pacific, 2.2% as Asian, and 1.6% as some other ethnicity. Regarding gender, 62.5% were women, while 36.5% were religious, 72.9% were parents, and 76% were in a relationship. The mean age was 50.16 (SD = 13.11). Among respondents in the pooled sample (independent respondents who had responded to two adjacent waves between Time 1–8; see Modelling Strategy), 91.5% reported their ethnicity as European, 9.2% Māori, 2.0% Pacific, 4.1% Asian, and 3.0% some other ethnicity. Regarding gender, 63.6% were women, while 36.7% were religious, 71.2% were parents, and 74.9% were in a relationship. The mean age was 48.94 (SD = 13.74). Overall, these samples tend to over-represent women and European ethnicities in particular, relative to 2018 New Zealand census estimates (51.4% and 70.2%, respectively; see Sibley, 2023).

Questionnaire Measures

Sexual orientation identity (sexual identity) was assessed with the question: “How would you describe your sexual orientation?” The question was open-ended, and responses were coded into identities according to a structured classification scheme (Greaves et al., 2017). A full copy of this coding scheme and the sample sizes for missing or non-codable responses are presented in the Supplementary Online Materials (see Table S2). We classified participants into three categories for inclusion in our analyses: Heterosexual, lesbian/gay, and plurisexual (which refers to answers with codes under Bisexual, Bicurious, Pansexual, and LGB+ not otherwise specified). Table 2 displays the sample sizes at each wave for each of these three categories.

Table 2. Sample sexual orientation identity characteristics at each time point.

	Time 1 (2013/14)	Time 2 (2014/15)	Time 3 (2015/16)	Time 4 (2016/17)	Time 5 (2017/18)	Time 6 (2018/19)	Time 7 (2019/20)	Time 8 (2020/21)
Heterosexual	13,618 (94.7%)	11,868 (94.1%)	11,060 (93.5%)	17,491 (93.1%)	14,874 (93.4%)	42,464 (93.3%)	38,357 (91.9%)	34,712 (91.5%)
Lesbian/Gay	323 (2.2%)	294 (2.3%)	285 (2.4%)	474 (2.5%)	410 (2.6%)	1,089 (2.4%)	1,149 (2.8%)	1,116 (2.9%)
Plurisexual	391 (2.7%)	406 (3.2%)	425 (3.6%)	736 (3.9%)	586 (3.7%)	1,752 (3.9%)	1,998 (4.8%)	1,837 (4.8%)
Asexual	39 (0.3%)	46 (0.4%)	57 (0.5%)	81 (0.4%)	57 (0.4%)	185 (0.4%)	240 (0.6%)	257 (0.7%)
Total N	14,371	12,614	11,827	18,782	15,927	45,490	41,744	37,922

We also had a category for those who identified as asexual (i.e., those who experience no sexual attractions or low levels of attraction; see Bogaert, 2015; Greaves et al., 2017). Because of the small size of this group (see Table 2), we did not analyze consistency and changes to or from asexual identification. However, we did include asexual participants in the analyses assessing overall fluidity.

We used several variables to predict sexual identity change and fluidity. Age was measured in years based on provided birthdate. The gender measure changed over time (see Supplemental Online Material; SOM): in Time 1, gender was measured with a “female” or “male” binary forced choice question. In Time 2–8, gender was measured with the coded, open-ended question: “What is your gender?.” The full gender coding scheme is presented in Table S3 (SOM). The change to an open-ended measure was intended to allow participants to describe their gender identity using whatever terms they preferred, in line with a growing need for research “counting” transgender and gender-diverse participants (Fraser et al., 2020). That said, very few participants self-identified as transgender or gender-diverse in our sample (see the SOM for more information). Due to sample size considerations and our targeted gender-based hypotheses, we excluded gender-diverse participants (n = 236, 0.3% of the total NZAVS sample) from analyses relating to gender, although we emphasize the importance of future work looking at sexual fluidity in gender-diverse populations.

Personality was measured using the Mini-IPIP (Donnellan et al., 2006), a short form inventory assessing the Big-Five facets on a 1 (very inaccurate) to 7 (very accurate) scale (αs at Time 1 for Extraversion = .75, Agreeableness = .70, Conscientiousness = .66, Emotionality = .70, and Openness = .69).

Subjective health was assessed through the items “I seem to get sick a little easier than other people” and “I expect my health to get worse” (α at Time 1 = .61; Ware & Sherbourne, 1992), which participants rated from 1 (strongly disagree) to 7 (strongly agree). Finally, psychological distress was measured with the Kessler-6 scale, where participants rated how often during the last 30 days they felt “hopeless,” “so depressed that nothing could cheer you up,” “restless or fidgety,” “that everything was an effort,” “worthless” and “nervous” a scale from 1 “none of the time” to 4 “all of the time” (α at Time 1 = .84; Kessler et al., 2010).

Religiosity was assessed by asking participants whether they “identify with a religion and/or spiritual group” (Yes/No), and political conservatism was measured with the single item “Please rate how politically liberal versus conservative you see yourself as being” from 1 (Extremely liberal) to 7 (Extremely conservative).

Modelling Strategy

Because of the very small sample sizes of participants changing their stated sexual identity over time, we pooled participants across waves to examine the one-year rate of change within the entire sample. Inclusion in the pooled sample required participants to complete two consecutive waves of the NZAVS. Shifts (or consistency) in identity were first coded from Time 1 to Time 2. Shifts were then coded from Time 2 to Time 3 for participants not previously coded from Time 1 to Time 2 (either because they did not complete the survey or did not provide a codable sexual identity at one or both waves), and so on. This sample thus examines the one-year rate of change among unique participants, irrespective of whether they entered the study at Time 1 or a later wave. By taking this approach, we obtained an overall sample size of N = 45,856 respondents, which consisted of 42,848 heterosexual respondents (n = 27,248 women; 63.6%), 1,192 lesbian/gay respondents (n = 556 women; 46.6%), and 1,816 plurisexual respondents (n = 1,284 women; 70.7%) at the first measurement occasion.

We conducted two types of models to examine the variables associated with sexual fluidity. First, logistic regressions were conducted with the complete set of correlates of maintaining vs transitioning between heterosexual and plurisexual over a one-year period using the pooled sample. Models assessing transitions to and from lesbian/gay identities, however, only included gender and age due to a much smaller sample of people transitioning to and from this category. Predictor variables were always coded to match each participant’s first wave of data, and we utilized listwise deletion for missing data across predictor variables. To examine general fluidity in sexual identities, we conducted a negative binomial regression of the number of times participants (who had responded to every wave) reported a change in sexual identity across the eight waves of data, using the full set of variables (measured at Time 1). This sample consisted of 4,269 participants, including 4,041 heterosexual (n = 2,531 women; 62.6%), 108 lesbian/gay (n = 49 women; 45.4%), and 7 plurisexual (n = 5 women; 71.4%) participants as at Time 1. Finally, results from these formal models are supplemented with descriptive data to provide a more complete evaluation of sexual fluidity.

Results

Sexual Identity Stability and Change

We first present general rates of change in sexual identity over one – eight years. Table 3 displays Time 1 (2013/14) sexual identities as columns and later sexual identities at Times 2 (2014/15) to 8 (2020/21) as rows. Overall, the heterosexual category was the most fixed over time (97.7% of those who identified as heterosexual in Time 1 also identified as heterosexual seven years later). Lesbian/gay identification was slightly less stable, with 88.5% of those identifying as lesbian/gay also identifying this way seven years later. Plurisexual identities were the least stable, with 63.6% of those identifying as plurisexual in Time 1 also identifying this way seven years later. Note that in each case, however, more participants may have changed identities prior to Time 8 before changing back again at Time 8.

Table 3. Sexual orientation identity by gender with Time 1 (2013/14) as columns and later time points as rows.

					Time 1 (2013/14)
	Heterosexual				Lesbian/Gay			Plurisexual
	Women		Men	Total	Women	Men	Total	Women	Men	Total
	n (%)		n (%)	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)
Time 2 (2014/15)
Heterosexual	6,211 (98.8%)		3,792 (99.2%)	10,007 (99.0%)	0 (0.0%)	2 (1.6%)	2 (0.8%)	45 (21.7%)	19 (23.8%)	65 (22.1%)
Lesbian/Gay	4 (0.1%)		4 (0.1%)	8 (0.0%)	106 (93.8%)	127 (98.4%)	235 (96.3%)	6 (3.0%)	9 (11.3%)	15 (5.1%)
Plurisexual	71 (1.1%)		27 (0.7%)	98 (1.0%)	7 (6.2%)	0 (0.0%)	7 (2.9%)	156 (75.4%)	52 (65.0%)	214 (72.8%)
Time 3 (2015/16)
Heterosexual	5,714 (98.3%)		3,498 (98.8%)	9,216 (98.3%)	1 (0.9%)	0 (0.0%)	1 (0.4%)	40 (22.9%)	16 (22.5%)	56 (22.3%)
Lesbian/Gay	4 (0.1%)		6 (0.2%)	10 (0.1%)	104 (90.4%)	109 (96.5%)	215 (93.5%)	7 (4.0%)	8 (11.3%)	15 (6.0%)
Plurisexual	92 (1.6%)		38 (1.1%)	131 (1.4%)	10 (8.7%)	4 (3.5%)	14 (6.1%)	128 (73.1%)	47 (66.2%)	180 (71.7%)
Time 4 (2016/17)
Heterosexual	5,795 (98.3%)		3,461 (98.7%)	9,260 (98.5%)	4 (3.7%)	1 (0.9%)	5 (2.2%)	38 (22.4%)	24 (31.2%)	63 (24.8%)
Lesbian/Gay	5 (0.1%)		8 (0.2%)	13 (0.1%)	95 (87.3%)	113 (97.4%)	209 (92.1%)	5 (2.9%)	11 (14.3%)	16 (6.3%)
Plurisexual	93 (1.6%)		39 (1.1%)	132 (1.4%)	10 (9.2%)	2 (1.7%)	13 (5.7%)	127 (74.7%)	42 (54.5%)	175 (68.9%)
Time 5 (2017/18)
Heterosexual	4,921 (98.1%)		2,958 (99.0%)	7,883 (98.4%)	2 (2.2%)	2 (1.9%)	4 (2.0%)	37 (23.9%)	20 (28.6%)	57 (24.5%)
Lesbian/Gay	7 (0.1%)		6 (0.2%)	13 (0.2%)	82 (90.1%)	100 (96.2%)	183 (92.9%)	11 (7.1%)	6 (8.6%)	17 (7.3%)
Plurisexual	90 (1.8%)		23 (0.8%)	113 (1.4%)	7 (7.7%)	2 (1.9%)	10 (5.1%)	107 (69.0%)	44 (62.9%)	159 (68.2%)
Time 6 (2018/19)
Heterosexual		5,401 (98.5%)	3,253 (98.7%)	8,658 (98.6%)	2 (2.2%)	1 (0.9%)	3 (1.4%)	38 (23.2%)	17 (26.2%)	55 (23.2%)
Lesbian/Gay		3 (0.5%)	11 (0.3%)	14 (0.2%)	82 (90.1%)	112 (97.4%)	195 (93.8%)	7 (4.3%)	7 (10.8%)	16 (6.8%)
Plurisexual		79 (1.4%)	31 (0.9%)	111 (1.3%)	7 (7.7%)	2 (1.7%)	10 (4.8%)	119 (72.6%)	41 (63.1%)	166 (70.0%)
Time 7 (2019/2020)
Heterosexual		4,660 (98.1%)	2,750 (98.5%)	7,414 (98.2%)	2 (2.3%)	1 (1.0%)	3 (1.6%)	40 (27.4%)	19 (29.7%)	59 (26.9%)
Lesbian/Gay		12 (0.3%)	10 (0.4%)	24 (0.3%)	73 (83.9%)	91 (93.8%)	165 (88.7%)	8 (5.5%)	8 (12.5%)	19 (8.8%)
Plurisexual		79 (1.7%)	33 (1.2%)	112 (1.5%)	12 (13.8%)	5 (5.2%)	18 (9.7%)	98 (67.1%)	37 (57.8%)	141 (64.4%)
Time 8 (2020/2021)
Heterosexual		4,603 (97.6%)	2,760 (98.1%)	7,366 (97.7%)	3 (3.3%)	1 (1.0%)	4 (2.1%)	35 (22.7%)	21 (36.2%)	56 (25.5%)
Lesbian/Gay		16 (0.3%)	12 (0.4%)	28 (0.4%)	74 (82.2%)	94 (94.9%)	169 (88.5%)	12 (7.8%)	8 (13.8%)	22 (10.0%)
Plurisexual		83 (1.8%)	36 (1.3%)	119 (1.6%)	13 (14.4%)	4 (4.0%)	18 (9.4%)	105 (68.2%)	29 (50.0%)	140 (63.6%)

As our formal statistical models examined one-year rates of change, Figure 1 displays the probabilities of transitioning between each sexual identity over a one-year period in the pooled sample. Across participants, there was a probability of .989 for heterosexual participants to report the same identity one year later, compared to .935 for lesbian/gay participants and .803 for plurisexual participants. Thus, stability was highest for heterosexual and lesbian/gay participants over a one-year period and lower for plurisexual participants.

Figure 1. Probabilities of transitioning between (vs. remaining consistent in) each sexual identity over a one-year period. Average full sample probabilities are bold face, with probabilities for men and women samples below.

Table 4 identifies a similar pattern in one and seven-year rates of change among respondents to all eight waves. These proportions reveal high stability in heterosexual identities, with very low rates of change over both one and seven-year periods. Lesbian/gay identities showed less stability over a seven-year period, with 14.6% of lesbian/gay participants at Time 1 changing their identity at least once. Notably, however, a high proportion of plurisexual participants at Time 1 changed their identity over a one-year period (21.3%), and a higher proportion changed their identity at least once over a seven-year period (41.6%).

Table 4. Proportions of the heterosexual, lesbian/gay, and plurisexual samples at Time 1 reporting a different sexual orientation at least once over a one- or seven-year time period, across age groups.

	Heterosexual Time 1		Lesbian/Gay Time 1		Plurisexual Time 1
	One-year n (%)	Seven-year n (%)	One-year n (%)	Seven-year n (%)	One-year n (%)	Seven-year n (%)
Age Group
<26	33 (2.9%)	115 (10.3%)	1 (2.9%)	9 (26.5%)	20 (19.2%)	35 (33.7%)
26–49	53 (0.9%)	232 (3.8%)	3 (1.7%)	21 (12.1%)	37 (19.4%)	78 (40.8%)
>49	34 (0.5%)	139 (2.2%)	5 (4.3%)	17 (14.8%)	27 (27.3%)	51 (51.5%)
Gender
Women	86 (1.0%)	337 (4.0%)	7 (4.5%)	34 (22.1%)	54 (19.4%)	109 (39.1%)
Men	33 (0.6%)	145 (2.8%)	2 (1.2%)	12 (7.2%)	28 (26.9%)	49 (47.1%)
Total n (%)	120 (0.9%)	486 (3.6%)	9 (2.8%)	47 (14.6%)	84 (21.3%)	164 (41.3%)

Based on participants who responded to all waves of data collection. Percentages are for each gender or age group within each sexual identity. Totals may not add to totals across age groups and gender due to missing data on these variables. Some participants may have changed back again, or changed multiple times, over the seven-year window (but are counted once).

Finally, when examining fluidity in terms of the number of times participants who completed all waves changed their stated identity over the seven-year period, 94.1% (n = 4,019) did not change at all. A similar proportion of participants changed once (n = 95; 2.2%) or twice (n = 109; 2.6%). A further 21 participants (0.5%) changed three times, 19 (0.4%) changed four times, and 5 participants (<0.1%) changed five times. One participant changed six times (<0.1%), and no participants changed seven times.

Linearity and Fluidity of Sexual Identity Change

To test whether sexual identity change over time reflects a directional, linear change or “true” fluidity (RQ1), we examined patterns of change in participants’ sexual identity over the seven-year period (among participants who responded to all eight waves and recorded at least one identity change; Figure 2) and directions of change in sexual identity over a one-year period in the pooled sample (Figure 1). As shown in Figure 2, 22.8% of respondents followed the linear developmental model (moving from heterosexual to plurisexual or gay or lesbian, or plurisexual to lesbian/gay, but not in the reverse direction). However, 77.2% (n = 193) of those who changed identity labels did not follow this linear progression. Instead, they moved fluidly between labels, changing either both toward and away, or only away, from same-gender attracted identities.

Figure 2. Patterns of change in sexual identity over an eight-year period among participants who reported at least one identity change. The results contrast patterns of change consistent with the linear developmental model with remaining patterns of change (i.e., those involving, or consisting solely of, change away from same-gender attracted identities).

Figure 1 also indicates some directionality in sexual identity changes over a one-year period. This is most notable for plurisexual participants, who had a higher probability (.159) of shifting to a heterosexual identity over time compared to the probability of shifting to a lesbian/gay identity (.037).

Stability and Fluidity Among Plurisexuals

We then assessed the rates of stability and change among plurisexual participants (RQ2). The one-year (21.3%) and seven-year (41.3%) rates of change among plurisexual participants (see Table 4) suggest a relatively high degree of movement out of this identity. Yet, although plurisexual identities were the least stable over both one-year and seven-year periods, 78.7% (the inverse of total fluidity in Table 4) of those with a plurisexual identity in Time 1 maintained their identity over a one-year and 58.7% over a seven-year period. Moreover, as noted above, those with a plurisexual identity in Time 1 had a higher probability (.159) of shifting to a heterosexual identity over time than a lesbian/gay identity (.037).

Age Differences

We examined evidence for the association between age and fluidity (RQ3) through a negative binomial regression, including age as a predictor of the overall number of identity changes over a seven-year period, the corresponding raw data on change among age groups, and logistic regressions of one-year rates of change predicted by age in the pooled sample.

The negative binomial regression of identity changes indicated that age (measured at Time 1) was negatively associated with the number of times participants reported a change in their sexual identity over seven years. This number of sexual identity changes, broken down by age groups, is reported in Table 5. Younger people were generally the most fluid, with 77.6% of the younger age group reporting no changes in identity over seven years, compared to over 90% in the other age groups. A greater number of sexual identity changes also tended to be reported within the youngest age group, compared to the older age groups (for example, 10.2% of the youngest group reported two identity changes, compared to 2.6% and 1.3% of the middle and older age groups respectively).

Table 5. A breakdown of sexual fluidity (number of identity changes reported over eight-years) within age groups.

	Number of changes
	None	One	Two	Three	Four	Five	Six
Age group	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)
<26	159 (77.6%)	21 (10.2%)	19 (9.3%)	1 (0.5%)	4 (2.0%)	1 (0.5%)	0 (0.0%)
26–49	1,532 (93.0%)	43 (2.6%)	45 (2.7%)	15 (0.9%)	9 (0.5%)	2 (0.1%)	1 (0.1%)
50+	2,328 (96.3%)	31 (1.3%)	45 (1.9%)	5 (0.2%)	6 (0.2%)	2 (0.1%)	0 (0.0%)
Gender
Women	2,491 (93.6%)	64 (2.4%)	75 (2.8%)	14 (0.5%)	14 (0.5%)	4 (0.2%)	0 (0.0%)
Men	1,523 (95.4%)	29 (1.8%)	31 (1.9%)	7 (0.4%)	5 (0.3%)	1 (0.1%)	1 (0.1%)

Counts (and percentages) of participants changing their identity within each age group or gender. No participants changed the maximum number of times (seven) across the eight years.

The logistic regression models allowed us to examine whether age was predictive of specific directions of change in sexual identity over a one-year period relative to retaining a consistent sexual identity over time (i.e., to and from each of the three focal sexual identities). Age was negatively associated with the odds of shifting to a plurisexual identity relative to retaining a heterosexual identity over one year (p < .001) and positively associated with the odds of shifting to a heterosexual identity relative to retaining a plurisexual identity over time (p = .009; see Table 6). It was not, however, associated with shifting to a lesbian/gay identification relative to maintaining a consistent plurisexual identification (p = .706; see Table 7). Instead, age was negatively associated with shifting in the opposite direction, from a lesbian/gay to a plurisexual identity over time (p = .012). Finally, age did not predict the odds of shifting from a heterosexual to lesbian/gay identification (p = .134) or the odds of shifting in the opposite direction (from lesbian/gay to heterosexual identification; p = .841; see Table 7).

Table 6. Logistic regression models of: changing to plurisexual, relative to maintaining heterosexual identification (Model 1) over a one-year period, changing to heterosexual, relative to maintaining a plurisexual identity (Model 2) over a one-year period, and a negative binomial regression model of the number of changes in sexual identity over an eight-year period.

	Model 1 Change to Plurisexual (vs. consistent Heterosexual)				Model 2 Change to Heterosexual (vs. consistent Plurisexual)				Negative binomial regression of number of identity changes
	B	SE	OR [95% CI]	p	B	SE	OR [95% CI]	p	B	95% CI (B)	SE	p
Intercept	3.013	.653		<.001	2.782	.759		<.001	0.212		.897	.813
Age	−0.047***	.004	0.955 [0.947, 0.962]	<.001	0.013**	.005	1.014 [1.003, 1.024]	.009	−0.028***	−0.041, −0.016	.006	<.001
Gender	−0.265*	.122	0.767 [0.604, 0.974]	.030	0.061	.162	1.063 [0.774, 1.461]	.706	−0.054	−0.392, 0.284	.172	.755
Emotionality	0.099	.060	1.104 [0.981, 1.241]	.100	0.096	.073	1.101 [0.954, 1.271]	.188	−0.032	−0.198, 0.135	.085	.710
Subjective health	−0.075	.044	0.928 [0.851, 1.012]	.093	0.096	.060	1.101 [0.978, 1.240]	.110	−0.003	−0.139, 0.133	.069	.970
Kessler-6	0.376***	.091	1.456 [1.218, 1.741]	<.001	−0.266*	.127	0.766 [0.597, 0.983]	.036	0.471**	0.139, 0.802	.169	.005
Extraversion	−0.019	.042	0.981 [0.903, 1.065]	.645	0.098	.055	1.103 [0.991, 1.229]	.074	−0.001	−0.130, 0.129	.066	.993
Agreeableness	0.060	.057	1.062 [0.950, 1.187]	.292	−0.070	.073	0.933 [0.808, 1.076]	.339	−0.035	−0.197, 0.128	.083	.675
Conscientiousness	−0.176***	.045	0.838 [0.767, 0.916]	<.001	0.038	.061	1.039 [0.922, 1.170]	.534	−0.162*	−0.309, −0.014	.075	.032
Openness to Experience	0.316***	.053	1.372 [1.237, 1.521]	<.001	−0.090	.071	0.914 [0.796, 1.050]	.204	0.091	−0.058, 0.241	.076	.232
Religious	−0.393	.125	0.675 [0.529, 0.862]	.002	−0.032	.156	0.968 [0.713, 1.315]	.837	−0.006	−0.328, 0.315	.164	.896
Political conservatism	−0.233	.044	0.792 [0.726, 0.864]	<.001	0.177**	.055	1.194 [1.071, 1.330]	.001	−0.241***	−0.361, −0.121	.061	.001

N_{(Model 1)} = 40,660, N_{(Model 2)} = 1,605, N_{(Identity Changes)} = 4,03;

*p < .05, **p < .01, ***p < .001. B values refer to unstandardized (log-odd or log-count) estimates.

Table 7. Logistic regression models of changes to and from Lesbian/Gay identification (relative to remaining stable) over a one-year period in the pooled sample.

	Model 3 Change to Lesbian/Gay (vs. consistent Heterosexual)				Model 4 Change to Lesbian/Gay (vs. consistent Plurisexual)
	B	SE	OR [95% CI]	p	B	SE	OR [95% CI]	p
Threshold	7.888	.716		<.001	3.201	.383		<.001
Age	0.020	.013	1.020 [0.994, 1.048]	.134	−0.004	.010	0.996 [0.977, 1.016]	.706
Gender	−0.052	.319	0.949 [0.509, 1.772]	.871	0.968***	.268	2.634 [1.558, 4.452]	<.001
	Model 5 Change to Heterosexual (vs. consistent Lesbian/Gay)				Model 6 Change to Plurisexual (vs. consistent Lesbian/Gay)
Intercept	3.644	1.058		.001	1.221	.568		.032
Age	0.004	.020	1.004 [0.966, 1.044]	.841	−0.030**	.012	0.970 [0.947, 0.993]	.012
Gender	−0.523	.428	0.593 [0.256, 1.371]	.222	−2.110***	.473	0.121 [0.048, 0.307]	<.001

N_{(Model 3)} = 42,394, N_{(Model 4)} = 1,439, N_{(Model 5)} = 1,123, N_{(Model 6)} = 1,139.

*p < .05, **p < .01, ***p < .001. B values refer to unstandardized (log-odd or log-count) estimates.

Overall, age was negatively associated with fluidity assessed as the number of changes in sexual identity reported over time but was more inconsistently related with changes between sexual identities over one year, suggesting the effect of age may depend on initial sexual identity. Because heterosexual people are the majority of the sample, when looked at in the aggregate, the association between age and fluidity will be driven largely by initially heterosexual people shifting out of the heterosexual category (even if the base rate of change in this category is low). Therefore, it is useful to examine descriptive data of change within each sexual identity.

The data in Table 4 shows that the patterns for originally heterosexual, lesbian, and gay people largely mirror the results reported above – change was most frequently reported by people in the youngest age group (i.e., those entering the survey at <26 years of age). For plurisexual identities, however, there is no evidence of shifts being reported by the youngest group. Instead, the pattern is relatively “flat” for one year change, with between 19.2–27.3% of participants reporting a change over one year in each age category. For the seven-year rate of change, the highest percentage of change is reported by the oldest originally plurisexual age group (>49 at entry into the survey). The highest rate of change in the one-year timeframe also mirrored this pattern (27.3% of those who entered the survey at 50 years or older reported change over a year).

Gender Differences

To assess gender differences (RQ4), we examined both base rates of sexual identity identification, as well as the association between gender and changing identities over a one-year period and the number of changes over the seven-year period. When pooled across waves, a greater percentage of women (4.7%) reported that they were plurisexual than did men (2.6%). Descriptively, as shown in Figure 1, plurisexual identities appeared slightly more stable over a one-year period among women (P(consistent plurisexual) = .813) than men (P(consistent plurisexual) = .745). The probability of maintaining a consistent heterosexual identity over a one-year period was similar for both women (P(consistent heterosexual) = .988) and men (P(consistent heterosexual) = .992), while the probability of maintaining a consistent lesbian/gay identification appeared lower for women (P(consistent lesbian) = .905) compared to men (P(consistent gay) = .972). These patterns are largely the same when looking at one and seven-year rates of change between men and women in Table 3.

Formally testing potential gender differences in the direction of change, men were less likely than women to shift to a plurisexual identity (vs maintaining a heterosexual identity) over a one-year period (p = .030; see Table 6). However, for both men and women, the probability of shifting in this direction was low (see Figure 1 for the unadjusted probabilities). Men were more likely than women to shift from plurisexual identification to gay identification over a one-year period (B = 0.968, SE = .268, OR[95% CI] = 2.634 [1.558, 4.452], p < .001), while women were more likely to shift in the opposite direction (from lesbian/gay identification to plurisexual identification, relative to continuing to identify as lesbian/gay (B = −2.110, SE = .473, OR[95% CI] = 0.121 [0.048, 0.307], p < .001; see Table 7). As shown in Figure 1, these gender differences were larger than the gender difference in shifting from heterosexual to plurisexual identities.

Finally, there was no evidence of gender differences in the odds of shifting to lesbian/gay identification relative to maintaining a consistent heterosexual identification (B = −.052, SE = .319, OR[95% CI] = 0.949 [0.509, 1.772], p = .871), or vice versa (B = −.523, SE = .428, OR[95% CI] = .593 [0.256, 1.371], p = .222; Table 7). Notably, there was also no evidence of gender differences in sexual identity fluidity when measured as the number of identity changes over a seven-year period (p = .133; see Table 6). This is further highlighted in Table 5, which shows relatively equal percentages of men and women who made each number of identity changes.

Personality, Health, and Political Differences

In addition to age and gender effects, we examined the associations personality, health, and political ideology had with the magnitude of sexual fluidity (number of changes over a seven-year period; negative binomial regression) as well as with movement between heterosexual and plurisexual sexual identities (but not between other sexual identities due to sample size constraints) over a one-year period (logistic regression; RQ5). The results of a negative binomial regression model assessing the full range of psychological correlates of the number of identity changes (regardless of direction) are displayed on the right side of Table 6. The effects of emotionality (p = .710) and subjective health (p = .970) were small and positive yet not significantly associated with the number of times people reported a change in their sexual identity over the seven-year period. However, psychological distress was positively associated with the number of changes (p = .005). Big-Five personality traits were also not uniquely associated with the number of identity changes (ps > .232), except conscientiousness, which was negatively associated with the number of identity changes (p = .032). Finally, political conservatism was negatively associated with fluidity (p = .001).

We also examined the associations between these same variables and identity change between heterosexual and plurisexual identities over a one-year period (Table 6; left-hand columns). Higher scores on psychological distress were associated with increased odds of shifting from a heterosexual to plurisexual identity over time (p < .001). Higher scores on emotionality were associated with increased odds of this transition, but the effect was not significant (p = .100). Subjective health ratings were negatively associated with transitioning from heterosexual to plurisexual identities, but the effect was not significant (p = .093). Conscientiousness was associated with lower odds (p < .001), and openness to experience higher odds (p < .001), of shifting from heterosexual to plurisexual identification over time. Finally, both religiosity and political conservatism were associated with lower odds of shifting from heterosexual to plurisexual identities over time (p = .002 and p < .001, respectively). Looking at transitions in the opposite direction (i.e., from plurisexual to heterosexual identification), only political conservatism (p = .001) and psychological distress (p = .036) were associated with increased odds of shifting in this direction over time, with all other variables not associated with transitions in this direction.

The analyses reported above included all variables simultaneously and thus removed shared variance to look at unique associations between these constructs and indices of change. Given the targeted question about emotionality and psychological distress’s associations with identity fluidity, we conducted additional analyses looking at the association between emotionality (i.e., neuroticism) and psychological distress separately, each predicting the number of identity changes. When not taking other variables into account, higher levels of both emotionality (B = .218, SE = .064 p = .001) and psychological distress (B = .668, SE = .112, p < .001) predicted a greater number of identity changes.

Discussion

The current study presents a large-scale longitudinal examination of the frequency and probabilities of sexual identity fluidity among heterosexual and LGB+ adults in New Zealand. Critically, our study measured sexual identity and fluidity in a context where people are relatively safe and free to explore their sexual identity (Spartacus, 2020), allowing us to examine as close to “true” levels of fluidity as possible in a survey design. Moreover, we examined a broad age range of heterosexual and LGB+ participants to thoroughly explore individual differences that may predict sexual fluidity (or lack thereof). By doing so, we contribute to a growing body of work that examines how, for some people, sexuality may be dynamic, fluid, and situationally dependent. Likewise, we address a growing need for nationally representative studies to consider narratives of sexual identity that represent those who exist beyond societal norms about sexuality (Browne, 2010; Ruberg & Ruelos, 2020).

We first explored the frequency and direction of sexual identity change. Our findings revealed that, over a one-year period, heterosexual men were the least likely to change identities, followed by heterosexual women and gay men (all had >97% probabilities of continuing to identify as the same identity after one year). In contrast, lesbians (9.5% after one year), plurisexual women (18.7%), and plurisexual men (25.5%) were most likely to change identity after one year. However, it is important to note that identity was still relatively stable over the one-year period. Indeed, over the course of seven years, only 5.7% of people reported at least one sexual identity change. These results indicate that for the vast majority of participants, the way one describes their sexual identity is stable over time. However, these results also show that sexual identity is fluid for some, corroborating previous research that reveals fluidity among a minority of people (see Table S1; see also Katz-Wise & Todd, 2022). While on the surface, 5.7% of the sample may seem like a small proportion of the sample, it is important to note that the proportion of the overall sample who identified as LGB+ (at Time 1) was 5.4%. Thus, those demonstrating sexual fluidity reflect a sizable minority in our sample that would be erased or misrepresented in constructions of sexual identity as static, discrete categories.

Are People Directionally Sexually Fluid?

As our sample consisted of a sizable group of participants who already identified as LGB+ at the time of the first assessment and, thus, were already “out,” we could examine whether change occurred across identities. Moreover, we could examine whether change was limited to a linear pathway consistent with people “coming out” and moving from stated heterosexual to LGB+ identities (Cass, 1979; Eliason, 1996) or whether change was more consistent with multidirectional sexual fluidity. Our findings revealed that approximately 23% of those who changed their self-reported sexual identity displayed a linear trajectory consistent with “coming out.” The majority of fluidity was not, however, corroborating previous assertions that sexual identity change, when it occurs, does not predominantly occur in one direction and may reflect true fluidity (Diamond, 2006).

Is Plurisexuality Stable, Transitional, or Fluid?

Our findings also challenge the common perceptions that plurisexual people are “in denial” (Matsick & Rubin, 2018; Rieger et al., 2005) or that plurisexual identities are transitional, bridging the gap between heterosexual and lesbian or gay identities (MacDonald, 1981). Our results contradict these ideas: only 3.6% of people showed an identity transition from a plurisexual to a gay/lesbian identity (either beginning at heterosexual or not). Likewise, 63.6% of participants who entered the study as plurisexual reported the same identity seven years later. Consistent with past work (e.g., Savin-Williams et al., 2012), however, plurisexual participants were the most likely to show fluidity in all directions, potentially supporting the idea that plurisexual identities can be open to change and may shift in response to different relationships or experiences (Blumstein & Schwartz, 1977; Diamond, 2016; Ross et al., 2012; Weinberg et al., 1995). Overall, these findings are broadly consistent with understanding plurisexual identities as relatively non-transitional sexual identities for some, but fluid or open to lifelong change for others. In other words, while some plurisexual people demonstrate fluidity, plurisexuality is not synonymous with sexual fluidity (Diamond et al., 2020).

Diamond (2006, 2008a) has long argued that sexuality can be dynamically fluid for some and, further, that some people may change year to year in how they understand and define their sexual identity or orientation. The above findings corroborate this assertion by demonstrating a minority of people that shift sexual identity labels multiple times across adulthood and different heterosexual and LGB+ identities. Given that prior research has predominantly relied on studies of restricted samples over relatively short periods (see Table S1), we also advance prior research by utilizing a large, nationwide panel sample of heterosexual and LGB+ adults across seven years. In doing so, we were able to examine a) the relative stability and fluidity of sexual identity over a one-year period, b) the prevalence of sexual identity change over seven years, and c) the direction of this change. We thus highlight the considerable variability in fluidity across time that would otherwise be underestimated in shorter longitudinal investigations. Likewise, we demonstrate how static constructions of sexual identity may not accurately portray the experiences of people who vary in their identities over time. While future research is needed to understand how this fluidity persists over time (i.e., over the life span), these findings support Diamond’s assertions of the context-dependent nature of sexual identity and the need to consider constructions of sexual orientation beyond static categories.

Does Fluidity Change with Age?

Concerning age, younger (versus older) adults were more likely to show change, particularly toward plurisexual identities. While expected (see Greaves et al., 2017), whether this pattern reflects a cohort effect or a general tendency for young people to be more fluid than older people requires further investigation. An important caveat also emerged, such that initially plurisexual people showed high rates of change across the lifespan. This underscores the possibility that plurisexuality may represent an inherently fluid identity for some (Diamond, 2016; Ross et al., 2012; Weinberg et al., 1995) and adds nuance to our understanding of the age-fluidity link by showing this is qualified by sexual identity.

Are Women More Sexually Fluid Than Men?

We also examined the associations between gender and fluidity in sexual identity over time. While there has been much excitement about women showing erotic plasticity (Baumeister, 2000), our study did not support this notion. Indeed, while more women reported being plurisexual than men, they did not show more sexual identity fluidity over the eight assessment occasions (cf. Dickson et al., 2013). Moreover, among those who reported a plurisexual identity, men were more fluid than women. These findings challenge the notion of the inherent fixedness of male relative to female sexuality and demonstrate that a focus on aggregate levels of fluidity both a) reveals a lack of overall gender differences and b) masks gender differences within sexual identities, such that plurisexual men are more fluid than plurisexual women.

What are the Ideological, Health, and Personality Correlates of Sexual Fluidity?

Interesting results also emerged regarding the associations mental health and personality had with fluidity. Indeed, our findings revealed that emotionality was related to the number of sexual identity changes people reported but not when accounting for shared variance with other constructs. Psychological distress, however, was related to the number of sexual identity changes people reported and uniquely predictive of shifting between heterosexual and plurisexual identities. While we could not test the potential explanations for this finding in the present study, it is possible that these findings are explained by minority stress (Hoy-Ellis, 2023; Meyer & Schwitzer, 1999) or that people who experience distress may be more likely to explore their identities in attempts to find meaning. Future research is needed that considers these possibilities – among others – when examining the relationships between mental health and sexual identity fluidity. Moreover, openness to experience and less conscientiousness predicted changes toward plurisexual identities and the number of changes reported. These findings potentially speak to the inherent openness of plurisexual identities (Ross et al., 2012) and corroborate past research demonstrating the personality correlates of different sexual identities (e.g., Allen & Robson, 2020).

Finally, more politically conservative participants were less likely to change, reported fewer changes if they changed at all, were more likely to switch from plurisexual to heterosexual, and were less likely to change from heterosexual to plurisexual (relative to those more politically liberal). Moreover, more religious people also reported fewer changes from heterosexuality to plurisexuality (relative to those who were less religious). Each finding is consistent with the idea that conservatives are generally more negative toward LGB+ people and identities (Herek, 2000). At first glance, this finding might indicate that conservatives who “come out” are more likely to go “back in,” potentially because of their beliefs or pressure from their ingroup (Pachankis et al., 2017; PRRI, 2023; Scheitle & Wolf, 2018). While this is a possibility, future research is needed to examine the cause of these changes in identity over time.

Caveats and Future Research Directions

While the current study presents a rigorous investigation of sexual identity change in a large, national probability sample of adults, there are limitations worth considering. First, we assessed sexual identity fluidity with the underlying assumption that identity reflects (although imperfectly) attractions and behavior. Behavior and attraction, however, show more change over time than identity (Dickson et al., 2003, 2013; Spittlehouse et al., 2019). Thus, our results can only speak to identity fluidity, and future research should examine these three dimensions of fluidity in tandem to thoroughly examine sexual fluidity in adulthood.

We should also note that societal constraints affect one’s reporting of their sexual orientation. While we argued that sexual fluidity needs to be examined in contexts free from legal and social constraints, we could not entirely do so. Sexual prejudice is a serious (and ongoing) problem (Herek, 2000; Teal & Conover-Williams, 2016), and some of the change away from LGB+ identities toward heterosexual identities may reflect societal or familial pressure and discrimination rather than genuine fluidity. On the other hand, people may have also felt able to report their sexual identity more freely in later waves of the study (although our decision to pool data across waves would help to adjust for this). Examining sexual fluidity change and reported experiences of discrimination and hardship would help further understand how and why one’s reported sexual identities may change over time. Notably, as any deviation from heterosexuality (including stable LGB+ identities) will likely vary based on the social or political context, cross-national studies are needed to examine how different socio-political contexts affect one’s sexual identity and fluidity (or lack thereof).

We also assessed fluidity over eight annual waves of data, which, although relatively extensive, still provides a limited window of assessing identity change. It is possible, for example, that more people could follow the linear developmental model over the entire lifespan. Likewise, some of the plurisexual mobility we observed may be part of a longer journey toward a gay or lesbian identity (or vice versa). Shifts in sexual identity may also reflect “experimentation” among LGB+ individuals, given that many people experience continued sexual orientation development throughout young adulthood (Diamond, 2006; Kaestle, 2019; see also Rosario, 2019, for discussion) and changes in one’s socio-political context (namely, greater equality and legal protections) may lead to delayed experimentation in later adulthood. Finally, while our findings largely corroborate previous sexual fluidity research, our analyses of sexual identity changes as a proxy for fluidity may not account for random change or “noise” in participants’ responses. While there is no way to resolve these issues at present, future research examining sexual fluidity over a longer period of time will help further elucidate how sexual identity changes and develops across the lifespan.

Although we used a large national probability panel study that aimed to be representative of the New Zealand population, there were some discrepancies (such that our sample consisted of more women and European people than in the general population). Sociocultural norms and beliefs about sexual identity may preclude whether, when, and how sexual fluidity occurs (Garnets, 2002). As such, understanding the intersections between sexual identity and ethnicity is a necessary avenue for future research. Moreover, we were limited by the small number of LGB+ participants and lacked the sample size to explore sexual identity for those identifying as asexual. Indeed, asexuality is a vastly neglected avenue for sexual fluidity research (for a recent exception, see Su & Zheng, 2023), and future research should examine the experiences of a broader range of LGB+ people.

Finally, previous research suggests that transgender and gender-diverse (TGD) people have markedly different sexual identity journeys than their cisgender counterparts (e.g., Galupo et al., 2016; Katz-Wise et al., 2016, 2017). While TGD participants were included in our overall analyses of sexual fluidity, we were unable to assess prevalence and change in sexual fluidity among TGD people specifically, nor whether they differed from their cisgender counterparts. Additionally, our single-item gender measure likely underestimates the number of TGD participants in our sample (see Fraser, 2018, for discussion). Accordingly, caution is needed when generalizing our results to TGD people, and future research could utilize two-step gender measures (e.g., measures of gender identity and sex) to appropriately assess the relationship(s) between sexual fluidity and gender diversity.

Conclusion

In sum, our results provide data on sexual identity fluidity, in line with the preeminent work by Diamond (2008b, 2021). Overall, our work shows that sexual identity is fluid for a meaningful minority of people and challenges the notion that such change is directional or the domain of women. We also showed that change itself was predicted by demographics, political conservatism, and personality, highlighting the dynamic nature of sexual identity across adulthood. This is important, given that many national censuses are moving toward measuring sexual orientation. Yet, our results show that who is counted as part of the LGB+ community in any given snapshot will likely change over time. Thus, the present study provides the foundation for future investigations into the expanding and ever-changing ways we understand human sexualities.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Data Availability Statement

The data described in the paper are part of the New Zealand Attitudes and Values Study (NZAVS). Full copies of the NZAVS data files are held by all members of the NZAVS management team and advisory board. A de-identified dataset containing the variables analyzed in this manuscript is available upon request from the corresponding author, or any member of the NZAVS advisory board for the purposes of replication or checking of any published study using NZAVS data. The Mplus syntax used to test all models reported in this manuscript will be available on the Open Science Framework upon publication: https://osf.io/75snb/.

Supplementary Data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00224499.2023.2289517.

Additional information

Funding

The current study was supported by a Templeton Religion Trust grant [TRT-2021-10418] awarded to the New Zealand Attitudes and Values Study (NZAVS). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the article.