Introduction
There have long been popular narratives on both sides of the political aisle about LatinoFootnote 1 partisanship. Democrats often appeal to Latinos on the topics of racial prejudice and immigration, especially when Republicans are much more conservative on such issues (Corral and Leal Reference Corral and Leal2024; Garcia-Rios, Pedraza, and Wilcox-Archuleta Reference Garcia-Rios, Pedraza and Wilcox-Archuleta2019; Gutierrez et al. Reference Gutierrez2019; Ocampo, Garcia-Rios, and Gutierrez Reference Ocampo, Garcia-Rios and Gutierrez2021). Conversely, outreach from Republicans towards Latinos often emphasizes economic self-sufficiency alongside conservative positions on social issues like abortion, LGBT rights, and crime and law enforcement (Cadava Reference Cadava2020; Cortez Reference Cortez2020; de la Garza and Cortina Reference de la Garza and Cortina2007; Hickel, Alamillo, et al. Reference Hickel and Alamillo2020; Hickel, Oskooii, and Collingwood Reference Hickel, Oskooii and Collingwood2024; Wakefield Reference Wakefield2025). However, both the “(Dem)ography as destiny” and “Republicans but they just don’t know it” narratives that these strategies reflect have largely failed to be borne out in studies of aggregate Latino voting behavior in 2016 and elections afterward, which have consistently found that the vote choice and partisan identity of Latinos has either remained relatively stable, or shifted Republican (Corral and Leal Reference Corral and Leal2020; Reference Corral and Leal2024; Dyck and Johnson Reference Dyck and Johnson2022; Fraga, Velez, and West Reference Fraga, Velez and West2024; Hopkins, Kaiser, and Pérez Reference Hopkins, Kaiser and Pérez2023). Furthermore, rates of Independent identification among Latinos remain significantly higher than among their White and Black counterparts, indicating that many Latinos remain on the partisan sidelines (Fraga Reference Fraga2018; Hajnal and Lee Reference Hajnal and Lee2011).
In this paper, we address these narratives regarding Latino partisanship and extend the existing literature by examining shifts over a three-decade period—1989 to 2023—using available high-quality survey data on Latino political opinions. This time frame covers large shifts in each party’s positions on race generally and immigration specifically (Abrajano and Hajnal Reference Abrajano and Hajnal2015; Tesler Reference Tesler2016), and also extends into the most recent available data (2023). Additionally, we take special care to account for the extensive diversity of the Latino population by using demographic data in each survey in conjunction with Census demographic data to correct for potential imbalances (Beltrán Reference Beltrán2010; Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024). Many of our analyses also emphasize change among subgroups, as significant shifts among even small groups can potentially drive aggregate change (Fraga, Velez, and West Reference Fraga, Velez and West2024; Grimmer, Marble, and Tanigawa-Lau Reference Grimmer, Marble and Tanigawa-Lau2024). In particular, we focus on subgroup effects based on immigrant generation, age, and national origin (Fraga et al. Reference Fraga2011; García Bedolla and Hosam Reference García Bedolla and Hosam2021; Jones-Correa Reference Jones-Correa1998; McCann and Jones-Correa Reference McCann and Jones-Correa2020; Wong Reference Wong2006). The Latino population is one of the fastest-growing ethnic groups in the country and is also disproportionately young, which means that the Latino proportion of the electorate will only continue to increase while simultaneously becoming younger on average (Igielnik and Budiman Reference Igielnik and Budiman2020; Funk and Lopez Reference Funk and Lopez2022). Furthermore, while previous cohorts of Latinos were more likely to be foreign-born, younger Latinos are more likely to be U.S.-born (Igielnik and Budiman Reference Igielnik and Budiman2020; Funk and Lopez Reference Funk and Lopez2022). This suggests that even absent any individual-level opinion changes, Latino partisanship could shift significantly due to changes in the overall composition of the Latino electorate. Finally, we examine differences between national origin groups, with a focus on groups with sufficiently large and representative samples (Mexican, Puerto Rican, and Cuban origin). Ultimately, our goal is to provide an updated snapshot of Latino partisanship from 1989 through 2023 while taking into account the specific dynamics affecting Latinos regarding immigration, national origin, and generational change.
To create our dataset, we collected data from 35 different public opinion surveys ranging from 1989 to 2023 to create a sample of 103,561 Latino survey respondents.Footnote 2 We additionally used demographic and geographic data from these surveys along with Census data to calculate post-stratified survey weights, which help correct for potential demographic non-representativeness in any individual survey (Erikson, Mackuen, and Stimson Reference Erikson, Mackuen and Stimson2001; Ghitza and Gelman Reference Ghitza and Gelman2013; Leemann and Wasserfallen Reference Leemann and Wasserfallen2017; Mercer, Lau, and Kennedy Reference Mercer, Lau and Kennedy2018). Then, we estimated rates of Democratic, Republican, and Independent identificationFootnote 3 among Latinos over time along with additional figures for how this differs based on nativity, age/birth cohort, and national origin (Mexico, Puerto Rico, Cuba, or another country).Footnote 4 While this dataset is not a true panel, the inclusion of multiple datasets from several providers means that our results are less likely to be biased for an individual year or by an individual survey provider’s methodology. Our post-stratification weights also help correct for any individual survey having less representative demographics. We thus argue that this data can be taken as a synthetic survey panel that represents the closest possible estimate, given available data (L. R. Fraga et al. Reference Fraga2006; Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024), about how Latino partisanship has shifted both overall and for stable individual cohorts over the 1989–2023 time period.
In our results, we find notable shifts in Latino partisanship during particular time periods—although not always in the expected direction. While we find a shift toward greater Democratic identification around 2008, which corresponds with Latino outreach by the Obama campaign, we do not observe evidence of an anti-Trump backlash in 2016 and afterward. Instead, after 2016 and through 2023, we find that Latinos overall and especially those who are foreign-born became less likely to identify as either Democratic or Republican, and instead were even more likely to identify as Independents. We then emphasize trends among Latinos born in the continental United StatesFootnote 5 to address changes among this growing group. Our age subgroup analyses show that older U.S.-born Latinos are shifting from a historic trend of Democratic identification from the 1990s through the 2000s towards relatively higher Republican identification after 2008, while younger U.S.-born Latinos are becoming more likely to identify as Independent in the post-2016 period. Across subgroups, we find large increases in Independent identification from 2016 through 2023. Lastly, while we find distinct national origin subgroup differences among U.S.-born Latinos during the 1990s and 2000s—with Mexican Americans generally Democratic, Puerto Ricans even moreso, and Cuban Americans generally Republican—these differences largely disappeared by 2016. This is most surprising among U.S.-born Cuban Americans, who we find are actually now more Democratic than Republican identifying on average, and we also find a slight increase in Republican identification among Puerto Ricans from 2016 to 2023. Overall, these results indicate an accelerating generational divide in Latino partisanship, the declining importance of national origin, and growing rates of Independent identification among Latinos generally.
We conclude by discussing how our results relate to the extant literature on Latino political behavior, identifying and addressing potential caveats to our study, and exploring implications for Latino politics and American elections more broadly. As a large and relatively young population, Latinos are growing in electoral importance but are also still relatively less likely to have partisan attachments compared to non-Hispanic Whites and African Americans (Fraga Reference Fraga2018; García Bedolla and Michelson Reference García Bedolla and Michelson2012; Hajnal and Lee Reference Hajnal and Lee2011; Wong Reference Wong2006). Every election, more and more young Latinos are entering the electorate, and it is crucial to examine how existing literature does—or does not—provide leverage on their political experiences (Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024). By examining these temporal trends in the partisan identities of a diverse range of Latinos during a time of intense political change, we provide a snapshot of the potential future of Latino political behavior and partisanship.
Theories of partisan identity
Partisanship and partisan identity have become an increasingly powerful factor for explaining political behavior in American politics (Campbell, Miller, and Stokes Reference Campbell, Miller and Stokes1980; Green, Palmquist, and Schickler Reference Green, Palmquist and Schickler2004; Huddy, Mason, and Aarøe Reference Huddy, Mason and Aaroe2015; Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2019). As partisan polarization has increased overall, partisan attachments have similarly become an even more important factor for political behavior and even areas beyond politics such as social relationships (Bartels Reference Bartels2018; Huddy, Mason, and Horwitz Reference Huddy, Mason and Horwitz2016; Mason Reference Mason2015). This is relevant because theories of political socialization emphasize how partisanship is shaped by one’s sociopolitical environment, is learned at an early age, and is often highly affected by the political views of parents and peers (Berelson, Lazarsfeld, and McPhee Reference Berelson, Lazarsfeld and McPhee1954; Campbell, Miller, and Stokes Reference Campbell, Miller and Stokes1980; Green, Palmquist, and Schickler Reference Green, Palmquist and Schickler2004; Sears and Valentino Reference Sears and Valentino1997). Furthermore, while partisanship and individual views on various policies tend to be strongly correlated (Webster and Abramowitz Reference Webster and Abramowitz2017), some studies have argued that partisan attachments are actually the primary driver of policy views, which would suggest that identification with one’s respective partisan “team” is an important driver of political engagement (Achen and Bartels Reference Achen and Bartels2017; Lenz Reference Lenz2013). These studies demonstrate that an individual’s partisan identity is largely reliant on their lived context and that such identities are fairly durable once developed.
Still, there are important empirical and theoretical counterexamples to this general trend of individual partisan stability that suggest that both individual and macro-level partisanship can shift in response to various stimuli (Erikson, Mackuen, and Stimson Reference Erikson, Mackuen and Stimson2001; Tesler Reference Tesler2015). These include major policy shifts by either party, salient political events, and even politically relevant personal experiences. Examples of these include being exposed to the Vietnam draft and becoming more Democratic and switching parties based on partisan polarization on salient issues like race or abortion (Carmines and Stimson Reference Carmines and Stimson1989; Carmines and Woods Reference Carmines and Woods2002; Carsey and Layman Reference Carsey and Layman2006; Reny, Valenzuela, and Collingwood Reference Reny, Valenzuela and Collingwood2019; Reny, Collingwood, and Valenzuela Reference Reny, Collingwood and Valenzuela2019; Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2019). A related study by Peterson, Smith, and Hibbing (Reference Peterson, Smith and Hibbing2020) tests the “folk wisdom” that individuals tend to be more liberal (and thus Democratic) when they are younger and to become more conservative (and thus Republican) over time; they find that while an individual’s pre-existing partisan environment and identity tends to matter more than age by itself, there is some evidence that older people are more likely to shift their partisanship in response to changes in the political context (Peterson, Smith, and Hibbing Reference Peterson, Smith and Hibbing2020). These studies demonstrate that while partisanship is fairly fixed and crystallized (Tesler Reference Tesler2015), there are many potential ways it can still shift at the individual level and in aggregate.
What makes Latino partisan identity unique?
While these overarching studies are helpful for informing our questions about Latino partisan socialization, there are also specific factors—like the large share of Latinos who are either immigrants or the children of immigrants—that play an important role (Funk and Lopez Reference Funk and Lopez2022; García Reference García2016; Igielnik and Budiman Reference Igielnik and Budiman2020). Scholars have often argued that such immigrant ties can cause Latinos to react strongly in favor of politicians supporting immigrant rights and against anti-immigrant politicians and policies, which is then posited to lead to increased Democratic and decreased Republican support among Latinos overall (Barreto and Collingwood Reference Barreto and Collingwood2015; García Bedolla Reference García Bedolla2005; Garcia-Rios, Pedraza, and Wilcox-Archuleta Reference Garcia-Rios, Pedraza and Wilcox-Archuleta2019; Gutierrez et al. Reference Gutierrez2019; Pérez Reference Pérez2015; Saavedra Cisneros Reference Saavedra Cisneros2017). However, Latino parents who are immigrants often do not have as much socialization into American politics to pass on to their children (Carlos Reference Carlos2018; Fraga Reference Fraga2018; Hajnal and Lee Reference Hajnal and Lee2011; Pérez and Cobian Reference Pérez and Cobian2024). Consequently, younger (and predominantly U.S.-born) Latinos oftentimes take on the role of informing their older family members about politics (Carlos Reference Carlos2021; García-Castañon Reference García-Castañon2018; García-Castañon et al. Reference García-Castañon2019). This means that examining the political behavior of younger and U.S.-born Latinos, relative to their older and foreign-born counterparts, can provide a crucial snapshot into some of the drivers of partisan socialization that would not be emphasized by a parental socialization model.
Immigration
Given the number of Latinos with immigrant ties, immigration policy and rhetoric are especially important for considering Latino partisanship and partisan change (McCann and Jones-Correa Reference McCann and Jones-Correa2020; Pérez and Cobian Reference Pérez and Cobian2024).Footnote 6 Over recent decades, each party has shifted quite significantly in their stances on immigration—Democrats away from enforcement-oriented approaches toward policy stances that are more supportive towards undocumented immigrants, and Republicans away from relatively moderate stances on immigration during the Reagan era towards supporting increasingly militarized immigration enforcement and xenophobic rhetoric (Abrajano and Hajnal Reference Abrajano and Hajnal2015; Golash-Boza Reference Golash-Boza, Brotherton and Kretsedemas2018; Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2019). Still, while such shifts in each party’s position on immigration were expected to lead to ensuing shifts in Latino partisanship, evidence of this remains mixed. Previous research on anti-immigrant contexts and policies, such as California’s Proposition 187 (1994) or HR 4437 (2006), found that Latinos responded forcefully to such threats through community organizing, protests, naturalizing as citizens, and registering to vote (Barreto, Manzano, et al. Reference Barreto and Manzano2009; García Bedolla Reference García Bedolla2005; HoSang Reference HoSang2010; Pantoja, Ramirez, and Segura Reference Pantoja, Ramirez and Segura2001; Pantoja and Segura Reference Pantoja and Segura2003; Zepeda-Millán Reference Zepeda-Millán2017). However, a study by Hui and Sears (Reference Hui and Sears2018) of shifts in partisan registration among Latinos in California during the same time as Proposition 187 failed to find significant effects of the proposition on Latino partisan identification (Hui and Sears Reference Hui and Sears2018). More recently, while some research suggested Obama appealed to Latinos in 2008 and 2012 through immigration outreach and Trump angered many Latinos with his racist and xenophobic rhetoric in 2016 (Barreto and Collingwood Reference Barreto and Collingwood2015; Collingwood, Barreto, and Garcia-Rios Reference Collingwood, Barreto and Garcia-Rios2014; Gutierrez et al. Reference Gutierrez2019; Garcia-Rios, Pedraza, and Wilcox-Archuleta Reference Garcia-Rios, Pedraza and Wilcox-Archuleta2019), other studies of aggregate Latino partisanship during and after the 2016 election have instead found that Latino partisanship has remained relatively stable during this time period (Corral and Leal Reference Corral and Leal2020; Dyck and Johnson Reference Dyck and Johnson2022; Fraga, Velez, and West Reference Fraga, Velez and West2024; Hopkins, Kaiser, and Pérez Reference Hopkins, Kaiser and Pérez2023; McCann and Jones-Correa Reference McCann and Jones-Correa2020). While the canonical expectation from existing literature may be that partisan polarization on immigration would lead to a shift toward Democratic identification among Latinos overall (Corral and Leal Reference Corral and Leal2024; Jones-Correa, Al- Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018), evidence of this effect is less apparent when examining Latino macro-partisanship and is even less apparent in the post-2016 time period.
Two factors potentially complicate the relationship between immigration politics and Latino partisanship. First, recent research has identified a small but still significant minority of Latinos who express racially conservative views, oppose undocumented immigration, and tend to identify as Republicans (Cortez Reference Cortez2020; Fraga, Velez, and West Reference Fraga, Velez and West2024; Hickel, Oskooii, and Collingwood Reference Hickel, Oskooii and Collingwood2024; Hickel, Alamillo, et al. Reference Hickel and Alamillo2020; Pérez, Robertson, and Vicuna Reference Pérez, Robertson and Vicuna2023). Second, the Latino population is fairly young and increasingly U.S.-born, which means they are potentially less likely to view immigration as highly salient without active effort to maintain such attachments. As of 2020, only one-third of the Latino population was foreign-born, with more recent generations being even more likely to be born in the United States (Funk and Lopez Reference Funk and Lopez2022). These younger Latino generational cohorts may be more likely to have politicizing social networks and greater internal political efficacy compared to their older peers (Carlos Reference Carlos2018; Carlos Reference Carlos2021; Raychaudhuri Reference Raychaudhuri2020). However, younger people in general tend to participate in politics and identify as partisans at lower rates than those who are older (Schlozman, Verba, and Brady Reference Schlozman, Verba and Brady2013; Wattenberg Reference Wattenberg2020). Therefore, the impact of generational change may lag behind shifts in population size, as younger cohorts steadily become older and potentially become more engaged and partisan-identified over time. As the Latino electorate continues to grow while also remaining disproportionately younger, it is thus crucial to account for age when estimating Latino partisanship.
Age
Beyond examining trends in the overall and native-born Latino population, another goal of this paper is to answer these unaddressed questions about changes in partisanship among Latinos based on their age. Due to our focus on generational change, we emphasize shifts within commonly used generational birth cohorts. We use definitions for birth cohorts established by the Pew Research Center, which classifies individuals into generations based on year of birth to create non-overlapping categories. For most of this study, we focus on four birth cohorts: the Silent Generation (born 1928–1945), the Baby Boom Generation or “Boomers” (born 1946–1964), Generation X (born 1965–1980), and the Millennial Generation or “Millennials” (born 1981–1996). We also include respondents from outside of these four generational categories if they are present in a survey of Latinos: the Greatest or “G.I.” Generation (born 1901–1927) and Generation Z (born 1997–2012).
Table 1 displays the share of the adult Latino population that each birth cohort made up in 1990, 2000, 2010, and 2020, which we calculated using U.S. Census Bureau microdata. The top shows these for the entire population and the bottom for only Latinos who were born in the United States.Footnote 7 Overall, we find distinct periods when certain generations are almost a majority—Baby Boomers in 1990, Generation X in 2000, and Millennials in 2020. Even Generation Z, which only appears in the 2020 time period, makes up an appreciable share of the adult Latino population. The most notable difference is between the overall Latino population and the U.S.-born over time: while generational shares are largely similar in 1990 and 2000 regardless of nativity, the U.S.-born population has significantly more Millennials and Generation Z’s in 2010 and 2020 than the overall Latino population. While the Latino population previously did not have large generational differences between the U.S. and foreign-born subsets through the 2000s, this table reveals a disproportionately large cohort of Millennial and Gen Z Latinos among the U.S.-born (49% vs. 62% in 2020). These younger U.S.-born Latinos could potentially create large shifts in Latino partisanship during our observed time frame if their behavior is distinctive from older generations, who are also disproportionately more foreign-born.
Table 1. Percent of Latinos in each generation
Note: Cells indicate each generation’s share of the 18+ Latino population. Years indicate birth years included in each generational cohort.
National origin
Another important trend to consider for Latino partisanship is the role of national origin. Partisan differences among Latino subgroups are often related to the divergent experiences of different national origin groups: Mexican Americans and Puerto Ricans have generally been more consistently Democratic in their attachments overall while Cuban Americans have historically been the most Republican Latino group (Cadava Reference Cadava2020; Francis-Fallon Reference Francis-Fallon2019; García Bedolla and Hosam Reference García Bedolla and Hosam2021; Mora Reference Mora2014; Sánchez Reference Sánchez2007).Footnote 8 These different experiences include dynamics like regional settlement patterns, differential treatment by the U.S. political system, and pre-migration differences in socioeconomic status (Jones-Correa and Leal Reference Jones-Correa and Leal1996; Mora Reference Mora2014; Oboler Reference Oboler1995; Padilla Reference Padilla1985; Torres Reference Torres2001). For example, the large and growing Mexican American population and their concentration in pivotal states such as California and Arizona, alongside anti-immigrant efforts by Republican politicians, led to Mexican Americans becoming increasingly important to Democratic coalitions (Barreto, Collingwood, and Manzano Reference Barreto, Collingwood and Manzano2010; Francis-Fallon Reference Francis-Fallon2019; Garcia-Rios, Pedraza, and Wilcox-Archuleta Reference Garcia-Rios, Pedraza and Wilcox-Archuleta2019; Mora Reference Mora2014). Similarly, Puerto Ricans also often faced issues of poverty and racial discrimination, which led them to support Democrats (Beltrán Reference Beltrán2010; Padilla Reference Padilla1985; Sánchez Reference Sánchez2007). Meanwhile, earlier generations of Cuban Americans came primarily from middle and upper-class exiles from the country’s communist revolution and became staunch Republicans (Torres Reference Torres2001; Mora Reference Mora2014; Francis-Fallon Reference Francis-Fallon2019). A detailed examination of these different histories is beyond this paper’s scope,Footnote 9 but in our analyses, we will examine how the partisanship of each of these three groups shifts over time to help provide a historical-to-contemporary snapshot of how the trends described above affect Latino macro-partisanship.
Data
Our analysis draws on all available and high-qualityFootnote 10 surveys of Latinos from the earliest of such surveys (the 1989 LNPS)Footnote 11 and leverages the post-2000 growth in the number of large-sample surveys covering the Latino population (Fraga et al. Reference Fraga2006; Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024). We draw from 35 nationally representative surveys that contain a large sample of Latino respondents, include demographic information relevant to the Latino population, and ask respondents about their party identification and partisan leanings.Footnote 12 Our data includes the 1989 Latino Political National Survey (LNPS), the 2006 Latino National Survey (LNS), the 1999 Kaiser Family Foundation Survey, Pew Research Center’s National Survey of Latinos which was conducted from 2002 through 2018,Footnote 13 the Collaborative Multiracial Post-Election Surveys (CMPS) in 2016 and 2020, the Congressional Cooperative Election Study (CCES) conducted from 2006 through 2018, and the Cooperative Election Study (CES) conducted from 2019 through 2023.Footnote 14 A summary list of these datasets can be found in Table 2.Footnote 15 Altogether, these 35 surveys produce a dataset that captures information about political behavior, partisanship, and demographics from over 103,000 Latino respondents over a more than 30-year time span.
Table 2. Surveys Included in the Main Analysis
Note: “Latino Sample Size” excludes Latinos whose response for party identification was either “Don’t Know,” other party, a third party, or blank and did not express a leaned partisanship or “Independent” status in a follow-up.
To measure partisanship, we begin with the standard three-point party identification question, which asks whether respondents identify as either Democrat, Republican, as Independent, as “don’t know”, or as a third party. We then use the standard 7-point party identity followup question about the partisan leanings of those who initially do not identify with either party.Footnote 16 We then classified Democratic respondents as those who either initially identified as Democrats or who expressed a lean towards the Democratic Party in the independent leaner followup question, and likewise for Republican identifiers and Republican leaners. Our classification of Independents, then, includes two groups: those who identified as “Independent” in the first party identity question, and those who identified as “Independent” in the followup partisan leaning question. We then excluded respondents who answered “Don’t Know”/“Other Party” for both the partisan identity and partisan-leaning followup questions.Footnote 17 Throughout our analyses, we will focus on differences in the share of our samples who identify as or lean Democratic (Democrats), identify as or lean Republican (Republicans), and identify as Independent without any partisan leaning (Independents).
Our three main independent variables of interest are the respondent’s nativity, their age/birth cohort, and their national origin (Mexican, Puerto Rican, Cuban, and Other).Footnote 18 Using variables about where a respondent was born, we identify whether they were born in the United States (excluding Puerto Rico) or were born outside of the U.S./in Puerto Rico.Footnote 19 For the birth cohort, we reference a respondent’s age and/or birth year to identify which generation they were born in (see Table 1).Footnote 20 We determined national origin by creating a four-part variable classifying each respondent as either Mexican, Puerto Rican, Cuban, or “Other” based on their responses to Latino national origin questions.Footnote 21 We then examine how partisanship among Latinos varies when taking into account these various subgroups over our observed time period.
Analysis
In our analyses, we leverage our large dataset and use demographic data from the Census to calculate post-stratified survey weights. This is important for our study because we use results from surveys that vary substantially in terms of their sampling design, interview type, and overall representativeness.Footnote 22 Fortunately, the surveys we draw from also include individual-level demographic data that helps us to correct for potential biases by adjusting our weights and estimates based on the known demographics of the Latino population from the Census/American Community Survey (ACS). Calculating these post-stratification weights and using them in our analyses helps correct for a potentially biased survey sample (Ghitza and Gelman Reference Ghitza and Gelman2013; Leemann and Wasserfallen Reference Leemann and Wasserfallen2017), which is especially relevant given the limitations of existing data on Latino political behavior (Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024). Using 1990 and 2000 U.S. Census Public Use Microdata Sample (PUMS) data, and 2001–2023 ACS PUMS data for more recent surveys, we estimated post-stratification weights for each survey that account for how close that survey’s Latino sample corresponds to Census/ACS data from the year that the survey was collected. Our post-stratification variables include national origin (Mexican, Puerto Rican, Cuban, Salvadoran, Dominican, and “Other”), nativity (foreign/Puerto-Rican born, or born in continental U.S.), generational age group, education (college graduate or not), gender (male or female), and state/region.Footnote 23 These new weights help correct each individual survey sample, which varies widely in their sample size and how well they represent both the national Latino population and specific subgroups.
Our analyses are descriptive and fairly straightforward given the post-stratification weights that we described above, which help correct for potential imbalances and biases in the survey data. This then provides us with 35 individual surveys with large Latino samples (N = 103,561) over the 1989–2023 time period. Our analyses and figures were conducted using R/RMarkdown and the ggplot2 package. For each figure, we first plot individual survey estimates as a set of three points corresponding to the estimated percentage of Democrats/Democratic leaners, Republicans/Republican leaners, and pure Independents in the given sample.Footnote 24 We then use a loess smoother with each point weighted inversely to its variance to estimate a moving average for each of the three partisan identity subgroups over time, which means that surveys with lower variance/larger samples have a proportionately larger effect on the overall estimate. We argue that given our post-stratification weights and these additional methodological approaches, this dataset represents a synthetic panel of Latinos that accounts for many of the demographic and generational dynamics discussed above and allows us to track change both overall and for stable individual cohorts. We also leverage these demographic variables to provide insights into how partisanship trends differ based on a respondent’s nativity, age, and national origin. Ultimately, we use this data to provide an updated estimate of Latino partisanship that can help link past and recent research on Latino political opinions (Fraga et al. Reference Fraga2006; Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024).
Results
We first display results for all Latino respondents in our sample. Figure 1 displays trends in Democratic, Republican, and Independent identification among Latinos in all of the survey samples we have collected. This figure and all ensuing figures are calculated using our post-stratified demographic survey weights.Footnote 25 Our analyses begin in 1989, where we estimate that Latinos overall were 66% Democratic, 25% Republican, and 9% Independent.Footnote 26 Democratic identification then drops until 2004 while Republican identification slightly increases. From 2004 through 2012, we then observe a steady increase in Democratic identification (60%) and a steady decrease in Republican identification (25%). Finally, after 2016 through 2023, we find declining rates of Democratic identification (51%), slightly increasing rates of Republican identification (28%), and a slight but notable increase in Independent identification (20%).
Figure 1. Partisanship Trends Among Latinos, 1989–2023.
Note: This graph shows a weighted estimate from 35 public opinion surveys of the percentage (Y-Axis) of Latino adults identifying as either Democrat/lean Democrat (blue), Republican/lean Republican (red), or Independent with no partisan leanings (green) over the 1989–2023 time period (X-Axis). Respondents who replied “Don’t Know” or “Other Party” for both the 3-point party identity and partisan leaning followup questions are excluded from these analyses. Points show the average partisanship estimate for each individual survey, which are calculated using post-stratification weights based on Census/ACS demographics. Point size is proportional to survey N. Lines show a loess best-fit estimate using inverse variance weights and a 95% confidence band.
By nativity
Next, we examine trends based on a respondent’s nativity. Figure 2 displays the same graph for partisanship among Latinos but separated with respondents born outside of the United States or in Puerto Rico on the left, and those who were born in the mainland United States on the right. This comparison allows us to examine if there are particular trends for Latino respondents who are not first-generation immigrants themselves.Footnote 27 As with the previous figure, we begin by plotting point estimates (not shown for clarity; see Supplementary Appendix “Additional Results”). These points are calculated using post-stratified weights for each survey and nativity subgroup. We then estimate a loess smoother line through those points with a 95% confidence interval ribbon. In 1989, U.S.-born Latinos were both more Democratic (65%) and less Republican (24%) than their foreign/Puerto-Rican born counterparts (55% and 35%, respectively). Rates of Independent identification were also fairly low among both groups during this period (10%).Footnote 28 From 2000 through 2012, we observe a slight turn among foreign/Puerto Rican-born respondents towards higher levels of Democratic identification (55–60%), more rapidly decreasing Republican identification (20–25%), and increasing Independent identification (18–22%). For U.S.-born Latinos, meanwhile, we find slightly decreasing rates of identification as either Democrat (60%) or Republican (26%) with a corresponding increase in Independent identification (14%). Then, after 2016 and through 2023, we find that foreign/Puerto Rican-born respondents shifted back towards their previous rates of slightly higher Republican identification (35%) with a corresponding drop in Democratic identification (50%). Meanwhile, among U.S.-born respondents over the entire time period and especially post-2016, we find a sharp decrease in Democratic partisanship (50%), a slight decrease in Republican partisanship (22%), and starkly increasing rates of Independent identification (28%). This figure indicates that while foreign-born Latinos were more likely to shift toward Democrats from 2008 to 2012 and towards Republicans post-2016, such trends are less apparent among the U.S.-born, who have become less Democratic and more Independent over time and especially after the 2016 election. The large drop-off in Democratic partisanship, and corresponding increase in Independent identification, is especially notable in surveys after 2020 among U.S.-born respondents.
Figure 2. Partisanship Among Latinos by Nativity, 1989–2023.
Note: This graph shows a weighted estimate from 35 public opinion surveys of the percentage (Y-Axis) of Latino adults identifying as either Democrat/lean Democrat (blue), Republican/lean Republican (red), or Independent with no partisan lean (green) over the 1989–2023 time period (X-Axis). Respondents who answered “Don’t know“/”Other Party” for both the initial partisan identity and the partisan leaning followup questions are not included. Points (not shown for clarity) are based on the average partisanship estimate for each individual survey, which are calculated using post-stratification weights based on Census/ACS demographics. Lines show a loess best-fit estimate using inverse variance weights and a 95% confidence band. The left graph shows results for respondents born outside of the U.S. or in Puerto Rico and the right graph shows results for those born in the US.
By birth cohort
Next, we analyze our data by birth cohort to address our questions related to generational differences. To simulate a panel study tracking each generation over time, we use the respondent’s birth year/age, place them in a given generational cohort for each survey-year, and then calculate estimates for each subgroup over time. Due to our emphasis on generational trends among Latinos born in the United States, this figure only includes results from native-born respondents.Footnote 29 We plot this in Figure 3 and show trends for each generational birth cohort separately. We again first plot point estimates with post-stratification weights for each survey/generation/yearFootnote 30 and estimate a loess best-of-fit line to create a moving average of Latino partisanship over time by birth generation.
Figure 3. Partisanship Among U.S.-Born Latinos by Birth Cohort, 1989–2023.
Note: This graph shows a weighted estimate from 35 public opinion surveys of the percentage (Y-Axis) of U.S.-born Latino adults identifying as either Democrat/lean Democrat (blue), Republican/lean Republican (red), or Independent (green) over the 1989–2023 time period (X-Axis). Respondents who answered “Don’t know“/”Other Party” to both the initial partisan identity and the partisan leaning followup questions are not included. Points (not shown for clarity) are based on the average partisanship estimate for each individual survey, which are calculated using post-stratification weights based on Census/ACS demographics. Lines show a loess best-fit estimate using inverse variance weights and a 95% confidence band. The graphs are then arrayed vertically for each generational subgroup, which is determined based on respondent age and survey year, such that older generations are at the top and younger generations are at the bottom.
We first describe our results in Figure 3 for the three generations in the electorate in 1989. For the Silent generation of U.S.-born Latinos, who are the oldest in our sample, we observe significantly high rates of Democratic identification (75–80%) and low Republican identification (10–15%) from around 1989 through 2008. However, for Latinos in the Silent generation from 2008 through 2012, we find a decrease in Democratic identification (60%) and a notable increase in Republican identification (25%). From 2012 through 2023, we then observe a continued increase in Republican identification (32%) and decrease in Democratic identification (52%). Independent identification among the Silent generation remains fairly low across our observed time period (10–12%), with a slight increase post-2020 (16%). Among U.S.-born Latinos in the Baby Boomer generation in the 1990s and 2000s, we find weaker Democratic identification (65%) and stronger Republican identification (23%) than their Silent counterparts. But while Democratic identification decreased for U.S.-born Latino Boomers over the observed time period through 2023 (55%), rates of Republican identification remained more or less stable (25–29%) whereas Independent identification steadily increased (to 17%). GenX respondents were even less Democratic (48%) and even more Republican (37%) than older generations during the earliest time periods. Then, their rates of Democratic partisanship remained stable while their rates of Republican partisanship slightly decreased (28%). As with other groups, we find GenX Latinos became more Independent after 2016 (24%). For these older three generational cohorts, we find large generational differences at the earliest time periods—with older U.S.-born Latinos being more Democratic and younger cohorts being more Republican—but also find that these trends actually flipped around the 2012 and 2016 elections, when the Silents became the most Republican overall with Boomers following close behind.Footnote 31 More notably, rates of Independent identification generally increased across all subgroups in the post-2016 time period.
We next examine trends in Figure 3 for the youngest two generations of U.S.-born Latinos, who became old enough to vote during our observed time period. In 2000, which is the earliest time period when Millennials were voting-age, they were—like GenX—relatively less Democratic (60%) than their older counterparts. However, they were also less Republican-identified (27%), and instead were more likely than any other generation at the time to identify as Independent (18%). Then, while rates of Democratic identification among U.S.-born Millennial Latinos remained stable through 2008, it actually began steadily dropping off to a low of 48% in 2023. Republican identification also remains stable through 2023 (24%). Meanwhile, rates of Independent identification among U.S.-born Latino Millennials sharply increased to a maximum of 29% in 2023. For the youngest generation in our sample, GenZ, we find a broadly similar trend. While Democratic identification was higher pre-2016 (65%), it steadily decreased by 2023 to 49%. Rates of Republican identification also remain very low for this youngest cohort (18%) relative to older generations. Lastly, by 2023, rates of Independent identification among this youngest cohort are higher than for any other generational subgroup (33%).
We observe two distinct generational trends in our analyses of U.S.-born Latinos. First, while older generations tended to be relatively more Democratic than younger generations from 1989 into the early 2000s, they flipped towards increasing Republican identification during the Obama years and maintained their relatively higher level of Republican support through Trump and Biden (although it is important to note that even these older generations of U.S.-born Latinos remained at least 20% more Democratic-identified than Republican-identified). Second, for younger generations, rates of both Democratic and Republican partisanship are lower than for any Latino group in past decades. Instead, these younger cohorts of Latinos born in the United States are increasingly identifying as Independent. Across subgroups, we find that Independents are the fastest-growing partisan identity group.
By national origin
Lastly, in Figure 4, we examine trends in partisanship among U.S.-born Latinos with Mexican, Puerto Rican, and Cuban national origins.Footnote 32 Again, we restrict our results to just respondents who were born in the continental United States (i.e., not on the island of Puerto Rico).Footnote 33 We then use survey-level point estimates for each partisan identity/national origin (not shown)Footnote 34 to plot a smoothed loess line, which provides a moving estimate of partisanship within each national origin subgroup over time. For Mexican American respondents at the earliest time period, we find that they were fairly likely to identify as Democratic (56–59%), somewhat as Republican (25–28%), and not very often as independent (9–12%). From 2000 through 2012, we then observe stable rates of Democratic and independent identification and a slight decrease in Republican identification among Mexican Americans. After 2012 through 2023, however, we then find decreasing rates of both Democratic (50%) and Republican identification (23%) and instead find a sharp increase in Independent identification (27%). It is notable that these U.S.-born Latinos with Mexican heritage only exhibited a slight drop-off in Republican identification after 2016, when Donald Trump was on the ballot. Furthermore, rates of Democratic identification dropped during this same time-period. Overall, our estimates show that U.S.-born Mexican Americans are increasingly distancing themselves from either party and instead identifying more as Independent.
Figure 4. Partisanship Among U.S.-Born Latinos by National Origin, 1989–2023.
Note: This graph shows a weighted estimate from 35 public opinion surveys of the percentage (Y-Axis) of native-born Latino adults identifying as either Democrat/lean Democrat (blue), Republican/lean Republican (red), or Independent (green) over the 1989–2023 time period (X-Axis). Respondents who answered “Don’t know“/”Other Party” to both the initial partisan identity and the partisan leaning followup questions are not included. Points (not shown for clarity) are based on the average partisanship estimate for each individual survey, which are calculated using post-stratification weights based on Census/ACS demographics. Lines show a loess best-fit estimate using inverse variance weights and a 95% confidence band. The left graph shows results for Mexican-origin respondents, the middle for Puerto Rican-origin respondents, and the right for Cuban-origin respondents. Note that sample sizes for other national origin groups (e.g., Dominicans, Salvadorans) were too small to produce reliable estimates.
Among Puerto Ricans born on the U.S. mainland in the earliest time periods, we find slightly higher Democratic identification (70–73%) and slightly lower Republican identification (19–22%) than the full sample. We also observe an increase from 2008 to 2012 in Democratic identification (73–76%) and a decrease in Republican identification (10–13%). This is then followed by an even sharper post-2012 decrease in Democratic identification (49–52%) and an increase in Independent identification (25–28%). Rates of Republican identification among U.S.-born Puerto Ricans is quite volatile from 2008 through 2023, with a large dip in 2012–2016 followed by a return to modest rates (22–25%). Overall, while U.S.-born Mexican and Puerto-Rican-origin respondents were quite different during earlier time periods, their rates of partisan identification are largely indistinguishable by 2023.
Finally, while Cuban Americans have often been seen as the quintessential example of Latino Republicans, we find some interesting complications to this narrative in our results focusing on U.S.-born Cuban Americans. In the earlier time periods, we observe the expected lower rates of Democratic identification (22–28%) and higher rates of Republican identification (64–70%) along with fairly low rates of Independent identification (10–16%). However, we then find a consistent trend of decreasing Republican and increasing Democratic identification with an intersection point around 2000 when the number of U.S.-born Cuban American Democrats actually outnumbered Republicans. By 2012, U.S.-born Cuban Americans leaned towards Democratic identification (58–62%), although they still remained relatively more Republican-identified (30–34%) than Mexican American and Puerto Rican respondents. While Cuban Americans remain more Republican than other national origin subgroups, rates of Democratic identification still also remain marginally ahead of Republican identification in the most recent data. Rates of Independent identification among Cuban Americans also slightly increased after 2016, although this increase is smaller in magnitude than among Mexican and Puerto Rican origin respondents.
These results from Figure 4 demonstrate that while national origin was once a strong determinant of partisan divergence even among the U.S.-born Latino population, these trends have largely disappeared more recently and especially after the 2016 election. U.S.-born Mexican Americans and Puerto Ricans have instead begun shifting away from their previous Democratic loyalties–but instead of becoming more Republican, both groups have become significantly more Independent over time. Meanwhile, although U.S.-born Cuban Americans shifted strongly towards Democratic identification with a peak during Obama’s second re-election bid in 2012, their partisanship has remained highly unstable in the post-2016 period. Furthermore, as with the overall results, the increase in Independent identification across groups is important to note, as this trend suggests declining rates of partisan identification for multiple subgroups of the Latino population.
Discussion
In this paper, we contribute to a growing literature addressing trends in Latino partisanship and extend its time frame into the most recently available data (Alvarez and García Bedolla Reference Alvarez and Garcia Bedolla2003; Corral and Leal Reference Corral and Leal2020; de la Garza and Cortina Reference de la Garza and Cortina2007; Fraga, Velez, and West Reference Fraga, Velez and West2024; Hopkins, Kaiser, and Pérez Reference Hopkins, Kaiser and Pérez2023; Huddy, Mason, and Horwitz Reference Huddy, Mason and Horwitz2016; Saavedra Cisneros Reference Saavedra Cisneros2017; Dyck and Johnson Reference Dyck and Johnson2022). Using our aggregate analyses of 35 Latino political opinion datasets that ranged from 1989 through 2023, we find distinctive trends in Latino Democratic, Republican, and Independent identification that help illuminate findings from previous research. Both Republican appeals toward Latinos during the Bush era and Democratic appeals during the Obama era are reflected in commensurate shifts in Latino partisan identification towards either party (Barreto and Collingwood Reference Barreto and Collingwood2015; Cadava Reference Cadava2020; de la Garza and Cortina Reference de la Garza and Cortina2007). However, these shifts are largely confined to immigrant Latinos—among Latinos born in the United States, we find decreasing rates of Democratic identification and increasing Independent identification across our observed time frame. Even more notably, we do not find similar shifts during and after the 2016 elections, when much of the existing literature and lay wisdom suggested that the anti-immigrant rhetoric of Trump would alienate Latinos (Garcia-Rios, Pedraza, and Wilcox-Archuleta Reference Garcia-Rios, Pedraza and Wilcox-Archuleta2019; Gutierrez et al. Reference Gutierrez2019 but c.f. Hickel, Alamillo, et al. Reference Hickel and Alamillo2020; Hui and Sears Reference Hui and Sears2018). Instead, from 2016 through 2023, Republican identification among Latinos remained stable while Democratic identification decreased (Corral and Leal Reference Corral and Leal2020; Fraga, Velez, and West Reference Fraga, Velez and West2024; Hopkins, Kaiser, and Pérez Reference Hopkins, Kaiser and Pérez2023). Meanwhile, Independent identification among Latinos, which was stable from around 2004 through 2012, has steadily increased post-2016 (Fraga Reference Fraga2018; Hajnal and Lee Reference Hajnal and Lee2011). The trends we observe among Latinos in recent years suggest weakening Democratic loyalty, stable but still relatively low identification as Republicans, and rapidly accelerating Independent identification.
We also find notable trends when looking at subgroup differences based on birth cohorts and national origin. Older generational cohorts like the Silents and Boomers tended to be more Democratic in the earliest time periods while the younger generations like GenX were more Republican and Independent overall, which continued through 2004 (Alvarez and Garcia Bedolla Reference Alvarez and Garcia Bedolla2003). During the Obama years, however, the older generations began shifting away from their previous Democratic identification towards relatively higher levels of Republican support, which continued through 2023. But while the older generations have become more Republican in the most recent years, younger generations have not become more Democratic; instead, they have increasingly identified as Independent. For national origin, while we find large subgroup differences at the beginning of our observed period—with Mexican Americans generally leaning Democratic, Puerto Ricans even moreso, and Cuban Americans instead identifying as Republicans—these are much less noticeable by the end of our observed time period. Instead, across national origin groups, U.S.-born Latinos remain more Democratic overall but are also increasingly likely to identify as Independent, indicating declining partisan attachments across subgroups.
There are some important limitations and caveats to our findings. Some datasets in our study, especially before the pathbreaking 2006 Latino National Survey, are smaller or have unrepresentative Latino samples. Fortunately, this is much less of an issue in 2006 and after, when political opinion surveys began including more Latino respondents (Fraga et al. Reference Fraga2006; Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024). Furthermore, our usage of multiple surveys and post-stratification weights based on Census demographics help correct for potential imbalances in any individual survey sample (Erikson, Mackuen, and Stimson Reference Erikson, Mackuen and Stimson2001; Lau Reference Lau2018; Leemann and Wasserfallen Reference Leemann and Wasserfallen2017). Due to these concerns about sample size, some of our generational and national origin figures should also be interpreted with more caution than figures with fewer subgroups. Lastly, while geographical effects remain an important factor for Latino political behavior, the region/state data in our datasets was too sparse and often weighted in a way that further biased partisanship estimates.Footnote 35 With these caveats in mind, we assert that our inclusion of multiple datasets from a wide range of providers and our post-stratification weights help provide the closest possible simulated panel of Latinos and their partisan identification across a long and politically relevant time period.
Conclusion
This study demonstrates that Latino partisanship remains a moving target, and further suggests that while Democratic identification among Latinos is slightly declining across age groups and Republican identification has remained stable, rates of Independent identification have steadily increased after 2016. This has deeply important implications for upcoming elections as the Latino population continues to grow and age into the electorate. Latinos are concentrated in many battleground states such as Arizona, Georgia, Pennsylvania, and North Carolina and are approaching becoming a majority of the overall population in other key states like California, Texas, and Florida (Krogstad et al. Reference Krogstad2023). But while their population has grown, broad assumptions about the future behavior of Latinos risk falling into the “demography is destiny” trap presuming either that they are inherently Democrats due to supporting liberal stances on race and immigration, or reluctant Republicans who pinch their noses while voting due to alignment on other issues like social conservatism and the economy (Corral and Leal Reference Corral and Leal2024; de la Garza and Cortina Reference de la Garza and Cortina2007; Odio and Stein Reference Odio and Stein2021; Wakefield Reference Wakefield2025). Our results for older Latinos indicate that this group has become more Republican since around the 2008 election, after which each party experienced significant polarization on race and immigration (Abrajano and Hajnal Reference Abrajano and Hajnal2015; Tesler Reference Tesler2016). While we cannot conclusively confirm any causal relationship given our data and approach, this temporal trend of older Latinos becoming more conservative around the Obama era potentially aligns with other research finding an increase in conservatism and Republican identification among certain segments of the Latino population because of, not in spite of, Republicans adopting conservative stances on race and immigration (Alamillo Reference Alamillo2019; Fraga, Velez, and West Reference Fraga, Velez and West2024; Hickel, Alamillo, et al. Reference Hickel and Alamillo2020; Hickel, Oskooii, and Collingwood Reference Hickel, Oskooii and Collingwood2024; Pérez, Robertson, and Vicuna Reference Pérez, Robertson and Vicuna2023). Future studies into Latino political behavior and partisanship should continue to examine the behavior of these more conservative Latinos, who remain a small but significant minority of the Latino electorate and have increasingly supported Republicans in recent elections (Cadava Reference Cadava2020; Corral and Leal Reference Corral and Leal2024; Jones-Correa, Al-Faham, and Cortez Reference Jones-Correa, Al-Faham and Cortez2018; Pérez and Cobian Reference Pérez and Cobian2024).
Meanwhile, our results show that younger generations of U.S.-born Latinos are increasingly distancing themselves from either party and instead identifying as Independent. This trend implies that despite being a growing population, Latinos may not have as much of an impact politically due to persistently low rates of partisan identification and political engagement among voting-eligible Latinos and particularly younger Latinos (Fraga Reference Fraga2018; García Bedolla and Michelson Reference García Bedolla and Michelson2012; Hajnal and Lee Reference Hajnal and Lee2011). Given that Latinos tend to receive less parental political socialization due to the number of Latinos who are immigrants, the decline in partisan engagement among young Latinos is even more troubling because peer and social networks oftentimes play an important role in bolstering their political efficacy (Carlos Reference Carlos2018; Carlos Reference Carlos2021; García Bedolla Reference García Bedolla2005; García-Castañon Reference García-Castañon2018; García-Castañon et al. Reference García-Castañon2019; Raychaudhuri Reference Raychaudhuri2020; Wong Reference Wong2006). If this group is becoming less partisan-identified overall, there may be significant negative downstream effects on Latino political engagement—such as lower rates of turnout and partisan registration among Latinos overall—that would only further exacerbate inequities in rates of Latino political participation and resources spent on Latino campaign outreach (Fraga Reference Fraga2018; García Bedolla and Michelson Reference García Bedolla and Michelson2012; Hersh Reference Hersh2015; Hajnal and Lee Reference Hajnal and Lee2011; Wakefield Reference Wakefield2025). Future studies must continue to address what strategies might be most effective at reaching the increasingly politically disengaged Latino electorate, especially among younger generations and those born in the United States.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/rep.2025.31
Data Availability
Data, replication code and supplementary appendix can be found at https://dataverse.harvard.edu/dataverse/latinopid89to23
Funding statement
This research did not entail additional costs. There are no additional funders or funding sources.
Competing interests
The authors do not have any competing interests to declare.
Open access
