Diversity Matters: The Election of Asian Americans to U.S. State and Federal Legislatures

Despite substantial research on descriptive representation for Blacks and Latinos, we know little about the electoral conditions under which Asian candidates win office. Leveraging a new dataset on Asian American legislators elected from 2011 to 2020, combined with pre-existing and newly conducted surveys, we develop and test hypotheses related to Asian American candidates’ ingroup support, and their crossover appeal to other racial and ethnic groups. The data show Asian Americans preferring candidates of their own ethnic origin and of other Asian ethnicities to non-Asian candidates, indicating strong ethnic and panethnic motives. Asian candidates have comparatively strong crossover appeal, winning at higher rates than Blacks or Latinos for any given percentage of the reference group. All else equal, Asian American candidates fare best in multiracial districts, so growing diversity should benefit their electoral prospects. This crossover appeal is not closely tied to motives related to relative group status or threat.


Q
uestions about descriptive representation have anchored normative debates in modern democratic theory and animated empirical studies of how racial and ethnic groups can translate political interests into political action.Do minority group members want to be represented by those who share their race or ethnicity? 1 Does having such candidates on the ballot make them more likely to engage in politics?Does achieving descriptive representation shape the legislative process and its outputs, and if so, how?
Research on these questions has mainly examined African Americans and, to a lesser degree, Latinos in the United States.Some have looked at Black and Asian representation in the United Kingdom (Nixon  1998; Saalfeld and Bischoff 2013; Sobolewska, McKee,  and Campbell 2018) and Muslims and members of Scheduled Castes and Scheduled Tribes in India (Ansari 2006; Jaffrelot 2003; McMillan 2005).On the other hand, we know relatively little about descriptive representation for Asian Americans. 2In this paper, we 1 Like Lien, Conway, and Wong (2003), we use "race and ethnicity" to refer to the mixture of minorities usually described as part of the "communities of color" in the United States with African Americans or Blacks often described in racial terms, whereas Latinos or Hispanics are contrastingly described in ethnic terms.We recognize race and ethnicity as socially constructed terms laden with meaning, often complicated, in flux, and sometimes imposed (Espiritu 1992; Lien,  Conway, and Wong 2003; Lien et al. 2021, 232; Okamoto 2014; Wong  et al. 2011, 243-5).
2 Asian Americans have been described as a racial or a (pan)ethnic group.In Lien et al. (2021, 232), Le describes violence committed against Asian Americans "by virtue of membership in an otherimposed racial group in a racial hierarchy over which we do not have control."In ethnic terms, "Asian American" identity is conceived as a panethnic identity spanning ethnic origin identities.Again, see Le's usage in Lien et al. (2021, 232), "While respecting individual ethnic groups that comprise Asian America, we also need to be conscientious about continuing to build the commonalities that serve as the basis for a panethnic, superordinate identity."This description of Asians or Asian Americans as "panethnic" is not new (Espiritu  1992).We utilize Asian American for the panethnic identity as outlined by these same scholars and others (e.g., Lien et al. 2021).In the context of Asians or Asian Americans, we follow Lien, Conway, and  Wong (2003) and use "ethnicity" or "ethnic origin" to refer to individual groups (e.g., Hmong and Japanese).As panethnic identity remains a type of ethnic identity, we may occasionally describe Asians or Asian Americans as an ethnic identity in an appropriate context.
Like past scholars, we acknowledge that the Asian American community is "a political construct" and "a fluid, malleable phenomenon that has many layers" (Lien, Conway, and Wong 2003, 468; Lien  et al. 2021, 232).As Espiritu (1992, 6-7) explains, Asian American panethnic identity started as an "imposed category that ignores subgroup boundaries," but racial violence led to Asian American organizations adopting "protective panethnicity."Similarly, Okamoto (2014, 4) argues that "Asian Americans adopted a panethnic label and organized to challenge inequalities and build new communities" and "not simply because the state had assigned them to a racial category." We use "Asian" and "Asian American" interchangeably but tend to prefer "Asian American" except when the use of "Asian" helps maintain parallelism with other groups (e.g., Blacks and Latinos).We also tend to utilize "Asian" in reference to origins in Asian countries and languages as well as current U.S. Census racial categories (e.g., "Asian Alone population") following current Census terminology.
ask where Asian American candidates win and why.We approach this question through the prism of two distinct phenomena: ingroup support and crossover support.The former refers to support for Asian American candidates from Asian American voters-either from within their own ethnic origin group or "panethnic" support from other ethnic groups in the Asian American community-whereas the latter to votes gained from non-Asian American voters.From these basic notions, we advance a number of theoretical propositions about where Asian Americans win at the aggregate (i.e., district) level, and about the psychological roots of support for Asian American candidates in the minds of individual voters.To test them, we bring to bear a new, large aggregate-level dataset of election outcomes matched with demographic data, pre-existing surveys on Asian American political attitudes, and a set of original survey experiments administered to large samples of Asian, white, Black, and Latino voters.
The following emerges in the aggregate: (1) Asian American candidates win with a lower proportion of Asian Americans in their district than other U.S. minority groups win with the same share of their group; (2) the more racially and ethnically diverse a constituency,3 the more Asian American candidates win; and (3) ethnic diversity within the Asian American community has no impact on electoral success by Asian American candidates.We find that crossover support for Asian American candidates comes from the absence of anti-Asian discrimination from white, Black, and Latino voters and, indeed, these groups' slight favorability for Asian over white candidates.With respect to ingroup voting, Asian American candidates benefit from loyalty tied to ethnic origin while simultaneously appealing in panethnic fashion to Asian American voters of other ethnic origins.
Taken together, these results consistently support theories suggesting that multiracial4 environments provide more coalition building opportunities for members of small minority groups like Asian Americans.At the same time, they negate competing ideas that Asian American candidates especially benefit from whites' identifying and supporting them as the "model minority."Moreover, while Asian Americans may identify more strongly by ethnic origin rather than with "Asian" or "Asian American" as a broader category (Lien et al. 2021; Lien, Conway,  and Wong 2003; 2004; López, Ruiz, and Patten 2017), the latter is sufficiently strong to serve as a basis of political support.More broadly, these findings provide a striking example of how this fast-growing group, and one that is highly ethnically diverse, can gain descriptive representation notwithstanding major challenges.

BACKGROUND AND EXPECTATIONS
A cornerstone of modern democratic theory is descriptive representation, or the notion that minority group interests are furthered with the election of group members to political office (Dovi 2002; Mansbridge 1999;  Pitkin 1967).The assumption is that peoples' interests are more likely to be articulated by representatives who share common experiences, history, and social networks.Racial redistricting in the United States is intended to make it possible for minority groups, especially Black and Latino voters, to elect their candidates of choice.This, in turn, has substantive payoffs: Black and Latino representatives are more likely to reflect the community's preferences in roll-call votes (Casellas  2011; Lublin 1999; Whitby 1997).By the same token, Black representatives are more likely to serve on committees, participate in oversight hearings, sponsor legislation addressing Black interests, and use liberal policy frames when speaking on these issues (Canon  1999; Gamble 2011; Haynie 2001; Minta 2011).
Restrictive immigration laws kept the Asian American population in check for almost two centuries.Since passage of the Immigration and Nationalities Act of 1965, however, their population has grown sharply.Today, more than 20 million people currently living in the United States trace their origin to a diverse set of Asian countries (López, Ruiz, and Patten 2017).The concept of an Asian American identity, as opposed to identities rooted in individual Asian ethnic origin groups, only emerged alongside the wave of immigration that began in the 1960s (Cho and Lad 2004;  Okamoto 2014).
Whether or not Asian Americans support or benefit from descriptive representation is less well known.Reviews of extant surveys argue that Asians' feeling and acting upon "linked fate" is highly contextdependent (Lien 2001; Lien, Conway, and Wong  2004).Another recent study argued Asian Americans appear to think about descriptive representation in much the same way as Latino respondents: favorable in principle, and increasingly as a function of strong ethnic identification, ethnic linked fate, and lower levels of "acculturation" (Schildkraut 2013).
Whether (and where) Asian Americans achieve descriptive representation, notwithstanding their preferences, is another matter.Due to the comparative paucity of Asian American elected officials, most of what we know comes from in-depth case studies of local and statewide races (Cho 2001; Collet 2008a;  2008b; Lai and Geron 2006; Ông and Mayer 2008;  Tam 1995; but see Cho 2002; Hardy-Fanta et al. 2016;  Lai et al. 2001; Min 2014; Sadhwani 2022). 5These studies engage, in a localized way, questions of whether Asians appear to be polarized with respect to whites, whether they vote as a bloc to support candidates of their own ethnic origin, and whether they support Asian American candidates regardless of their ethnicity as means of advancing the collective political interests of that larger category (Espiritu  1992; Okamoto 2014).Here, our goal is to draw an unprecedentedly wide-ranging picture of Asian American candidates' political success, under the aegis of two broad themes: Ingroup Support and Crossover Support.

Ingroup Support
A bedrock expectation is that Asian American candidates will find political success where Asian American voters are more numerous (Espiritu 1992; Hardy-Fanta  et al. 2006; Lai and Geron 2006; Lien 2015; Sadhwani  2022).Recent empirical research on statewide races in California sustains the idea that support from Asian American voters matters, but also that the relationship is highly contextual (Sadhwani 2022).One meaningful theoretical distinction we can draw is between groupcentric and non-group-centric motives.By "groupcentric," we mean that the central motive underpinning the willingness to vote for an Asian candidate is tied to perceived group interest and group hierarchy (Masuoka and Junn 2013).In the present example, Asian American voters want to vote for Asian American candidates because the former have an intrinsic preference for the latter.If so, we would expect Asian American candidates to do well in districts with proportionately more Asian American voters.At the individual level, we would expect to find a preference (among Asian Americans) for Asian American candidates, holding constant issues of policy and competence.
Ingroup support for Asian American candidates may be subdivided by affinity with ethnic origin group versus panethnic identification with Asian Americans as a broader social category (Collet 2008a; 2008b). 6We know quite a bit about the social and psychological precursors of identity choice among Asian Americans (e.g., Lien, Conway, and Wong 2003; 2004).Yet for all that, the link between these identities and voting is not well understood.Past studies find instances of ingroup support deriving from both sources in local and statewide elections, but we find relatively little guidance about which factor results in Asian American candidate victories more and when.Analyzing survey data, Min  (2014) observes that high levels of panethnic solidarity depress Asian American voter registration and turnout yet also stimulate higher rates of participation in political activities other than voting; on the other hand, Wong et al. (2011) find minimal differences in the rates that panethnic and ethnic origin identifiers vote and participate in forms of political activism besides voting.Finally, Sadhwani (2020) finds that Asian American candidates attract higher panethnic turnout, but that the effect on co-ethnics varies.
For our purposes, we might expect Asian American candidates' ingroup support to be rooted primarily among voters of their own ethnic origin.Surveys show that Asian Americans overwhelmingly identify with their ethnic origin rather than as Asian American (Lien, Conway, and Wong 2003; López, Ruiz, and  Patten 2017).Furthermore, a relatively homogenous ethnic origin group should provide more cohesion, as well as a tighter network for organizational and financial support.Taken together, these considerations drive the ethnic-origin hypothesis (H 1 -EOH).At the aggregate level, it implies that Asian American candidates will do better in districts where the Asian American community is largely made up of a single ethnic group.At the individual level, H 1 -EOH suggests that Asian American voters will, when offered the choice, prefer common-origin Asian American candidates over those of other ethnic origins.
On the other hand, panethnic Asian American identity may drive candidate support, at least in the contemporary era, for a few reasons: first, the Asian American electorate (vs. the population) skews younger and second-or third-generation immigrant, and as such are more likely to opt for a panethnic Asian American identity than their ethnic origin group (Pew  Research Center 2013; Sadhwani 2022; Tam 1995;  Wong et al. 2011).Second, Asian American community financial networks are national rather than local (Cho  2003; Cho and Lad 2004), so candidate fundraising is not necessarily driven by co-ethnic supporters within the electoral constituency.Finally, ethnic (and panethnic) voting is always a matter of perspective (Okamoto  2014).While Asian American voters might prefer someone of their own ethnic origin to another Asian American ethnic group head-to-head, they may still favor the latter over a non-Asian American candidate.For Asian American voters, this is especially likely to matter given their relatively small numbers and very high ethnic diversity.Taken together, these considerations underpin the panethnic hypothesis (H 2 -PEH).At the aggregate level, the H 2 -PEH hypothesis would be a better fit than the H 1 -EOH if Asian American candidates did just as well as (or better) in districts where the Asian American population was diverse rather than monolithic.At the individual level, H 2 -PEH suggestsagain, contra H 1 -EOH-that support for candidates of other Asian origins should be just as strong as it is for candidates of one's own Asian origin, and that support for both should be higher than support for candidates of other minority groups.
It may be that Asian American ingroup preferences are not especially group centric in the way defined here, if (for example) their motives are not intrinsically tied to conceptions of group identity and interest, but rather reflect informational assumptions about competence or the kinds of policies candidates of various ethnic groups are likely to support.The non-group centric hypothesis (H 3 -NGC) has two key individual-level implications: First, we would find comparable levels of ethnic and panethnic support, and neither especially stronger than support for white candidates.Second, we would find that, when offered the choice between purely descriptive ethnic representation and substantive representation, the latter wins out.Of course, these do not have to be mutually exclusive.

Crossover Support
Because Asian American voters almost never form a majority, that they succeed politically at all suggests that they must generate crossover appeal, that is, electoral support from other minority groups (e.g., Espiritu  1992; Hardy-Fanta et al. 2006; Lien 2015; Sriram and  Grindlife 2017; Takeda 2001).A useful and natural benchmark in assessing this is how likely an Asian American official is to hold a district for a given Asian share of the population relative to the share held by members of other minority groups.In other words, do Asian Americans succeed disproportionately-versus America's other major racial and ethnic groups-in generating crossover support?There are reasons to think that they might.In general, the literature has emphasized Asian American candidates' crossover appeal to non-Asian voters in winning both mostly white and more multiracial districts (Lien et al. 2007).Some have described the crossover appeal of Asian American candidates as stemming from their ability to "toggle" between ethnic and panethnic appeals (Collet  2008a), or, on the other hand, to adopt a "deracialized" strategy (Franklin 2010; Gillespie 2010; 2012; King-Meadows 2010; Lewis 2010; Sriram and Grindlife  2017; Yon 2010).
Yet we have little systematic knowledge of where and how crossover helps Asian American candidates win at the district level, or the psychological foundations of their crossover appeal.We return, as before, to the theoretical distinction between group-centric and non-group-centric motives.For a white voter, a groupcentric crossover vote for an Asian candidate would occur if Asians are perceived as less ethnically "threatening" to whites than candidates of other minority groups.This is the model minority hypothesis (H 4 -MMH) (see, e.g., Visalvanich 2017a; 2017b). 7mpirically, at the aggregate level, we would expect, if this hypothesis is valid, that Asian American candidates should do better in districts with white majorities.At the individual level, evidence for H 4 -MMH would consist of white voters preferring Asian candidates to Black or Hispanic candidates, all else equal.
Other ethnic or racial minority voters might also have group-centric reasons to crossover to Asian American candidates.8It is plausible, for instance, that the crossover appeal of Asian American candidates to Black and Latino voters lies in shared experience as minority group members in America's racial hierarchy (Hardy-Fanta et al. 2016; Masuoka and Junn 2013;  Phillips 2021).This is the shared minority status hypothesis (H 5 -SMSH).Another possibility is that group members would prefer one of their own, but rate Asian Americans more highly than they rate all other competing groups.This is the next best choice hypothesis (H 6 -NBCH).Both H 5 -SMSH and H 6 -NBCH suggest that-contra H 4 -MMH-at the district level, Asian American candidates will do better in multiracial districts than in mostly white ones.If they have similar aggregate implications, however, the two minority crossover hypotheses have different implications at the individual level: if other minority voters pick Asian American candidates at comparable rates to themselves, but prefer any minority candidate to white candidates, this would be consistent with H 5 -SMSH.On the other hand, if other minority voters rate Asian American candidates less highly than themselves, but more highly than all other groups, this would support H 6 -NBCH.
The previous hypotheses are all premised on crossover support for Asian American candidates as a function of people voting their group interests.But crossover appeal may lie less in notions of racial or ethnic competition and threat, and more in judgments of competence or substantive similarity on issue preferences.That is, group preferences are not intrinsically tied to conceptions of group identity and interest, but rather reflect informational assumptions about competence or the kinds of policies candidates of various ethnic groups are likely to support.Indeed, the variant of H 4 -MMH presented by Visalvanich (2017a; 2017b) is primarily of this character rather than racial per se: in short, whites' preferences are tied to their stereotypes about Asian competence as held against negative stereotypes of members of other minority groups.This echoes a well-established understanding in the literature that Asian American political preferences are more or less in line with whites on many key issues (Cho and Cain 2001; Uhlaner, Cain, and Kiewet 1989).This lack of polarization with whites implies that, overall, Asian American candidates should succeed at winning office disproportionately versus candidates of other minority groups.This argument also, in principle, applies to preferences of other minority voters if they view Asian Americans as (on average) holding similar ideological and political interests.Together, we label these motivations Non-Group-Centric Crossover (H 7 -NGCC).Empirically, this pattern would be consistent with a finding that, in the aggregate level (i.e., district level), Asian American candidates do better in districts with white majorities (versus multiracial ones).At the individual level, however, we would find-contra the group-centric hypotheses raised above-the absence of strong group differences, once we have controlled for the informational roots of whites' putative preference for Asian American candidates-assessments of competence and issue preferences.

DATA AND APPROACH
Our empirical approach triangulates based on comprehensive, aggregate-level data and national opinion surveys that extensively overrepresent minority groups.For the former, we identified Asian Americans elected in regular state legislative elections held for both chambers between 2011 and 2020 for 33 states: 21 with the highest Asian American populations along with 12 more with high Black or Latino populations to facilitate comparisons between groups. 9Data were similarly collected on Asian Americans elected to the two federal houses from 2012 to 2020.We also gathered information on the ethnic origin of Asian American officials.These data were obtained from a variety of sources, including the National Asian Pacific American Political Almanac (Nakanishi and Lai N.d.), the Asian Pacific American Institute for Congressional Studies, various Asian American legislative caucuses, and examination of photos and biographies on legislative websites. 10e matched data on legislator race and ethnicity with district demographic data from the American Community Survey (ACS) conducted by the U.S. Census Bureau.The ACS is preferable to the decennial census as this annual survey better tracks changes in the quickly evolving Asian American population.While most states held elections under the same district boundaries across the decade, some changed their boundaries mid-decade.Generally, the ACS provided data for both old and new boundaries, but observations were dropped when data were unavailable. 11or individual-level evidence, we turn to two public opinion surveys.The first is the 2016 pre-election round of the National Asian American Survey, or NAAS (Ramakrishnan et al. 2016), which interviewed a large, nationally representative sample of Asian American adults (n = 2,787), as well as sizable samples of white (n = 501), Black (n = 520), and Hispanic respondents (n = 514).The second is an online survey we conducted in September 2021 targeting a nationally representative sample of Asian American adults from the six largest ethnic origin groups in the United States (Chinese, Indian, Filipino, Japanese, Korean, and Vietnamese, total n = 1,000), as well as comparison national samples of whites (n = 502), Blacks (n = 501), and Latinos (n = 500). 12hese two sources of evidence complement each other.The aggregate data provide a thorough picture of where Asian American candidates win seats, and how this relates back to district demography.The limitations are that it is: (a) purely observational in nature and, thus, limits our ability to make strong causal inferences; (b) relatively ill-suited to getting at what are, fundamentally, psychological motives without making strong ecological assumptions tying these to district-level characteristics; and (c) by necessity focused on election winners versus election candidates or campaigns, meaning that we are unable to see those who prevailed ran against, and what kind of campaign it took to get them over the finish line.By contrast, the survey evidence sacrifices the "big picture," but it addresses the weaknesses of the aggregate data by: (a) implementing experimental strategies that allow for stronger causal identification; (b) focusing explicitly on individual-level analysis of motives; and (c) focusing, as we describe further below, on respondents' evaluations of prospective candidates rather than simply observing election winners, and thus not being vulnerable to the same censoring problems of the aggregate data.

EXPLORING THE PREVALENCE OF ASIAN AMERICAN LEGISLATORS
Figure 1 shows the total number of Asian American state and federal legislators broken down by party from 9 The data thus include all 11 southern states, the 10 non-southern states with the highest Black population, and the 12 states with the highest Hispanic population.The aggregate dataset can be found at the APSR Dataverse (Lublin and Wright 2023). 10Legislators with backgrounds in more than one minority group are categorized as members of both groups.Excluding these legislators does not alter the results.Legislators with origins in two Asian countries are also categorized as members of both ethnic groups.Following academic and U.S. Census convention, Asian American excludes people with origins in the Middle East.Though Pacific Americans, including Native Hawaiians, were often lumped with Asian Americans in the past, they are not classed as Asian American.
11 Pennsylvania adopted entirely new state legislative maps after the 2012 elections.Only a few of Alaska's districts remained unchanged after the 2012 elections.Some Georgia Senate and Texas House districts also changed after the 2012 elections, though most remained the same.Judicial orders altered a minority of Virginia House districts before the 2019 elections.North Carolina redrew before the 2020 election.Mid-decade changes also occurred to congressional districts in Florida, North Carolina, Pennsylvania, and Virginia.2011 to 2020. 13The number of state representatives increased from 70, or 2.2%, in 2012 to 114, or 3.2%, in 2020.Similarly, the share of state senators rose from 2.2% to 3.4% (29-45).Asian Americans, however, comprised 7.4% of the Asian population in these 33 states. 14he only Asian majority state, Hawaii, accounts for a disproportionate, albeit declining, share of Asian American state legislators.Over the decade, the Hawaiian share of Asian American legislators declined from 54% to 32% between 2012 and 2020.
Outside Hawaii, there is a great deal of variation in Asian American success across states.California, the state with the second highest Asian population at 17.1%, now has the most Asian American state legislators-14 or 11.7%.But six states-Arkansas, Delaware, Louisiana, Mississippi, Rhode Island, and South Carolina-elected no Asian American state legislators over the entire decade.Meanwhile, 11 states where Asians comprise under 5% of the population elected at least one Asian American state legislator. 15Just 13 Supplementary Table 2 shows the data for each state.New Jersey and Virginia hold state legislative elections in odd-numbered years.All Alabama and Maryland state legislators serve 4-year terms.State senators in many states serve staggered terms, but the table shows the total after each election but not taking into account special elections or appointments.A combination of two 4-year terms and one 2-year term is sometimes used to assure that all senate terms are completed within the 10-year apportionment cycle. 14This figure from the 2019 ACS includes Asian Americans who identified with multiple races. 152019 ACS Asian Alone or in Combination Population.The 11 states are Arizona, Colorado, Florida, Georgia, Michigan, Missouri, New Mexico, North Carolina, Ohio, Pennsylvania, and Tennessee.Colorado, Ohio, and New Mexico had no American Asian legislators after 2020.
4.9% Asian American, Georgia elected two senators and four representatives in 2020.
Asian American state legislators win overwhelmingly (almost 90% of the time) as Democrats.Eleven of the 24 states that elected at least one Asian American legislator-Alaska, Illinois, Maryland, Michigan, Minnesota, New Mexico, New York, North Carolina, Oregon, Pennsylvania, and Washington-did not elect any Asian American Republicans.Some states are exceptions to this pattern.The single Asian American legislators who won in Colorado, Florida, and Tennessee were all Republicans.In heavily Democratic California, Republicans leapt from 0% to 29% of Asian American state legislators over the decade.
Asian American representation in the U.S. House rose from 2.0% to 3.4% (9-15 members).The number of Asian American U.S. Senators tripled from one to three in 2016 but declined to two after Kamala Harris became Vice President.Until two Republican representatives won in 2020, all had been Democrats. 16he extent to which Asian Americans have achieved parity in descriptive representation-that is, a parity ratio (seats over population) of 1-depends upon the population measure.We rely on the proportion Asian alone or in combination with another race in the population. 17igure 2 shows parity ratios in the wake of the 2020 elections. 18waii is the only state in which Asian Americans were continuously overrepresented, holding an average of 21% more seats than expected based on population during the 2010s.This high level of electoral success partly reflects that Asian Americans, especially Japanese Americans, are disproportionately active within the state's Democratic Party, which wins an overwhelming share of elections (Coffman 2003).Moreover, the winner-take-all electoral system usually advantages majority groups like Asian Americans in Hawaii.
Elsewhere, the picture was more mixed.Asian Americans did not achieve parity based on population in both houses in any other state.Only the Connecticut, North Carolina, and Ohio Senates reached parity in 2020.Ten state houses (Arizona, California, Illinois, Maryland, Massachusetts, Michigan, Minnesota, Nevada, Oregon, and Washington) and five state senates (Alaska, Georgia, Maryland, Michigan, and Washington) were above one-half of parity at the decade's end.
In the U.S. Congress, Asian Americans have narrowed the gap but still fall well short of parity.Their share there was just 35% of the total population in 2012 and 51% by 2020.At 17% of parity, the share of Asian American senators was even lower than that expected based on total population after the 2012 elections.That rose to 44% in 2018 before Kamala Harris left the Senate to become Vice President.
Figure 3 presents the ethnic origin of Asian American state legislators over the 2010s. 19Eight legislators with origins in more than one ethnic group are counted    twice.Though composing 6.4% of Asians, Japanese Americans led the state legislative numbers throughout the decade due to their strength in Hawaiian politics.By 2020, however, their numbers dropped overall by two among state legislators and from five to three among federal legislators.Always heavily Democratic, only one Republican Japanese American state legislator remained after the 2018 elections.Chinese Americans, who comprise the largest share of Asians at 23.2%, consistently had the second highest numbers, and do (relatively) well in California, Hawaii, Illinois, and Massachusetts.Unlike Japanese Americans, more serve as Republicans and the GOP share remained stable at 19% as the number of Chinese American legislators grew from 26 to 37 over the decade.The number of federal legislators rose from 3 to 4, including the election of Illinois Sen. Tammy Duckworth in 2016.
Indians now form 18.8% of Asians and the number of Indian state legislators more than quadrupled from 6 to 27 during the last decade.They are not concentrated in any state, with three apiece serving in Michigan, North Carolina, and Washington and four in New York after the 2020 elections.As their numbers shot up, the Republican share fell from 33% to 11%.Four Indian Americans now serve in the U.S. House.Sen. Kamala Harris was the first Indian American ever to win election to the U.S. Senate in 2016.
Other ethnic origin groups had lower levels of representation.Filipino Americans are 18.1% of the Asian population.During the 2010s, the number of Filipino American state legislators rose from 14 to 18 with the number of Republicans holding steady at 2. Like Japanese American legislators, most win in Hawaii.After the 2020 elections, there are no U.S. senators but one U.S. House member with Filipino origins, a decline of one from 2018.
Only 8.2% of the Asian population, the number of Korean American state legislators increased from 11 to 16 over the decade.Two states, Hawaii and Maryland, elected more than one Korean state representative.Vietnamese Americans, who form 9.4% of the Asian population, saw a steep increase in state legislators, rising from 1 to 10. Around one-fifth of Koreans and Vietnamese sit as Republicans, but the numbers are small.In 2016, Florida Rep. Stephanie Murphy was the second Vietnamese American, and first Democrat, to win election to the U.S. House.Similarly, in 2018, New Jersey Rep. Andy Kim became the second Korean American, and first Democrat, to serve in that body.No Korean or Vietnamese Americans have ever won a U.S. Senate seat.Hmong, who form 1.4% of the Asian population, were the only other ethnic group to elect three state legislators simultaneously.All are Minnesota Democrats with five serving currently in the House and one in the Senate. 20ian Americans overwhelmingly win as Democrats with the number of Democratic state legislators increasing from 87 to 142 between 2011 and 2020.Meanwhile, the number of Republicans rose from 12 to 17 and their share of all legislators fell from 12.1% to 10.7%.All the Asian American federal legislators during this period were Democrats until two Republicans, Reps.Young Kim and Michelle Steel, won election in 2020.

Aggregate-Level Evidence
Figure 4 shows the share of Asian Americans elected to state legislatures after the 2020 elections by the share of Asian alone and combination population for Hawaii and the 32 states in the dataset besides Hawaii. 21In support of the bedrock idea that there is strength in numbers, Asian American legislators are more likely to hold districts as the Asian population share increases.The sole exception-the dip in 50%-60% range for the 32 states-reflects the small number of cases where one additional Asian American candidate success would result in the data fitting the general pattern.Asian American candidates win below their share of the population in state legislative districts roughly until reaching the 30%-40% range for the Asian alone or in combination population. 22o matter how the share of Asian Americans is measured, 23 outside of Hawaii, Asians form a majority in no more than six state legislative and one congressional district.Unlike for African Americans and Latinos (Lublin 1997; Lublin et al. 2020), the intentional creation of Asian majority districts does not account for the bulk of officials elected to state legislatures or the U.S. House.The vast majority of Asian American legislators elected outside of Hawaii represent districts where Asian Americans, let alone their ethnic origin group, do not represent either a majority or even a near majority.Lower rates not just of citizenship but also turnout among Asian Americans further suggest that the data overstate the share of Asian American voters.
We turn now to multivariate analysis of this aggregate-level dataset.The unit of analysis is state legislative seats in 33 states, from 2011 to 2020.The dependent variable is whether a seat is held by an Asian 20 There are now two Cambodian and two Pakistani state representatives.One Indonesian and one Nepali serve in state houses, as do one (non-Hmong) Laotian and one Bangladeshi in state senates. 21The results are similar regardless of how one measures the share of Asian Americans and also for the U.S. House, as shown in Supplementary Table 6 along with the number of cases.The dataset includes Asian American legislators from multiracial backgrounds, but Asian alone measures exclude them and underestimates the comparable Asian population.The problem is much less severe outside of Hawaii, which has by far the highest share of multiracial population and legislators.Asian alone or in combination VAP and CVAP were unavailable from the ACS. 22The same is true for the 10%-20% range for Asian Alone CVAP, and the 20%-30% range for the Asian Alone population or VAP. 23Besides Asian alone or in combination population, we looked at Asian alone population, Asian alone VAP, and Asian alone CVAP.
American legislator and the main predictors are district-level demographics.Our estimator is multilevel logistic regression with random effects by district to address error clustering.We estimated all specifications with and without year fixed effects to account for the possibility that time trends may confound any results.Since all results are consistent regardless, we present the version without year fixed effects.
We use four different measures of the share of Asians Americans-Asians alone or in combination population, Asian alone population, Asian alone voting-age population (VAP), and Asian alone citizen voting-age population (CVAP)-and examine Hawaii separately.The bottom row displays the estimated proportion of Asians required for a district to have a 50-50 probability of having an Asian American legislator.Figure 5 shows the predicted probability of an Asian American legislator in a district based on the proportion of Asians. 24Regardless of measure, the positive relationship is unambiguous.
Having established the fact of Asian Americans' elevated ingroup support, at least in the aggregate, the nature of that support remains to be seen.Does it flow mainly from voters of a given Asian American official's own ethnic origin, or are Asian American officials equally adept at drawing panethnic support from other Asian origin groups?In the aggregate, the question is how diversity within a district's Asian American community effects the probability of an Asian American candidate winning.Model 1 in Table 2 tests the impact of this by including a measure of diversity within the Asian American communityin this case a Herfindahl index ranging from 0 (maximally homogenous with respect to ethnic origin diversity) to 1 (maximally diverse)-and an interaction of this measure with the share of Asian Americans (since any negative effect of intragroup diversity should increase as their numbers grow). 25Asian American community diversity has no effect, all else equal, and, if anything, the probability of an Asian American legislator slightly increases with intragroup diversity, which flatly contradicts theories that suggest ethnic origin diversity among Asian Americans hinders their election. 26ack of common identity and even antagonistic relationships between countries of origin have apparently not prevented the construction of a panethnic Asian American identity (Espiritu 1992; Okamoto 2014).These results may also reflect the disproportionate weight of young and non-first-generation Asian Americans among Asian American voters who are more

FIGURE 4. Percent Asian American State Legislators by Asian Alone or in Combination Population
Note: See the text for information on the 32 included states.Source: Data on legislator race gathered by authors.Data on Asian alone and combination population from the American Community Survey. 24The curve for the Asians alone population is not shown because it overlaps so closely with Asians Alone VAP in the 32 states graph.For Hawaii, we present the results only for the Asians alone or in combination measure due to the very high share of the population that identifies as Asian as well as with another racial group. 25Like the models in Table 1, the estimator for the models in Table 2 is multilevel logistic regression with random effects by district to address error clustering.Specifications with and without year fixed effects produce extremely similar results with the latter presented here.The racial and ethnic diversity Herfindahl index equals 1 − Σ(x i / population) 2 with x i being the census ethnic origin categories.We have, however, combined Chinese and Taiwanese as well as Japanese and Okinawans. 26The curve is very close to a flat line.This remains unchanged if one groups Asians into regional categories, such as East Asian, South Asian, and Southeast Asian.A model of only open seats also shows an insignificant relationship between the intragroup diversity and the probability of an Asian American legislator.Note: VAP is the voting-age population.CVAP is the citizen voting-age population.Each curve shows the estimated probability of an Asian American state legislator based on the models in Table 1.Asian American winner is the dependent variable in all models with the only independent variables being different measures of the share of Asians.The estimator is multilevel logistic regression with random effects by district to address error clustering.The estimates based on the models including Asian Alone VAP and total population as independent variables are virtually identical, so only the former is shown here.Note: VAP is the voting-age population.The dependent variable is Asian American winner with Asian Americans coded as 1 and other as 0.
Asian American winners include legislators with multiple racial origins, but their exclusion does not alter the substantive results.The intergroup diversity Herfindahl index equals 1 − Σ(x i /population) 2 with x i being the standard census racial categories with Hispanics excluded and treated as a separate category.The non-Hispanic multiracial population is also treated as a distinct category.The intragroup diversity Herfindahl index equals 1 − Σ(x i /population) 2 with x i being the census ethnic origin categories.We have, however, combined Chinese and Taiwanese as well as Japanese and Okinawans.The model includes both state house and state senate elections for the 32 states outlined in the text.The estimator is multilevel logistic regression with random effects by district to address error clustering.We estimated all specifications with and without year fixed effects to account for the possibility that time trends may confound any results.Since all results are consistent regardless, we present the version without year fixed effects.The number of cases is lower for the intragroup diversity model due to missing data on ethnic origin for cases where racial data are available.Source: Data on the percentage of Asians from the American Community Survey.Data on legislator race gathered by author.See the text for more detail.
likely to identify as Asian Americans (Tam 1995; Wong  et al. 2011).They also reflect "points of convergence on a key set of public policies" across Asian American ethnic origin groups (Wong and Shah 2021, 75).

Individual-Level Evidence
Do these patterns hold up when we examine the preferences of Asian American voters? Figure 6 shows responses to a battery of questions fielded in both the 2016 NAAS and our 2021 survey, prompting respondents as follows: "Thinking about the next Congress and who should be elected.Please tell me how important the following statements are."Respondents were then asked, in sequence, about their own racial/ethnic group, women, their own party identification, and "people in Congress who represent the issues I care about most."As an added twist, Asian American respondents in both surveys were prompted, at random, to consider either their own ethnic origin group or "Asian Americans," allowing for a clean test of relative support for the two categories.What emerges is striking consistency across surveys, all the more remarkable given that they were fielded 5 years apart using very different modes of sampling and administration.Three results stand out in both cases: (1) across all minority groups, there is fairly strong support for more descriptive representation in Congress, with Asian respondents comparable to those who are Latino but less supportive than those who are Black; (2) all groups tend to favor substantive representation (i.e., partisan and issues) over descriptive representation; and (3) across categories of ingroup support, Asian American respondents do not appear to discriminate strongly between their own ethnic origin and "Asian American" representation, with no difference in the NAAS and a slight preference for the latter in our survey.
This last point is relatively strong evidence for our panethnic hypothesis (H 2 -PEH), and fully consistent with the aggregate-level findings shown earlier: Asian American candidates find as much support from Asian American voters outside their own origin group as from inside.To test this further, we turn to a vignette experiment embedded in our 2021 online survey asking respondents to evaluate hypothetical U.S. House candidates.We vary the candidate's race/ ethnic origin according to nine possible combinations -the six Asian origin groups for whom we have respondents, plus white, Black, and Latino-using a combination of surname, parent's country of birth,

FIGURE 6. Preferences on Different Kinds of Representations in Congress by Group
Note: Points represent means with 95% confidence intervals.Each x-axis value is a different item in the representation battery asked identically in both surveys.Possible scores range from 0 = not at all important to 1 = extremely important.Note that the "Paneth Asian" and "Asian Eth Origin" items were asked only of Asian respondents; these respondents saw either "Asian Americans" or their own ethnic origin, respectively, assigned at random.The figure was produced from the following tables, posted as individual files in the Harvard Dataverse associated with this article: "2021 Online Survey Aggregates for Figure 6.dta" and "NAAS 2016 Aggregates for Figure 6.dta."Each table was generated from 2021 Online Internet Survey and 2016 NAAS (respectively) per detailed instructions provided in the posted do-file "Survey Analysis Code.do."Source: 2016 NAAS and 2021 Online Internet Survey.and origin ethnicity.We also vary, in tandem with this information, the candidate's panethnic identity ("Asian American," "Anglo American," "African American," and "Hispanic American").27Finally, we varied the candidate's job history, toggling either a community organizer interested in protecting workers' rights or a small business owner devoted to reducing regulations to create jobs.Political experience is held constant as all candidates have "spent the last 3 years on his local city council."The basic vignette is as follows: This is Michael [surname], who is preparing to run for a seat in the U.S. Congress.Born in the U.S. to parents who immigrated from [ethnic origin], he is currently 37 years old, and married with two children.He is college-educated, and [work background].While proud of his [ethnic origin] roots and his [panethnic] identity, if elected he pledges to work hard on behalf of all those he represents.
Respondents are then asked three agree-disagree questions, which we combined into an additive favorability index based on their strong overall commonality (Cronbach's alpha = 0.89): • "Michael [surname] is the kind of candidate that understands the political issues that matter to me." • "If he won, Michael [surname] would represent my district well in Congress."• "I would like to see more people like Michael [surname] in Congress." Our interest is specifically in how our Asian American respondents evaluate candidates of different ethnic origins, plotted in Figure 7.The most relevant comparisons are among the three left-most bars, which reference candidates of the respondent's own ethnic origin group ("R's Eth Origin"), another of the five Asian origin groups ("Other Asian Origin"), and the average of all non-Asian groups ("Non-Asian W-B-L").Bars to the right indicate mean scores for Asian respondents applied to white, Black, and Latino candidates separately.Altogether, the results best fit the panethnic hypothesis: Asian Americans tend to favor Asian American candidates over others all else equal, contra the non-group-centric hypothesis (H 3 -NGC).And, while there is a slight preference for their own ethnic origin, the difference is not statistically significant as the ethnic origin hypothesis (H 1 -EOH) would FIGURE 7. Asian Voter Candidate Preference by Race/Ethnicity of Candidate Note: This figure maps support for vignette candidate (and additive index of three favorability questions asked), with scores ranging from 0 = completely unfavorable to 3 = completely favorable.Only Asian respondents are included."W," "B," and "L" stand for "white," "Black," and "Latino," respectively.Spikes represent 95% confidence intervals.The figure was produced from the following table, posted in the Harvard Dataverse associated with this article: "2021 Online Survey Aggregates for Figure 7.dta."This table was generated from the 2021 Online Internet Survey per detailed instructions provided in the posted do-file "Survey Analysis Code.do."Source: 2021 Online Survey.
have anticipated.In short, Asian Americans do seem to have group-centric political preferences and they are willing to support candidates of other Asian ethnic origins almost as much as candidates of their own.

Aggregate-Level Evidence
How do Asian American candidates fare, versus candidates of other minority groups, for a given underlying prevalence of that group in the district electorate?In contrast to the many studies indicating that Black and Latino candidates require districts with at least majority of their group (or close to it) to win election in most areas (e.g., Davidson and Grofman 1994; Lublin  et al. 2020; Parker 1990; Whitby 1997), Asian American candidates become more likely than not to emerge victorious at lower levels.The odds shift in favor of an Asian American legislator once Asians alone or in combination form more than 41% of the total population.The same is true when Asians alone surpass 33% of CVAP, a share that rises to 39% for the VAP and the total population.For Hawaii, the required proportion of Asian alone or in combination population for Asian American legislators to have an equal shot of holding the district is 40%, close to that outside the islands.
Looking at Asian CVAP, the measure that most closely mirrors the Asian American share of the electorate outside Hawaii, models indicate that Asian candidates should win one in seven contests in 20% Asian districts-odds that increase to one in four at 25% Asian.The chances of an Asian American candidate winning are lower for broader measures of the Asian population that include many noncitizens, 1 in 10 for 20%, and 1 in 6 for 25% Asian alone VAP or total population districts.The probabilities decline to 1 in 13 at 20% and 1 in 7 at 25% based on the model using the population of Asians alone or in combination, the most expansive measure of the Asian population.
These figures matter because, outside Hawaii, there are far more districts between 20% and 30% Asian than greater than 30% Asian-2.5 times as many based on the broadest measure of Asians alone or in combination. 28As a small, dispersed minority, Asian American candidates need to win in contexts where they are far from the majority to gain representation.Asian Americans can only very rarely leverage the Voting Rights Act (VRA) to create more electoral opportunities due to the inability to create Asian majority districts.
Similar models of the election of Black and Latino legislators confirm that Asian legislators succeed disproportionately for any percentage of Asians compared to equivalent percentages of Blacks or Latinos. 29igure 8 shows the predicted relationship between the proportion of same-group CVAP and the probability of an Asian, Black, or Latino state legislator.The probability of an Asian legislator is higher at any given proportion than for Black or Latino legislators.Due to much higher citizenship rates in the Black than Asian or Hispanic population, the probability of a Black legislator trails that of an Asian legislator by much less and roughly matches that for a Latino legislator for any proportion of same-group VAP (not shown here). 30 small group that is diverse within its own ranks, Asian American politicians must swim in political pools that vary considerably in racial and ethnic composition.Here, we leverage our aggregate-level data onto our specific hypotheses about crossover support for Asian American candidates.The core question is: do Asian American candidates get crossover support primarily from whites, or from members of other minority groups? Figure 9 shows the share of Asian Americans elected outside of Hawaii at different levels of intergroup diversity as measured by a Herfindahl index that ranges from 0 to 1 with higher numbers indicating greater diversity.31 Despite occasional dips in the proportion of Asian American officials as intergroup diversity rises, Asian Americans clearly hold a much higher share of state legislative and congressional districts at high levels of intergroup diversity.
The low rate of Asian American candidate success in the 0-0.1 category is not especially surprising.Outside Hawaii, there are no heavily Asian American districts, so these are homogenous districts populated by another racial or ethnic group.The failure of Asian American candidates to win a greater share of seats in the 0.1-0.4range is more interesting.If Asian Americans fared best in districts that contained only one other racial or ethnic group, such as whites, we would expect to see more Asian American winners in this range.Instead, Asian American candidates clearly experience their highest rates of success in multiracial districts with intergroup diversity scores above 0.6.
The descriptive data provide preliminary evidence against the model minority hypothesis (H 4 -MMH), which 28 Based on the 2019 ACS 5-year estimates.The ratio is 2.3 for the Asians alone population and 2.7 for both the Asian VAP and CVAP. 29See model estimates in Supplementary Table 9.The models of the election of Black legislators' control for the share of Hispanics as well as Blacks as past studies demonstrate that omitting it can lead to inaccurate estimates (Grofman and Handley 1989; Lublin 1999).The share of Hispanics is set to zero in Figure 10 to show the impact of changes in solely the Black population. 30The probability of a Black elected official holding a district is lower in the South than the non-South for districts less than 50% Black VAP or CVAP; see Supplementary Figure 1 based on models shown in Supplementary Table 9. Determination of the percent needed to provide a realistic opportunity to elect under the VRA requires casespecific analysis. 31The racial and ethnic diversity Herfindahl index equals 1 − Σ(x i / population) 2 with x i being the standard census racial categories with Hispanics excluded and treated as a separate category.The small non-Hispanic multiracial population is also treated as a distinct category.Across state legislative districts, intergroup diversity ranges from 0.04 to 0.82 with a mean of 0.43 and a standard deviation of 0.17.See Supplementary Table 6 for the number of cases.
suggests that Asian American candidates should do better in white majority districts, and in favor of both the shared minority status (H 5 -SMSH) and next best choice (H 6 -NBCH) hypotheses, in that Asian American candidates do better in multiracial districts than in mostly white ones.Assessing the solidity of any relationship between FIGURE 9. Percent Asian American Legislators in 2020 by District Diversity, Excluding Hawaii Note: All states but Hawaii included for the U.S. House.See the text for information on the 32 included states for state legislators.The intergroup diversity index equals 1 − Σ(x i /population) 2 with x i being the standard census racial categories with Hispanics excluded and treated as a separate category.The small non-Hispanic multiracial population is also treated as a distinct category.Source: Data on legislator race gathered by authors.Data on race and ethnicity from the American Community Survey.FIGURE 8.Estimated Probability of Asian, Black, or Latino Legislator by Group CVAP Note: CVAP is the citizen voting-age population.Each curve shows the estimated probability of either an Asian, Black, or Latino state legislator based on the appropriate models in Table 1 and Supplementary Table 9.In these models, the dependent variables are Asian, Black, or Latino winner with the corresponding CVAP for that group included as an independent variable.The model of Black winner also includes Hispanic CVAP.The estimator is multilevel logistic regression with random effects by district to address error clustering.Results are based on data from the 32 states included in the dataset as described in the text other than Hawaii.
intergroup diversity and the probability of an Asian American official requires a multivariate model, as the figure does not control for the share of Asian Americans.Model 2 in Table 2 presents a random effects multilevel logistic regression model that does so.
The marginal effect of intergroup diversity is expressed graphically in Figure 10.Even after controlling for the share of Asian Americans, intergroup diversity has a strong effect.An increase from a moderately low 0.25 to a moderately high 0.55 in intergroup diversity spurs a jump in the probability of an Asian American legislator from 0.58% to 1.00%-a substantively meaningful 72% gain considering that Asian Americans comprise only 2.5% of all cases. 32The results exhibit the same pattern for the U.S. House but are not statistically significant due to the smaller number of cases.
Tests of alternative models reveal that these results cannot be explained by changes in the share of any individual racial group.We examined models that, in addition to the percentage Asian VAP, controlled separately for the percentage of non-Hispanic white VAP, Black VAP, and Hispanic VAP.None of the models accord with the hypothesis that Asian Americans are more likely to win in districts with more non-Hispanic whites or the alternative that they fare better in districts with higher Black or Latino populations.Beyond falling well short of statistical significance, increasing the share of any of these non-Asian groups beyond a comparatively low level always results in a decline that has no effect on the probability of an Asian American legislator. 33GURE 10.Estimated Probability of an Asian American State Legislator as Intergroup Diversity Varies with 95% Confidence Intervals Asian American candidates benefit from greater intergroup diversity rather than the presence of members of any one other major racial or ethnic group.This aligns with theories suggesting that Asian American candidates have greater electoral opportunities the less a constituency is seen as "belonging" to any one racial or ethnic group.It does not support the notion that Asian American candidates do better in areas with more whites due either to favorable perceptions by whites of the so-called "model minority" or to frictions between Asian Americans and other minority community members.Nor do the results support the conclusion that Asian American candidates receive more support from other communities of color.

Individual-Level Evidence
Now, we examine Asian American candidates' crossover appeal to voters of different racial or ethnic groups, using our 2021 online survey.For this, we return to the candidate choice vignette, described above.Figure 11 leverages this vignette onto crossover voting by showing mean candidate support among white, Black, and Latino respondents, across every permutation of candidate race or ethnicity.Bars are labeled by respondent race/ethnicity (WR, BR, and LR) and candidate race/ethnicity (WC, AC, BC, and LC) such that, for example, "BR, AC" refers to mean support among Black respondents for an Asian candidate.
Assessments are broadly favorable and there is very little difference across the various combinations.There is some evidence of Asian American candidates' disproportionate crossover support in that white, Asian, and Latino respondents all seem to prefer Asian American candidates to whites all else equal; here, then, there is perhaps a hint of evidence for our model minority hypothesis (H 4 -MMH) in whites' preference for Asian American candidates over those who are members of other minority groups.There is also support for the shared minority status hypothesis (H 5 -SMSH) in Black and Latino preference for Asian over white candidates, all equal.But none of these differences are substantively large or statistically significant.Ultimately, then, evidence here is more consistent with the non-group-centric crossover story (H 7 -NGCC): there seems to be virtually no group differentiation per se, in the presence of other contextual information about the candidate and the election.
Taken together, unlike our evidence on ingroup support, the evidence on crossover presents a somewhat mixed picture.In the aggregate, we find no evidence of whites' preference for Asian Americans as a "model minority" (H 4 -MMH) and some suggestion that Asian American candidates do better in diverse, multiracial districts.This seems tentatively in favor of the notion that Black and Latino voters exhibit some kind of preference for Asian American candidates, either because of shared minority status (H 5 -SMSH) or as second preference versus a candidate of their own group (H 6 -NBCH).But at the individual level, the picture is somewhat different: there, we see very little in the way of statistically significant group differences, all else equal.This leaves us, at least tentatively, most closely in line with the non-group-centric-crossover model (H 7 -NGCC).

ROBUSTNESS IN THE AGGREGATE-LEVEL EVIDENCE
Beyond racial composition and diversity, we separately looked at a variety of alternative models designed to assess other theories.Some of these are tied to factors internal to the Asian American community that might be expected to amplify resources and turnout. 34To explore the effect of institutional arrangements, we also looked at single versus multimember districts, term limits, legislative professionalism, and population.Unlike in Casellas' (2011) study of Latino legislators, none demonstrated a significant relationship to the election of Asian American officials. 35Finally, we considered models that controlled for the education, as measured by the share of college graduates of non-Americans, as education has been linked to greater racial tolerance.These models revealed that education has a substantively and statistically meaningful positive link to the election of Asian American officials. 36Including this factor does not alter the central findings regarding the relationships of either intergroup or intragroup diversity to the election of Asian American legislators.

CONCLUSIONS
This paper has aimed to provide a comprehensive picture of descriptive representation of Asian Americans.Built on a new and unprecedented data collection effort, we map out where Asian American candidates have been able to succeed over the past few decades, how this success relates to district demographics, and, finally, gain some insight into why.In so doing, we build on a literature that has, for the most part, focused on deep dives on specific districts and specific races.For all the theoretical richness these studies provide, we simply have not known much-until now-about Asian American descriptive representation or the underlying dynamics of the election of Asian American legislators in the big picture.Nor do we know whether the claims made from these studies are generalizable.
In the aggregate, we show that: (1) Asian Americans have yet to attain descriptive parity with their population size in most U.S. contexts, but when they win, they are able to do so with a lower proportion of Asian Americans in their district than other U.S. minority groups win with theirs; (2) the more multiracial a constituency, the more Asian American candidates win; and (3) Intracommunity diversity has no impact on Asian American electoral success.Consistent with the aggregate-level findings, individual-level evidence shows a strong element of panethnic support for Asian American candidates that extends beyond the boundaries of that candidate's ethnic origin group.Because the experiments controlled for candidate policy position and competence, we interpret this difference as reflecting group-centric motives.In terms of crossover, we find-also consistent with the aggregate-level evidence-that Asian American candidates do not do especially well among white voters; rather, all racial groups exhibit a slight bias toward Asian American candidates (vs.white candidates), but the lack of statistically significant differentiation indicates that Asian American candidates' crossover support is less about race per se and more about other political attributes they bring to bear.The picture is less clear than it was with ingroup voting, however, and thus more evidence, especially at the individual level, is called for.
In this paper, we brought to bear a diverse yet complementary array of evidence on the drivers of support for Asian American candidates.This gives us substantial leverage to make claims about the demographic factors that help or hinder descriptive representation for this group, and some of the psychological factors predisposing voters in favor of or against its candidates.Yet of course much remains to be learned, especially with respect to campaign-related dynamics: we cannot say much about structural features of American politics (its two-party system and various institutional arrangements), nor can we address in depth campaign-related dynamics such as candidate and opponent rhetoric, media coverage, and so on.
The broader implications of these findings are, perhaps most significantly, that the electoral patterns of Asian American legislators are profoundly different than what has been found for Black and Latino legislators.The overwhelming evidence is that Black and Latino candidates win in districts where their group composes a majority.Asian Americans are more dispersed and, outside Hawaii, rarely form a majority in legislative districts.Their success nonetheless is broadly 34 We created separate models that controlled for the shares of Asian college graduates, Asians living in poverty, aged 65 and older, and aged 30 and younger.As education, income and age have all been positively linked to turnout in individual-level studies, each might relate to the election of Asian officials.None had a significant effect either alone or in models that also interacted the variable with the proportion of Asians. 35To measure legislative professionalism, we turned the National Conference of State Legislatures (NCSL) classifications into a fivepoint scale.Models controlling for the NCSL scale as well as dummies for highly professional and professional legislatures did not indicate a statistically significant relationship.Nor did models that controlled for multimember districts, term limits, population, or log(population). 36See Supplementary Table 11 and Supplementary Figure 8. Controlling for any interaction between the share of non-Asian college graduates and the share of Asians does not change the impact of either non-Asian educational levels or intergroup diversity on the election of Asian American legislators.
in agreement with the idea, established in the case studies cited earlier, that they are better able to achieve crossover support.Second, this crossover support is limited neither to whites and white majority districts nor to people of color and districts where they form a majority.Indeed, Asian Americans perform even better in districts that are broadly diverse, signaling crossover potential not just with whites but with Black and Latino voters as well.Like African Americans and Latinos, Asian Americans face barriers to representation, but the nature of these barriers is not the same (Kim 2018). 37ll of this bodes well for increased representation of Asian Americans as demographic change works to diversify districts.The effects of more targeted political efforts to achieve representation, on the other hand, have been less uniformly positive.The federal VRA has mandated ballots and other voting materials in Asian languages where the size of the language group is sufficiently large. 38But Asian Americans have not benefited from favorable redistricting maps that have aided Black and Latino efforts to overcome institutional barriers, including those of the electoral system (Chen and Lee 2013; Phillips 2021).
In the bigger picture, this study has triangulated a great deal of evidence on theoretical questions about the nature of crossover voting, and the power of panethnic political support.Legislative representation is of perennial interest both inside and outside of academic debates and our findings here not only shed light on one particular fast-growing (yet still understudied) group, but also point to a broader understanding of how minority communities can make their voices heard in legislatures across political contexts.

SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit http://doi.org/10.1017/S0003055423000242.

FIGURE 1 .
FIGURE 1. Asian American Legislators by Election Year and Party a. State Legislators

FIGURE 3 .
FIGURE 3. Asian American State Legislators by Ethnic Origin and Year

FIGURE 2 .
FIGURE 2. Ratio of Asian American State Legislators to Asian Population after the 2020 Elections, 33 States

FIGURE 5 .
FIGURE 5. Percent Asian and the Estimated Probability of an Asian State Legislator in Hawaii and 32 Other States

FIGURE 11 .
FIGURE 11.Non-Asian Voter Preferences by Candidate Race/Ethnicity

TABLE 1 .
Random Effects Logistic Regression Models of the Election of Asian American State Legislators VAP is the voting-age population.The dependent variable is Asian American winner with Asian American legislators coded as 1 and others as 0. Asian American winners include legislators with multiple racial origins, but their exclusion does not alter the substantive results.The model includes both state house and state senate elections for the states outlined in the text.Our estimator is multilevel logistic regression with random effects by district to address error clustering.We estimated all specifications with and without year fixed effects to account for the possibility that time trends may confound any results.Since all results are consistent regardless, we present the version without year fixed effects.Source: Data on the percentage of Asians from the American Community Survey.Data on legislator race gathered by author.See the text for more detail. Note:

TABLE 2 .
Random Effects Logistic Regression Models of the Election of Asian American State Legislators that Include Intergroup or Intragroup Diversity, Excludes Hawaii