Immigrant threat and national salience: Understanding the “English official” movement in the United States

Amy H Liu, Anand Edward Sokhey, Joshua B Kennedy, Annie Miller


The passage of (and debate over) immigration laws in Arizona highlights the increasing linguistic diversity of the US. To date, 31 states have passed an English-official bill. In this paper, we test several hypotheses concerning the adoption of such legislation across the states. Using data spanning the past three decades, we present event history models on the timing of adoption since the start of the modern movement in 1980. Like previous works, we find that the timing of adoption in states is structured by immigrant population and the initiative process. However, we find a conditional story that has been overlooked to date: the effects of immigrant threat only increase the likelihood of English-official legislation adoption when the issue of immigration is nationally salient.

  • Language
  • immigration
  • state politics


When Tocqueville (1835/2002) wrote about democracy in America, he remarked, “The bond of [the English language] is perhaps the strongest and most lasting that can unite men. All the emigrants spoke the same tongue; they were all children of one and the same people” (p. 29). While English remains the language of the US today, it is no longer the speech of all emigrants. For example, Arizona’s recent passage of Senate Bill 1070 reignited a national debate over immigration generally and language specifically. This is not an Arizona-exclusive incident. Recently, Oklahoma became the 31st state to pass English-official legislation; the Florida and Nevada Houses debated whether to require English for driver’s license tests; Washington considered a bill requiring translation services in pharmacies (English First, 2011; ProEnglish, 2011). The promotion of one group’s language over all others—especially in the name of national unity—can be and has been contentious (Horowitz, 1985). In this paper, we consider this topic. Specifically, what explains the adoption of English-official legislation?

While a sizable immigrant population (Citrin et al., 1990) and the presence of direct initiatives (Schildkraut, 2001) are important, neither factor can explain how individuals come to perceive immigrants as threats, and then act upon this threat. Here, we argue for the importance of national salience. When the subject of immigration appears often in the media, English-speaking locals—especially those in states that have large immigrant populations and that allow for direct initiatives—are made aware of and cued to participate in demanding and getting English-official legislation. In the remainder of this paper, we review the literature on English-official legislation, followed by a discussion of our argument. We then test the conditions under which immigrant size and initiative processes affect the adoption of English-official legislation, finding that national salience plays an important moderating role. We conclude with a discussion of the findings and implications.

Immigrant threat and English-official legislation

Figure 1 illustrates the trend in the adoption of English-official legislation. Prior to 1981 when Senator Hayakawa (Republican-California) proposed making English an official language (Del Valle, 2003), only five states (Hawaii, Illinois, Louisiana, Massachusetts, and Nebraska) had recognized English as a language for official capacities. However, ever since then, the US has witnessed a new era of English-official legislation. The five-year period between 1986 and 1990 was especially busy: during these years, 11 states passed some type of legislation designating English as official. What explains these temporal developments?

Figure 1.

Passage of English-official legislation.

Gray shades denote states that have already adopted English-official legislation. Black shades denote states that adopted said legislation in the given five-year span.

*Hawaii, Alaska.

The first explanation is a behavior story. In the US, despite the “melting pot” rhetoric, proficiency in the English language is a “highly resonant symbol of American nationality” for most citizens (Citrin et al., 1990: 536). Group unity—an important component of national identity —depends on the relationships between members of the in-group (Schildkraut, 2011). At some level, there is a “desire to keep ‘strangers’ out” due to the perceived potential for fracturing such individuals could pose to a broader concept of a unified national identity (Theiss-Morse, 2009: 17). However, when the immigrant population is sizable, governments are forced to recognize other languages in the public sphere. When this is the case, these policies can generate strong emotional responses, and calls for English-official legislation can garner mass appeal (Citrin et al., 1990: 556–557).

This anxiety is especially pronounced in the US, a country with a limited multilingual citizenry that is also a popular destination for non-English speaking migrants (Newman et al., 2012; Schildkraut, 2011). Consequently, there is an increasing demand for English-official legislation adoption. However, whether these demands translate into policy depends on a second explanation. From an institutional perspective, what is of importance is if the state allows for direct initiatives. When states that allow for direct initiatives experience an influx of immigrants, they are more likely to adopt such legislation; in contrast, in states that lack a directive initiative mechanism, the same cannot be said (Schildkraut, 2001: 446).

What explains this? Initiatives can change the calculations of state politicians. Generally, legislators prefer to stay away from polarizing issues to avoid antagonizing certain segments of their constituencies. This was the case in the early years of the English-official language movement (Tatalovich, 1995). However, when the initiative process is a channel for institutionalizing contentious issues, politicians have to consider whether voters may evaluate their representatives based on where they stand on those same issues (Donovan et al., 2008; Nicholson, 2005). Hence, state governmental actors are more likely to support policies that accurately reflect their constituents’ preferences (Arceneaux, 2002; Gerber, 1996; Matsusaka, 2001). In the context of our paper, this pattern should make English-official legislation more probable in states with an initiative process than in those without one. In the latter cases, the presence of large non-English-speaking populations increases the likelihood of greater political engagement among minorities; politicians fearing the electoral repercussions of potentially causing offense have incentives to avoid such matters (Schildkraut, 2001). This claim is consistent with the argument that states which allow for direct democracy are also more likely to adopt anti-minority legislation (Haider-Markel et al., 2007).

Both the behavioral and institutional accounts hinge on the notion of immigrant threat. However, absent in both is a mechanism for how immigrants come to be seen as threats. For instance, a large immigrant population is neither a necessary nor sufficient condition for the passage of anti-immigration ordinances (Hopkins, 2010). In fact, in some works, it is not a statistically significant factor for explaining popular support for “official English” (Citrin et al., 1990). Following this, we contend that what matters in explaining the adoption of English-official legislation is not the mere immigrant population size or the simple presence of a direct initiative process, but the national salience of the immigration issue. This perspective contends that when “salient national rhetoric politicizes immigration, immigrants can quickly become the target of local political hostility” (Hopkins, 2010: 40). The result of such hostility, we argue, is the introduction—and passage—of English-official legislation.

The concept of national salience also ties into the institutional argument about the initiative process’s effects on politicians and voters. Research on state initiatives shows that high salience can reduce the Downsian (1957) costs to voters, thereby increasing both issue-learning opportunities and voting likelihood1 (Lacey, 2005; Nicholson, 2003; Smith and Tolbert, 2004). The mechanism here involves “inviting the citizen [into] the political process … [which] may increase political efficacy, knowledge, and participation” (Nicholson, 2003: 409). This effect hinges largely on political forces that bring an initiative into the public consciousness (Lacey, 2005). Accordingly, politicians may change their strategic calculations in reaction to these reduced costs, and we posit that the adoption of English-official legislation becomes more likely as a result. The following hypotheses reflect these arguments:

  • Hypothesis 1: English-official legislation is more likely to be adopted when the issue of immigration is of high national salience.

  • Hypothesis 2: When the issue of immigration is of high national salience, English-official legislation is more likely to be adopted where there is a large immigrant population and the state allows for direct initiatives.

Research design

Our unit of analysis is state-year. The sample includes all 50 states between 1980 and 2010. We begin in 1980 because that is the start of the modern “English Official” movement (Feder, 2007). Like Schildkraut (2001), we estimate event history models to consider the factors that increase the risk of a state passing legislation. However, our analysis advances on previous work in at least three different ways: it extends the period of interest through the 2000s (Schildkraut’s analysis ends in 1998); it uses a modeling technique that is more appropriate given the sample distribution; and it incorporates the notion of national salience.

Dependent variable

The key concept of interest is the adoption of English-official legislation. A state-year observation is assigned a value of 0 if there is no legislation designating English as the language of official functions. However, once a state passes such legislation, the observation is assigned a value of 1. In the parlance of survival analysis, adoption is considered a “failure,” and we model the timing of “failure” (i.e., passage) from a number of covariates.

Key covariates

We have three key covariates: immigration size, initiative process, and national salience. To measure immigration size, we use data from the Department of Homeland Security’s Office of Immigration Statistics. The advantage of this data source compared to the more frequently employed US Census is twofold. Firstly, the Yearbook of Immigration Statistics is available electronically on an annual basis dating back to 1973.2 In contrast, the US Census is neither available annually nor readily available for the entire time period. Secondly, the Yearbook measures the number of non-US residents per state filing for green cards or similar benefits. Conceptually, these are the individuals who are more rooted in their communities and are more likely to have been the ones pushing for the recognition of non-English languages.

Our measurement of direct initiative is straightforward. Although most states that have direct initiatives allow them for both statutes and state constitutions, there is some variation, as noted by Schildkraut (2001). For example, Idaho, Utah, and Washington allow initiatives only for statutes. In contrast, Florida and Michigan allow them only for their constitutions. Here, a state is assigned a value of 1 if the direct initiative exists at any point in the political process.

Finally, to get at the national salience of immigration issues, we create a count of the yearly number of mentions of the topic in The New York Times (NYT) over the period of inquiry. Figure 2 shows the general trend of this measure. The NYT is the most prominent newspaper source in the US and is often an agenda-setter for other media outlets. We believe the NYT accurately represents the national media climate. In addition, other scholars have used the NYT for similar purposes (e.g., Buell and Sigelman (2008) on negativity in presidential campaigns from 1960 to 2004 and Epstein and Segal (2000) on the salience of Supreme Court cases from 1946 to 2009). In our analyses, we split the count of NYT articles at its mean, making a dichotomous indicator for “high” versus “low” salience.3

Figure 2.

National saliency of immigration.

Control variables

There are, of course, other variables that may confound the effects of the three key covariates. To address this, we control for several factors based on prior theoretical expectations.

State citizen ideology

Research has indicated that states with conservative-leaning citizens are more likely to favor initiatives that negatively target immigrant populations (Nicholson, 2005; Tolbert and Hero, 1996). To measure state ideology, we employ a measure developed by Berry et al. (2010) because its data is available in a time-series cross-sectional format; because the measure effectively incorporates both public measures and legislator opinion (which can be disaggregated or utilized together); and because the measure has lower random error than the alternative measure by Erikson et al. (1993).4

Labor force

The Bureau of Labor Statistics (BLS) provides yearly unemployment data. We contacted BLS directly, as historic figures were not readily accessible online. In the event history models, we control for labor force size within a state, normalized by the state’s population. This variable reflects general employment conditions, where high unemployment has been shown to positively affect the probability that legislation or initiatives targeted towards immigrant populations will pass (Nicholson, 2005; Tolbert and Hero, 1996).


Following Schildkraut (2001), we control for regional effects by incorporating a dummy variable for whether a state is in the South. The theoretical expectation is that Southern states are more likely to adopt English-official legislation.

Descriptive statistics

Table 1 presents general descriptive statistics on the 31 states with English-official statutes presently in effect. Each state is assessed at the time the law is adopted for whether it has a large immigrant population, whether it has an initiative process, and whether national salience on immigration issues was high. For presentation purposes, all three variables in this table are dichotomized. The immigrant population and national salience (i.e., mentions of immigration in the NYT) variables are split at their means. For the five states that adopted English-official legislation before 1980 (Louisiana in 1807, Nebraska in 1920, Illinois in 1969, Massachusetts in 1975, and Hawaii in 1978), we only note whether initiatives had been in place at the time of passage.

Table 1.

Adoption of English-official legislation.

We observe fairly wide variation among the states. Of the 31 states with English-official legislation, 14 had initiative processes in place at the time of adoption. Excluding states that had adopted such legislation by 1980, eight states adopted English as their official language in years in which national salience on immigration issues was high. Meanwhile, only four states with large immigration populations (defined as an immigrant population above the mean) adopted English-official legislation.

There are, however, some patterns. Firstly, in states with large immigrant populations, English-official legislation was adopted only when those states also had an initiative process in place. This conforms to the story advanced by Schildkraut (2001) and Tatalovich (1995): adopting English as the official state language remained unpopular among legislators in states where there were high numbers of immigrants, but the presence of an initiative process served to offset this concern.

Arizona and Oklahoma both adopted English-official laws when immigration populations were above the national mean, an initiative process was in place, and when national salience on the subject matter was high. Another seven states that adopted these laws are located in the South: Georgia, Kentucky, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia. This is consistent with theoretical priors that Southern states are more likely to adopt English as the official language (Schildkraut, 2001).

The lack of a broad pattern among states with English as the official language suggests precisely the kind of interactive relationship that we posit. Immigrant population, initiative processes, and national salience have only limited effects individually when considering the likelihood of the adoption of English-official legislation. Instead, we argue that these factors should be considered collectively in order to gain a clearer picture of why states choose to make English their official language.

Cox-proportional hazards survival model

To examine the adoption of English-official legislation in the US states since 1980, we employ a Cox-proportional hazards survival model. Event history is the proper analytic technique when the research question centers on timing (Box-Steffensmeier and Jones, 2004). Moreover, the Cox model is semi-parametric—that is, it does not impose assumptions about the duration dependence for the process of interest (Cleves et al., 2010). The advantage of this non-assumption is presented in Figure 3, which graphs the risk of English-official legislation adoption in states between 1980 and 2010. It is clear that the hazard rate decreases fairly rapidly during the first 15 years of the observation period, but then increases during the next 15. Common parametric survival models either assume flat hazard rates over time (e.g., exponential), or monotonic forms (e.g., Weibull). This is problematic, as fitting a model with an incorrect distributional assumption can produce misleading estimates (Box-Steffensmeier and Jones, 2004).

Figure 3.

“Risk” of English-official legislation adoption 1980–2010.

Moderating effects of national salience

In Table 2, we present the results from our Cox-proportional hazards model. In the context of event history analysis, positive coefficient signs indicate that adoption risk increases as a function of changes in a covariate (and negative signs the opposite). Given our hypothesis, we split the sample by “low” versus “high” salience, presenting two sets of estimates. In the first model (in which the sample includes only observations where national salience on immigration issue was low), neither the model nor any variable is significant. This general lack of findings supports our ex ante expectations: the immigrant threat is no threat when the issue of immigration is of low national salience.

Table 2.

Predicting adoption of English-official legislation 1980–2010 (survival analysis; Cox regression estimates).

In contrast, looking at the right-hand side of the table, we see a very different set of results. The overall model for high salience is significant, and critically, so are the key covariates of interest. In states without direct initiatives, immigrant population size can decrease the risk of English-official legislation being adopted. When the percentage of immigrants is zero (theoretically, this is an unrealistic scenario, but empirically the percentage is indeed extremely low in some states, where the minimum is 0.02%), having a direct initiative also reduces the risk English-official legislation will be adopted. However, under conditions of high salience, the interaction between these two covariates is positive and significant. When immigration is discussed frequently in the national media, being in a state with a higher proportion of immigrants and with a direct initiative increases the risk of official-English legislation being adopted. Substantively, the effect of this interaction translates into about a 42% increase in the likelihood of adoption between (1) a state with sample mean immigrant percentage and no direct initiative, and (2) a state with the sample mean immigrant percentage and a direct initiative. It is also worth noting that the state citizen ideology emerges as statistically significant and in the expected direction—liberal states are less likely to adopt English-official legislation.

Figure 4 plots the baseline and cumulative hazard functions for the high salience model—these are the estimates, conditional on the covariates. The non-monotonicity of the graph affirms the choice to use the Cox-proportional model: the cumulative hazard indicates adoptions are growing increasingly over the period of study, which coincides with the sharp rise in the hazard in the mid-2000s (see bottom maps of Figure 1).

Figure 4.

Retrieved hazard and cumulative hazard functions for official legislation in the states (high salience model).


As the number of non-English-speaking immigrants to the US increases, it is understandable why many see this trend as a threat. Proficiency in the English language is, after all, nearly synonymous with American nationalism (Citrin et al., 1990). However, whether the English-speakers respond to this threat depends on factors other than the number of non-English-speaking immigrants. One is whether the state allows for direct initiatives. Initiatives provide channels for the majority—that is, the English-speakers—to institutionalize their preferences. However, these processes are not a sufficient condition. A large immigrant population and the presence of an initiative process cannot account for how immigrants are seen as threats. Here we identify another factor: national salience. As a temporal dynamic, the immigrant threat is only a “threat” when the topic of immigration figures prominently in the national media. When salience is high, states with large immigrant populations and that allow for direct initiatives are more likely to adopt English-official legislation.

Some have voiced concerns about the effects of initiatives on minority rights (Haider-Markel et al., 2007). We address these concerns by demonstrating a conditional relationship between the presence of the initiative process and the likelihood of adopting English-official legislation. Specifically, the process is moderated by national salience. Initiative states certainly have a greater propensity to adopt English-official legislation, but this propensity depends upon how relevant concerns about immigration are to citizens at large. Put differently, while direct democracy can make adoption more likely, the good news is immigrants do not always lose with initiatives.


A previous draft of this paper was presented at the 2011 Western Political Science Association’s annual meeting (San Antonio, TX). We would like to Beth Reingold, Sophia Wallace, and two anonymous reviewers for extensive comments. Many thanks to Daniel Hopkins for sharing his data. All errors remain our own.

Article Notes

  • Declaration of conflicting interest The authors declare that there is no conflict of interest.

  • Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

  • Supplementary material The complete data sets of this article are available at:


  • 1. Typically, these works have assessed salience at the state (as opposed to the national) level.

  • 2. While illegal immigrant population size is important theoretically (i.e., threat does not differentiate over the legal status of a non-native English-speaker), the lack of reliable data across the sample precludes us from including it in our analysis.

  • 3. Our results are robust to alternative operationalizations of “high” salience. One alternative operationalization distinguished observations in which the count increased (“high salience”) from the previous year. Another focused on whether the count in a particular year was greater than the moving average of past years (“high salience”). The direction and significance of the key interaction—and most other covariates—remain unchanged. The results are also robust to splitting on the sample median, although they are less stable when using higher arbitrary cut-off points in the sample (sign and size of covariates remain the same but begin to lose significance at conventional levels).

  • 4. Berry et al. (2010) measure the ideological leaning of the electorate, governor, and legislature in each state. In their comparative analyses, Meinke et al. (2006) find the time-series properties of the Berry et al. (2010) measure perform better than Erikson et al. (1993) measure.

This article is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License ( which permits non-commercial use, reproduction and distribution of the work as published without adaptation or alteration, without further permission provided the original work is attributed as specified on the SAGE and Open Access page (


View Abstract