Education Populism? A Corpus-Driven Analysis of Betsy DeVos’s Education Policy Discourse

Scholars of political economy have raised the question of whether recent populist movements around the world signal the decline of neoliberal hegemony. What would such a decline mean for education policy, an arena that has been dominated by a neoliberal common sense for several decades? This study investigates the policy discourse of former U.S. President Donald Trump’s Secretary of Education, Betsy DeVos, in order to assess the extent to which it aligns with the neoliberal common sense or draws upon discourses of populism that have been gaining traction in the last few years. Using methods of corpus linguistics, I engage in a critical discourse analysis of 59 of DeVos’s public speeches delivered between 2017 and 2019 in comparison with a reference corpus of speeches delivered by DeVos’s predecessors in the George W. Bush and Barack Obama administrations. The findings, informed by Ernesto Laclau’s theory of populism as political logic and discourse, suggest that DeVos deploys several features of populist discourse even as she advocates policies that are characteristically neoliberal. I consider the implications of this discourse for education policy in the US.


Introduction
Scholars of political economy have raised the question of whether recent populist movements around the world signal the decline of neoliberal hegemony (Fraser, 2017;Kiely, 2020;Rose, 2017;Schmidt, 2017). What would such a decline mean for education policy, an arena that has been dominated by a neoliberal common sense for several decades (Rizvi & Lingard, 2010)? In the US, this common sense dates back to the 1980s and has assumed that the best way to improve schools-and thereby the nation's economic competitiveness-is to: 1) impose higher curriculum standards, 2) hold schools and teachers accountable for students' performance on standardized assessments, and 3) introduce an array of market-based reforms promoting school choice, private contracting, and incentive pay for teachers and school leaders, to name a few (Burch, 2009;Hursh, 2007;Mathis & Trujillo, 2016;Mehta, 2013). While these principles gained significant bi-partisan traction during the decidedly neoliberal presidencies of George W. Bush (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) and Barack Obama (2009, the 2016 election of Donald Trump may have indicated the beginning of an ideological shift in education policy. From the beginning of his candidacy, Trump used right-wing populist rhetoric to argue against the neoliberal common sense in areas outside of education, advocating economic protectionist policies, censuring political and cultural elites and technocratic experts, and calling for empowerment of the people, where "the people" were defined implicitly as White citizens who had supposedly been left behind by free trade agreements and liberal immigration policies (Kiely, 2020;Schmidt, 2017). Similar arguments have become common in Europe in the last decade, used for example by populist leaders in the UK during the Brexit campaign and by elected officials and political candidates in France, Austria, Hungary, and Poland (Mounk, 2018). And yet, while a populist turn is now evident around the globe and in the US, education research has been relatively silent on the question of whether populist discourses have entered the education policy arena-and, if they have, whether such discourses indicate growing cracks in the neoliberal consensus of education reform.
This study investigates the policy discourse of President Donald Trump's Secretary of Education, Betsy DeVos, in order to assess the extent to which it aligns with the neoliberal common sense or draws upon the discourses of populism that have been gaining traction in the last few years. Prior to her appointment as secretary of education, DeVos was known chiefly for her school choice activism-particularly her support of vouchers, which enable families to use tax dollars for tuition at private or religious schools (Brown, 2016). As a billionaire philanthropist and major donor to the Republican party who lacked professional experience in education, DeVos proved to be a controversial pick for the post; her Senate confirmation was the first cabinet nomination in U.S. history to require a tie-breaking vote from the Vice President (Huetteman & Alcindor, 2017). For those on the Right, DeVos's appointment signified a promising future for school choice, while for those on the Left, DeVos's priority was nothing short of the dismantling and privatization of public education in the US (Brown, 2016).
Understanding DeVos's discourse is important, I want to suggest, because whether or not she was successful in accomplishing her policy goals, we need to recognize the cumulative power of repeated ideas and discourses (Lakoff, 2016), particularly when they come from someone who occupies a privileged position such as that of the secretary of education (Arce-Trigatti & A. Anderson, 2020). From a Gramscian (1973) perspective, such ideas and discourses may emerge as hegemonic, as the next new common sense, and so it is crucial that critical scholars pay attention. Furthermore, DeVos's policy preferences are hardly idiosyncratic; her support for vouchers, for minimizing (if not eliminating) the federal government's involvement in public education, and for deregulating for-profit K-12 schools and universities have long been promoted by influential conservative think tanks and policy advocacy organizations such as the Heritage Foundation, The Cato Institute, and the American Legislative Exchange Council (G. Anderson & Montoro Donchik, 2016). These organizations supported DeVos during her tenure as secretary, even suggesting at times that she ought to defend her policy proposals more forcefully (see, e.g., McCluskey, 2018). DeVos's discourse thus provides a window into a mode of thinking that is not hers alone, and regardless of her ability to enact her policy preferences, her position as secretary enabled her, at the very least, to lay an ideological groundwork for future policymakers with similar ideas.
Using methods of corpus linguistics, I engage in a critical discourse analysis of DeVos's public speeches in comparison with those of her predecessors in the Bush and Obama administrations. The question guiding this study is as follows: To what extent does Betsy DeVos's education policy discourse represent continuity or change in relation to the neoliberal consensus of education reform? The paper begins with a brief review of recent scholarship on populism as a political logic with a set of distinct discursive features. Next, I discuss the question of whether recent populist movements indicate an ideological break from the neoliberal consensus. After describing the study's corpus-driven approach to discourse analysis, I present an analysis of the study's findings and discuss their implications for education policy in the US. Ernesto Laclau's (2005) influential theory of populism emphasizes its discursive features. According to Laclau, populism is a political logic that deploys a dichotomous division between the elite establishment or institutional system and a group that comes to identify as "the people." The discursive construction of "the people" as a new political subject begins when a series of political demands-for example, more restrictions on immigration, lower taxes, tougher policing-remains unfulfilled by the current institutional regime. Although these demands may, on the surface, share little in common with each other, the groups who make these demands begin to develop a solidarity based on their opposition to the current regime. That is, they develop a collective identity based not on the specific content of their individual demands, but on their common rejection of the elite establishment, which they see "as having betrayed the public trust" (Bonikowski & Gidron, 2016, p. 1596. Populism thus tends to have an anti-institutional character, a rejection of institutions such as representative government, independent media, bureaucracies-anything that moderates the direct expression of the will of the people (Brubaker, 2017;Miró, 2019;Molyneux & Osborne, 2017). Laclau (2005) argues that the construction of the populist identity-of "the people"-is based on a logic of equivalence, whereby the groups in the populist camp come to see their various demands as fundamentally the same. Hence the frequent reliance on discourse that is vague and imprecise, as in the use of empty signifiers. In order for a populist movement to become politically mobilized, Laclau notes, its adherents need to coalesce around some symbols that represent their basic rejection of the establishment, words or phrases that will gloss over any surface-level differences among their particular demands. Signifiers such as freedom and justice, for example, serve this purpose well because they can mean a variety of things to different people; their emptiness, their lack of any strict denotative meaning, makes it possible for them to represent a wide range of social demands (Laclau, 2005). Schmidt (2017) demonstrates how Trump's campaign slogan, "Make America Great Again," succeeded because of its semantic flexibility (Schmidt, 2017). For those who saw globalism and deindustrialization as the roots of their economic woes, the slogan could signify a call for making America industrial again. For racists who rejected multiculturalism and opposed immigration, the slogan could covertly suggest a call for making America White and Christian again.

Populism as Political Logic and Discourse
The empty signifier worked because it was vague and imprecise enough to capture a wide variety of political demands.
Populist discourse is also distinctive in its preference for simplified representations of political issues. As Brubaker (2017, p. 367) explains, the populist style uses "rhetorical practices of simplicity, directness, and seeming self-evidence, often accompanied by an explicit anti-intellectualism… that valorizes common sense and first-hand experience over abstract and experience-distant forms of knowledge." What matters, then, is not what experts know and recommend, but what the people believe is right for themselves. In the context of the COVID-19 global pandemic that began in 2020, for example, Trump and his administration actively discredited and misrepresented scientific advice regarding the use of protective masks, suggesting instead that people should decide for themselves whether or not to wear one (Tollefson, 2020). Such distrust of experts may explain the prevalence of normative arguments in the discourse of recent populist movements (Schmidt, 2017). Unlike cognitive arguments, which "justify in terms of expert knowledge and logics of causation," normative arguments "legitimate through appeal to societal values and logics of appropriateness" (Schmidt, 2017, pp. 250-251). Thus, in populist discourse, we may expect to see policy proposals justified not by scientific evidence of their effectiveness, but rather by their simple congruity with a set of beliefs about what is right, or the way things ought to be.

Beyond Neoliberalism?
The rise of populist politics in the last few years may represent the most sustained and electorally successful attack against neoliberal hegemony that we have seen in recent decades. Fraser (2017) points out that in the US, where a neoliberal common sense had prevailed among both major political parties since at least the 1990s, those who resisted the common sense tended either to be marginalized or eventually contained by the neoliberal mainstream. Such was the case in 2012: Although President Barack Obama borrowed some of the class-conscious counter-discourse of Occupy Wall Street in his campaign for reelection, neoliberalism would remain hegemonic through his second term (Fraser, 2017). It was not until 2016 that two presidential candidates-Bernie Sanders on the Left, Donald Trump on the Right-would launch explicit attacks on neoliberalism, calling attention to its fundamental flaws and destructive effects on the working class (Schmidt, 2017). Sanders would lose the primary to Hillary Clinton, who represented the neoliberal mainstream, but Trump would win the presidency, bringing his conservative populist rhetoric to the White House.
The recent success of populist movements may signify what Kiely (2020, p. 398) calls the "beginning of the end of neoliberal hegemony," or what Streeck (2017) refers to as a sign of an interregnum, a chaotic and uncertain period between the neoliberal order and a new, yet-to-bedefined institutional order. Yet these scholars are, of course, tentative about these claims. It is too soon to know if neoliberalism is nearing its end; after all, one of its essential characteristics, and arguably a chief source of its resilience over several decades, is its flexibility and mutability (Schmidt, 2016). Kiely (2020) raises the question of whether today's conservative populists are truly rejecting neoliberalism or actually ushering in yet another new form of neoliberalism, one that attempts to resolve some of its internal contradictions-for example, its paradoxical vision of state power. While neoliberalism calls for a non-interventionist state and promotes market freedom, Kiely (2020, p. 405) notes, it also "relies on the state to carry out its project." David Harvey's (2005) definition of neoliberalism is relevant here. According to Harvey (2005, p. 2), neoliberalism is a theory of political economic practices that proposes that human well-being can best be advanced by liberating individual entrepreneurial freedoms and skills within an institutional framework characterized by strong private property rights, free markets, and free trade. The role of the state is to create and preserve an institutional framework appropriate to such practices. Harvey (2005) calls attention to the state's active role in creating and maintaining the conditions for free markets and privatization. In a similar vein, Kiely (2020) and Rose (2017) explain that despite neoliberal rallying cries for minimizing government and promoting free markets, neoliberalism has coincided with a rather activist state, as evident in the governing technologies of New Public Management. Since the early 1980s, neoliberal "freedom"-a watchword of the ideology-has ironically required a burgeoning of governmental apparatuses. In education, for example, the state has had to create markets, set performance standards and benchmarks, develop new assessments and data systems, and design sophisticated evaluation regimes to audit the outcomes of schools and individual teachers (G. Anderson & Cohen, 2015).
Kiely (2020) asks whether such excesses of the neoliberal state are the root of neoliberalism's decline or merely of a conservative populist movement calling for a new form of neoliberalism. Noting, for example, Trump's "rejection of any notion of 'the public' and his reduction of social interaction and diplomacy to private transactions," Kiely (2020, p. 399) suggests that Trump may represent "the culmination of neoliberal rule." For conservative populists, Kiely (2020, p. 407) argues, neoliberalism's "libertarian promise has somehow been betrayed"; what is needed, according to them, is not an alternative to neoliberalism, but a neoliberalism without interventionist government. Fraser similarly suggests (2017) that while Trump used an anti-neoliberal populist discourse to get into office, he has actually governed with a "hyper-reactionary neoliberalism"-a rejection of the sort of neoliberalism that supported relatively progressive policies on matters ranging from climate change to racial, ethnic, sexual, and religious diversity.
Which brings us to the question of where Betsy DeVos stands as the educational representative of Donald Trump's conservative populism. Although education scholars have produced a wealth of literature critiquing neoliberal policy over the last two decades, the field has had relatively little to say regarding the possibility that a post-neoliberal era is near or even underway. Perhaps this is because market-based reforms, audit cultures, and other typical manifestations of neoliberalism continue to mark the education policy landscape and show few, if any, overt signs of fracture. Wolgemuth et al. (2018, p. 3), however, open the question of whether the election of Trump and the rise of right-wing populism signify a "referendum against centrist liberals and their support for neoliberal policies in education." Means and Slater's (2019) recent analysis of neoreactionist discourses as a sign of crisis in progressive neoliberalism constitutes another exception to the relative silence on post-neoliberal formations. Importantly, they argue that "postneoliberal imaginaries are being produced that escape current mainstream academic debate," and that "this lacuna should be viewed as a point of vulnerability if educational studies are to be connected to political struggles for a just social order" (p. 172).
Taking a cue from these authors, the present study offers a critical analysis of education policy discourse as an indicator of the possibility of wider ideological change. In conceptualizing ideology, I am influenced by Eagleton's (1991, p. 29) point that it often involves "the promotion and legitimation" of the political interests of a social group in the face of competing or opposing interests. In this sense, ideological discourse has a persuasive or rhetorical quality, and it is concerned with "the sustaining or challenging of a whole political form of life" (Eagleton,p. 29). Although this study pays close attention to a set of keywords identified in DeVos's speeches, it is important to note that single words are not inherently ideological; rather, words can carry ideological significance only in their use (Fairclough, 2015). Thus, as I will explain below, while I used quantitative methods to analyze the saliency of certain words across the corpus of DeVos's speeches, I used qualitative methods to interpret their meanings-including ideological significance-in their contexts.

Data
The primary dataset for this study consisted of a research corpus of 59 speeches (100,316 words) delivered by U.S. Secretary of Education, Betsy DeVos, during the first three years of the Trump administration, 2017-2019, hereafter labeled the "DeVos corpus." In order to compare the ideological themes of this corpus with those of the two presidential administrations preceding this time period, I assembled a discourse reference corpus of 596 speeches: 364 speeches delivered by the two secretaries of education under the George W. Bush administration, Rod Paige (2001-2004) and Margaret Spellings (2005, and 232 speeches delivered by the two secretaries who served the Barack Obama administration, Arne Duncan (2009)(2010)(2011)(2012)(2013)(2014)(2015) and John King, Jr. (2016). Together, the DeVos corpus and reference corpus consisted of approximately 1.3 million words spoken by education secretaries over the course of 19 years. Table 1 summarizes the number of speeches and words spoken by the secretaries for each administration. The DeVos corpus and reference corpus included only prepared speeches that focused on policy matters or that conveyed beliefs about P-12 or higher education policy. These speeches, which were publicly available on the US Department of Education website and on an Internet archive site (Internet Archive, https://web.archive.org/), included testimonies before Congressional committees, speeches before special interest groups, keynote addresses before professional association meetings and conferences, and remarks at award ceremonies such as the National Teacher of the Year ceremony.
Of course, speeches like these represent only one discourse genre in which secretaries of education communicate with the public. Secretaries also sit for media interviews, publish Op-Eds, contribute posts to social media such as Twitter, and more. Given this study's method of comparing Secretary DeVos's discourse with that of her predecessors, it was important to select a discourse genre that was used consistently by all secretaries in the reference corpus and the DeVos corpusthis to rule out the possibility that differences across speakers were attributable to disparate genre conventions. Furthermore, as Arce-Trigatti and A. Anderson (2020, p. 4) note, speeches of U.S. secretaries of education are "delivered to a wide and varied audience and [are] covered by widely circulated media outlets. As such, they serve as an interesting medium through which to critically engage current trends in American education."

Methods
This study uses methods derived from corpus-driven critical discourse analysis. This computer-based approach, which combines quantitative procedures of corpus linguistics with qualitative analytical tools of critical discourse studies (Baker, 2006;Fairclough, 2015), enables a researcher to investigate patterns of ideological continuity and change in large bodies (corpora) of texts that were produced over a specific period of time (Mulderrig, 2012). The quantitative procedures of corpus linguistics provide a macro-level view of a corpus's lexical and (sometimes) grammatical features; in turn, these identified features can be analyzed and interpreted at a microlevel with the qualitative techniques of critical discourse analysis.
It is important to note that, as a mixed-method approach, corpus-driven critical discourse analysis privileges neither quantitative nor qualitative methods. As Baker et al. (2008, p. 274) have stated, each set of methods "contributes equally and distinctly to a methodological synergy." In this study, while I used quantitative measures of keyness to analyze the distinctive word choices of the DeVos corpus, I conducted qualitative analyses of concordances-lists of all occurrences of these words presented in their immediate context-to interpret their meaning and function in DeVos's speeches. For example, as reported in the findings below, keyness measures revealed that choice was a particularly salient word throughout the DeVos corpus, occurring with statistically significantly higher frequency than in the reference corpus. Yet a decontextualized list of words like choice provides only a surface-level indication of a corpus's content and style. Thus, qualitative methods of concordance analysis, which attend to context, were necessary to uncover the way DeVos tended to deploy choice in her speeches.
Given my purpose of tracing ideological continuity or change across a rather large body of texts, both quantitative and qualitative methods were necessary. The quantitative methods provided a macro-level view of the DeVos corpus, identifying its important linguistic features and helping to ensure that micro-level, qualitative analyses were focused on features that were representative of the corpus as a whole (Baker et al., 2008). Furthermore, the qualitative analyses facilitated theoryinformed interpretations of features that were sensitive both to text-internal context and sociopolitical trends. A fuller explanation of the two main analytical tools used in this study, keyword and concordance analysis, is provided below.

Keyword Analysis
After reading DeVos's speeches to develop a basic familiarity with their content and discourse, I used WordSmith Tools 7.0 (Scott, 2016) to identify the top 50 keywords of the DeVos corpus. In corpus linguistics, keyness is a quantitative measure of the saliency (as opposed to mere frequency) of certain words in a corpus in comparison with some norm-that is, with a reference corpus of similar types of texts (Baker, 2006). As Scott (2010) notes, keywords indicate the "aboutness" of a text or collection of texts and can also indicate an author's distinctive style.
A variety of statistical measures can be used to identify the keywords of a corpus based on a comparison between words' observed frequencies and their expected frequencies, where expected frequencies are derived from the reference corpus data (Brezina, 2018). For this study, I used the log-likelihood statistic, a common measure of keyness that indicates whether the difference between a word's observed and expected frequencies is statistically significant (Brezina, 2018). I used this measure to identify both the positive and negative keywords of the DeVos corpus. Positive keywords are those that occur with unusual frequency when compared with the baseline of a reference corpus; conversely, negative keywords occur with unusual infrequency. While my analyses focused chiefly on the positive keywords of the DeVos corpus (as these provided a more direct indication of the corpus's salient features), I also examined the list of negative keywords to identify the sorts of discourse that were common in the reference corpus but nearly absent from DeVos's speeches.
In identifying the keywords of the DeVos corpus, the significance level was set at p < 0.000001, signifying 99.9999% confidence that a given word's presence in the corpus "isn't due to chance but a result of the author's (conscious or subconscious) choice to use that word repeatedly" (Baker, 2006, p. 125). 1 I also employed an additional requirement that a positive keyword appear in at least 10% of the texts in the DeVos corpus. This threshold prevented the inclusion of keywords that may have been frequent only because they were the specialized topic of a few isolated speeches (Mockler, 2018). Furthermore, following Mockler (2018), I supplemented the log-likelihood statistic with Hardie's (2014) log ratio as the latter is an effect size statistic. In other words, while loglikelihood can establish statistical significance, the log ratio can tell us "how big / how important a given difference is" (Hardie, 2014;emphasis in original). Only keywords with a log ratio statistic of 1.50 or higher were included, signifying that the relative frequency of a given keyword was, at the very least, nearly 3 times greater than that in the reference corpus. Combined, these methods helped to produce results that were significant and relatively strong.
Although corpus-driven discourse studies sometimes use a "stop list" to exclude function words from the analysis (e.g., articles, pronouns, modal auxiliary verbs, and others that have little lexical meaning in themselves), I chose not to employ any such restrictions. Charteris-Black (2012) demonstrates in a keyword analysis of political speeches that function words can reveal important features of a speaker's rhetorical style (Charteris-Black, 2012). And, as noted above, theories of populism-as-discourse suggest that certain stylistic choices are germane to populist politics. Indeed, the findings of this study demonstrated that several function words-e.g., should, nor, everythingplayed important roles in conveying DeVos's ideological interests.
The use of a reference corpus of same-genre texts is crucial here because it provides a baseline of what we might expect to see in a collection of texts in that genre (Brezina, 2018). For example, when I compared the DeVos corpus with the reference corpus of all other secretaries of education in the dataset, the words "education" and "school" were not identified as keywords because they were (unsurprisingly) highly frequent in both the DeVos corpus and the reference corpus. The word "freedom," however, was identified as the first positive keyword (ranked by loglikelihood) because it was highly frequent in the DeVos corpus but rarely used in the reference corpus. Thus, we can say that "freedom" is a particularly salient term for DeVos, and as it is recognizable as a watchword of neoliberalism (Harvey, 2005), we would do well to analyze its use in the DeVos corpus with qualitative techniques.
The identification of specific keywords for in-depth analysis is not a purely quantitative procedure (Charteris-Black, 2012). Given my interest in assessing the extent to which DeVos's discourse represented continuity or change with respect to the neoliberal consensus, I selected for deeper analysis any keywords that could potentially hold ideological significance. As noted above, although some words are immediately recognizable as ideologically-contested terms (e.g., freedom, justice), words serve ideological ends only in their use (Fairclough, 2015). Thus, to determine whether a keyword has an ideological meaning or function in a corpus, one must analyze the keyword in its context.

Concordance Analysis
In corpus linguistics, analysis of keywords in their contexts is typically performed with the concordancing function of software such as WordSmith Tools 7.0 (Baker, 2006;Scott, 2016). A concordance is a list of every occurrence of a word of interest in a corpus; it is displayed as a set of lines of text that include the word of interest and a few words to its left and right, allowing the researcher to examine each occurrence of the word in its context. Sorting the concordance lines alphabetically by words appearing to the left or right of the keyword, the researcher can begin to detect patterns of usage. Table 2 shows a sample of 13 of the concordance lines for "should," the second-highest ranked keyword in the DeVos corpus. Because "should" appeared 318 times in the DeVos corpus, the entire concordance of "should" consisted of 318 lines of text; this extract includes only lines 86 through 98 to demonstrate one pattern of meaning. In order to reveal what "should" be done or "should" be the case according to DeVos, I sorted the lines alphabetically by the first and second words to the right of "should." Several patterns can be detected in the entire concordance of this word, but in this small sample, the most obvious pattern is the idea that students and families "should be free" to make their own choices about their education. This pattern, along with others in the remaining concordance lines, suggested an ideological significance-the promotion of an interest in the face of opposing interests-that needed further investigation. Learning needs to be lifelong, and students should be free to decide their education each 92 own students. And we believe students of all ages should be free to pursue the education that's 93 student learns differently. That every student should be free to do so is why I'm here. It's also 94 bureaucrats tasked to make decisions families should be free to make for themselves. Just 95 students so each of them can grow. Every option should be held accountable, but they should be 96 every day on behalf of their students. In fact, they should be honored, celebrated, and freed up to do 97 education. They are also important values that should be honored when we engage in 98 creates risks of its own. The reality is… we should be horrified of not changing. Our children Having used a preliminary round of concordance analysis to select keywords with apparent patterns of ideological significance, I analyzed the concordances of these words in greater depth to understand their meaning and function in the corpus. In cases where a concordance line provided insufficient context to determine meaning and function, I referred to the full text of the speech in question. I also examined the most frequent 2-and 3-word clusters within a 5:5 window (5 words to the left and 5 words to the right) of each keyword, which helped to indicate the sorts of phrases in which DeVos tended to use the keyword. Throughout this stage of analysis, I kept analytic memos for each keyword, noting patterns of meaning in the concordance, the most frequent 2-and 3-word clusters associated with the keyword, and my preliminary ideas about the keyword's significance. I then used an additional round of concordance and cluster analysis to refine my analytic memos.
These concordance analyses were informed by methods of critical discourse analysis. Using Fairclough's (2003Fairclough's ( , 2015 work, I was sensitive to lexical features such as ideologically-contested terms (e.g., "freedom," "choice"); grammatical features such as modality (cases where the speaker expresses degrees of obligation or certainty about a proposition); and the texturing together of different discourses (e.g., discourses of neoliberalism and populism). Ultimately, I used the refined analytic memos to group the keywords into thematic categories of ideological content (e.g., "antiinstitutionalism") and rhetorical function (e.g., "normative argument").
I employed a combination of emic and etic perspectives to develop these thematic categories (Fetterman, 2008). For example, I found that a number of keywords, when analyzed in their immediate context, seemed to express the value of individualism (e.g., individual, unique, different, fit/s,) and so I initially considered individualism as a theme in the DeVos corpus. Here I was interpreting DeVos's words from an emic perspective, treating them more or less at face-value. In further rounds of concordance analysis, however, I noticed that DeVos consistently celebrated individualism in the context of a more general rejection of institutions and a questioning of the experts and elites who administer them. Here, my familiarity with theories of populism and the sociopolitical context of Trump's presidency facilitated an etic or external perspective. I thus read DeVos's notion of individualism as part of a more general theme of a populist anti-institutionalism, a theme that was further confirmed as I identified additional populist tropes in the corpus. Themes generated in this way ultimately contributed to an overall interpretation of the ideological character of DeVos's discourse.
To summarize the contributions of quantitative and qualitative methods in this study, Figure  1 provides a visual representation of the various layers of analysis and interpretation. Concordance analysis provided a qualitative, micro-level, context-sensitive investigation of the keywords generated in the quantitative, macro-level analysis of the corpus. Themes, as described above, were built on quantitative salience of keywords and the qualitative analyses of their concordances. In turn, an overall interpretation of the ideological character of DeVos's discourse was based on the integration of the study's macro-and micro-level analyses of the corpus.

Limitations
Although the advantages of quantitative and qualitative methods complemented one another as described above, it is important to discuss some limitations of the research approach. This study focused on DeVos's discourse, using the discourse of preceding secretaries of education only as a reference corpus. While this approach foregrounded the salient features of DeVos's discourse in relation to that of her predecessors-revealing a distinct hybrid of neoliberal and populist discourses-it did not rule out the possibility that her predecessors may have deployed some of these same features, though to a smaller degree. Including the negative keywords of the DeVos corpus, however, helped to demonstrate that certain key features of her predecessors' discourse (e.g., the language of New Public Management) were nearly absent from DeVos's speeches. 2 This evidence provided further confirmation of substantive distinctions between DeVos's discourse and that of her predecessors, but separate studies of the speeches of each of these secretaries would be necessary to determine whether DeVos's discourse is a truly novel phenomenon.
Furthermore, the identified keywords of a research corpus are a function of the genre of its texts as well as of the characteristics of the reference corpus. My decision to use only prepared speeches as research and reference corpora-due in part to the availability of texts, but also to the Interpretation (ideological character of DeVos's discourse)

Themes
(patterns of ideological content and rhetorical function)

Keyword Analysis
(quantitative analysis of salient discursive features of corpus)

Concordance Analysis
(qualitative critical discourse analysis of keywords in context) importance I placed on comparing discourse within the same genre-assumes that such speeches are a reliable indicator of a given administration's education policy discourse and ideological interests. It is possible that the inclusion of other genres of discourse-e.g., interviews, memoranda, social media posts-would have produced somewhat different results. Finally, it should be noted that the research and reference corpora consisted only of prepared text; as such, they did not capture instances where a secretary may have diverted from the script.

Findings and Discussion
It is possible to discuss in depth only a handful of the keywords of the DeVos corpus in this limited space, but as the analyses below will demonstrate, each of these keywords is connected in some way to other keywords in the list-whether by patterns of mutual proximity within the speeches or by semantic or functional relation. Thus, in what follows, I will focus the discussion on several keywords that represent the major themes and discursive characteristics of DeVos's speeches, extending the analyses at times to demonstrate their relation with other keywords. Table 3 provides a list of the top 50 keywords in the DeVos corpus, grouped by four major themes and listed in order of their keyness. The first theme, anti-institutionalism, captures 20 keywords that DeVos used primarily to communicate ideological content. Although one could easily predict that some of these keywords would have ideological significance (e.g., freedom, choice), for others, ideological significance became evident only during the concordance analysis stage (e.g., buildings). Concordance analyses revealed that many of these words had to do with empowering individuals in the face of institutions that limit their freedoms or fail to honor their unique qualities and needs (e.g., different, individual, unique, fit/s, empower). The remaining three themes capture 14 keywords that were also used to express ideological content but primarily served a rhetorical function: normative argument, dichotomization, and imprecision. Most of the keywords representing these three themes are function words (e.g., should, nor), but a handful of content words are also included (e.g., rethink, embrace/d) because the concordance analyses revealed that their rhetorical functions in the speeches were more prominent than any patterns of specific content they expressed. Sixteen of the top 50 keywords were not categorized in any particular theme because the concordance analyses did not reveal any strong patterns of ideological meaning or function.

Anti-Institutionalism
The most highly ranked keyword in the DeVos corpus, freedom, appears 307 times in DeVos's speeches. Although the word freedom appears in the title of one of DeVos's specific policy proposals, "Education Freedom Scholarships," this three-word cluster (including the singular, "Scholarship") accounts for only 29 occurrences of freedom in the corpus. Thus, in 91% of its occurrences, freedom serves a purpose other than to identify a specific policy proposal. For example, the two-word cluster, "freedom to," occurs 82 times and "education/al freedom" occurs 57 times without the word "scholarship/s" immediately following.
It is noteworthy that the most highly ranked keyword in the DeVos corpus, freedom, is an ideologically-contested term, a watchword for neoliberalism, which has historically called for free markets, individual freedoms, freedom of choice, and freedom from government regulations (Harvey, 2005). Freedom is also an example of what Laclau (2005), in his theory of populism, has called an empty signifier, a word that has very little content of its own and can therefore be used by a group or its leader to capture the wide variety of political demands that are typical of populist movements.
The semantic flexibility of the word freedom is evident throughout the DeVos corpus. In a representative passage from a speech delivered in 2018, DeVos connects freedom to school choice (here and below, I have underlined all keywords in quoted passages in order to demonstrate their links to the keyword of interest): Choice is really about freedom! Freedom to learn, and to learn differently. Freedom to explore. Freedom to fail, to learn from falling and to get back up and try again. It's freedom to find the best way for you to learn and grow...to find the engaging combination that unleashes your curiosity and unlocks your individual potential. (July 25, 2018) This brief passage is something of a refrain for DeVos; it appears with several variations in 12 of the 59 speeches in the corpus. Here and in the 11 similar passages, DeVos provides a remarkably vague explanation of what freedom is. It is a descriptor for choice, but it is also connected to a series of verbs about learning and personal development: Freedom is about being able "to learn" in general, "to learn differently" (presumably from the way other people learn), "to fail," and to learn from failure. DeVos then links freedom to two additional verbs related to a general notion of seeking: freedom "to explore" (without a specified object of exploration) and freedom "to find" the kind of learning that is best for you as an individual. This expansive concept of freedom seems to allow DeVos to abstract away from any specific explanation of what she means by choice as a policy preference. Introducing a version of this refrain in 2019, DeVos criticizes the sort of solutions that, she claims, the Department of Education has used in the past to address U.S. students' low scores on standardized tests, but to no avail: "more spending, more regulation, more government" (October 1, 2019). For DeVos, a better solution is freedom, which she now splices with a variation on Trump's famous slogan, "Make America Great Again," itself an empty signifier (Schmidt, 2017): That's why now is the time to do something different. Something better. Embrace the thing that makes America great: Freedom. The freedom to learn. The freedom to grow. The freedom to rise. The freedom to pursue happiness. (October 1, 2019) Whereas DeVos initially seems on the verge of proposing a policy-of specifically identifying "something different" and "something better"-she instead offers the empty signifier freedom. Once again, DeVos defines freedom with a series of verbs that refer expansively to learning, personal development, and seeking: "learn," "grow," and "rise." Now, however, she adds an allusion to a famously vague phrase from the Declaration of Independence: the "unalienable rights" of "life, liberty, and the pursuit of happiness." Adams et al. (2019, p. 195) draw attention to the tendency of neoliberal systems to subject individuals to a "growth imperative": Neoliberal systems do so not only by providing a sense of freedom from constraints (including interference of oppressive others who would impose rules and regulations), but especially by providing freedom to pursue defining aspirations-to do what you want or what you like-and thereby to achieve happiness and well-being.
(emphasis in original) Through her references to freedom, DeVos readily deploys a neoliberal discourse that exhorts individuals to develop themselves and to find what satisfies them.
DeVos's concept of freedom, vague as it is, can make sense only through references to some unfreedom, explicit or implicit. As Laclau explains in his discussion of empty signifiers, "the semantic role of these terms is not to express any positive content but, as we have seen, to function as the names of a fullness which is constitutively absent" (2005, p. 96, emphasis in original). It would be wise, then, to search the corpus for explicit identifications of what DeVos sees as hostile to freedom. Although the two-word cluster, "freedom from," occurs in only four speeches, the words and phrases that follow this cluster are telling. For DeVos, the freedom we need is freedom "from a one-size-fits-all mentality," "from centralized control," "from 'the system'," "from Washington mandates" (January 16, 2018), "from government" (May 1 and May 17, 2019), and "from over reaching mandates from Washington" (March 13, 2017). Freedom's antagonist seems to be not only the governing establishment, but, more generally, institutional arrangements that demand the same behaviors from everyone, denying their individuality.
DeVos's anti-institutional orientation becomes especially clear when we examine her use of the word choice, another watchword of neoliberalism (see, e.g., Friedman & Friedman, 1980), which occurs 180 times in the corpus and ranks third on the list of keywords. Given DeVos's activism for school choice policies prior to her appointment as secretary, it is hardly surprising that she uses the word choice so frequently, but its status as a keyword is noteworthy. We should recall that keyness is not a measure of raw frequency, but a measure of saliency in comparison with a reference corpus. In this case, the reference corpus consists of speeches by four secretaries of education who served in the Bush and Obama presidential administrations, both of which initiated a variety of reforms to expand school choice. Yet the log ratio of choice in the DeVos corpus is 2.22, indicating that its relative frequency in her speeches is more than 4 times greater than that in the speeches of her predecessors.
As noted above, choice in DeVos's speeches is linked to the keyword freedom, but it is also linked with a number of other keywords-whether by semantic relation, proximity of occurrence, or both. These keywords include, for example, individual, options, unique, assigned, fit/s, closest, serve, choose, empower, and buildings. It is worth quoting at some length a speech from 2017 that is representative of the vision for school choice that DeVos communicates throughout the corpus: The future of choice lies in trusting and empowering parents-all parents, not just those who have the power, prestige or financial wherewithal to make choices.
No more "choice for me, but not for thee," from politicians in Washington, or in state houses.
The future of choice lies in caring less about the word that comes before "school" and more about the individual students that "school" seeks to serve.
The future of choice lies in funding and supporting individual students, not systems or buildings.
The future of choice lies in allowing students to progress at their own pace, to take charge of their learning, in recognizing them as the unique individuals they are.
The future of choice lies in embracing learning that fosters creativity, communication, collaboration and critical thinking-traits that prepare students for further education or the workforce, and for lifelong learning.
The future of choice lies in recognizing America-the greatest country in the history of mankind-can, and must, do better for our students-all of them. Because we must do better for our future. (September 28, 2017).
Similar to her use of freedom, DeVos offers a rather expansive concept of choice here. For DeVos, it seems, choice becomes an empty signifier that can include empowering parents, eliminating the control of the Washington establishment, dissociating schools from notions of publicness, seeing students as individuals and personalizing their education, embracing the skills necessary for students' future careers or further education, and even recognizing the failure of the nation to serve its students well. There is nevertheless an underlying common denominator in these significations, and it consists in DeVos's rejection of institutions and what she sees as their tendency to control the learning of students and constrain the choices of parents. The solution, for DeVos, is to empower the individual.
Indeed, the rejection of buildings and systems in addition to the Washington political establishment renders this anti-institutionalism clear. In an earlier speech, DeVos offers an even more explicit rejection of institutions as she explains what she means by choice: I am in favor of increased choice, but I'm not in favor of any one form of choice over another. I'm simply in favor of giving parents more and better options to find an environment that will set their child up for success.
I'm opposed to any parents feeling trapped or, worse yet, feeling that they can't offer their child the education they wish they could. It shouldn't matter what type of school a student attends, so long as the school is the right fit for that student.
Our nation's commitment is to provide a quality education to every child to serve the public, common good. Accordingly, we must shift the paradigm to think of education funding as investments made in individual children, not in institutions or buildings.
Let me say it again: we must change the way we think about funding education and instead invest in children, not in buildings. (March 29, 2017) In the reference to parents' "feeling trapped," there is an implicit connection between choice and freedom. Yet DeVos also sets up a dichotomous relationship between institutions (and the buildings that metaphorically represent them) and the needs of individual children, leading to her claim that education funding should be radically overhauled to prioritize individual choice. The suggestion here is that policy can meet the needs of institutions or of individual students; it cannot serve both.
In a 2018 speech, DeVos distinguishes her notion of choice from that of the two preceding administrations: "Choice in education is not limited to a student picking this building or that school-using this voucher or that scholarship. And it's not public versus private. Parochial versus charter. Homeschool versus virtual. Choice in education is bigger than that" (May 16, 2018). "Choice," DeVos adds, "is about freedom." Here, DeVos works to disengage choice from institutional language-whether it is the discourse of specific government policies (e.g., "vouchers") or the discourse of school types (e.g., "public," "parochial," "charter"). The suggestion here is that institutional notions of choice are far too limiting and therefore hostile to freedom.
For DeVos, students and parents are not the only ones whose choices are limited by institutions. Addressing a group of teachers in 2019, DeVos notes that she meets regularly with "excellent teachers like you who relate how their assigned professional development days are a 'waste' of their time. They long to choose their own professional development and make it fit their needs" (April 29, 2019). DeVos's solution to this lack of choice involves another call to defund institutions: "…we've asked that Congress invest in what you find useful for yourselves, as opposed to what others think is right for you. Our proposed 'teacher vouchers' recognize you as the professionals you are." DeVos uses a remarkably similar language to describe teacher choice and student choice: Teachers are stuck in "assigned professional development days" and want professional development to "fit their needs" (April 29, 2019); students are stuck in their "assigned neighborhood school" while their parents want to find "the right fit for them" (November 30, 2017). Furthermore, DeVos suggests that teachers, like students, would benefit from "vouchers" that would allow them to choose what works for them. Through similarity of word choice and policy solutions, DeVos applies the sort of "logic of equivalence" that is typical of populist discourse (Laclau, 2005). The demands of teachers, students, and parents may be different on the surface, but they are fundamentally equivalent in their opposition to institutions that purport to know what is best for them.
The rejection of institutions is thus related to another common feature of populist discourse: a distrust of experts, of those who claim to know what is best for individuals. Elsewhere in the corpus, DeVos dismisses experts overtly. In a 2018 speech about the future of choice, for example, DeVos notes that while the Bush and Obama administrations may have had their differences, they shared a single "false premise: that Washington knows what's best for educators, parents and students" (January 16, 2018). DeVos adds, "when it comes to education… those closest to the problem are always better able to solve it. Washington bureaucrats and self-styled education "experts" are about as far removed from students as you can get." Here, in dividing those who are "closest" to the educational process from those "far removed" people who consider themselves "experts," DeVos deploys the populist trope of valuing "first-hand experience over abstract and experience-distant forms of knowledge" (Brubaker, 2017, p. 367).
Thus far, I have discussed DeVos's use of keywords that demonstrate her anti-institutional orientation. But sometimes a corpus of texts is remarkable not only for the types of words it includes, but also for what it tends to exclude (Brezina, 2018). As noted earlier, a negative keyword is a word that occurs with unexpected infrequency in a corpus, given its relatively high frequency in a reference corpus of similar types of texts. Table 4 below provides a list of the 41 negative keywords in the DeVos corpus, in order of negative keyness-words that are highly salient in the speeches of Bush and Obama's secretaries of education but rarely used by DeVos. Most notably, 26 of these words might be categorized as "institutional" vocabulary (these are bolded in the table). Of these, 12 have to do with schools, school systems, or subjects: high (used in the reference corpus as part of the cluster, "high school"), college, early (used in the reference corpus as part of the clusters "early education" and "early childhood education"), teaching, districts, math, reading, principals, science, instruction, academic, and knowledge. Fourteen of the negative keywords have to do with neoliberal reforms and the New Public Management themes of standards, performativity, and monitoring quality and progress : standards, accountability, achievement, gap, progress, research, increase, level, reform, assessments, performance, performing, scores, and expectations. DeVos's avoidance of these words, I want to suggest, provides further indication of her rejection of institutions and the institutional common sense about school reform.

Table 4
DeVos Corpus Negative Keywords Negative keywords high, college, behind, standards, president, left, accountability, early, achievement, gap, teaching, progress, research, million, districts, math, reading, increase, level, programs, billion, language, reform, principals, science, technology, instruction, assessments, resources, performance, performing, president's, academic, grants, poverty, scores, knowledge, expectations, Texas, economic, build Note. Keywords relating to schools as institutions and keywords relating to themes of New Public Management appear in bold type.

A Populist Logic
Whereas freedom, choice, and their related keywords revealed the salient ideological content of DeVos's speeches, the remaining categories of keywords tell us more about DeVos's political logic and rhetorical moves. These keywords tended to be function words (e.g., should, nor), but they also included a few lexical words (e.g., rethink, embrace/d) that turned out to be interesting not for their own ideological implications, but for their rhetorical or logical functions-namely, DeVos's tendency to make normative arguments, to set up dichotomous relations between her views and those of her opponents or predecessors, and to rely on imprecise turns of phrase. Indeed, these functions are remarkably consistent with the characteristics of populist discourses (Laclau, 2005;Schmidt, 2017). I have already called attention to some dichotomies and examples of vague and imprecise language in the preceding discussion, but in what follows, I will present an analysis of keywords that are related more directly to these rhetorical features.

Normative Argument
It is noteworthy that the third highest ranked keyword in the DeVos corpus, should, is a function word. Although lexical keywords tell us about the salient content of a corpus, function keywords can reveal the distinctive style of a corpus (Baker, 2006). The prevalence of should throughout the DeVos corpus suggests her tendency to offer normative arguments. As noted above, whereas cognitive arguments make use of expert knowledge and draw relationships of cause and effect, normative arguments tend to be based on values or moralization (Schmidt, 2017). Schmidt (2017) also notes that populist discourses such as Trump's rely heavily on normative arguments because they resonate more with people who reject the language of technocratic debate.
Should is generally used in English to express obligation or necessity, sometimes to make moral/evaluative arguments about what ought to be the case or what is desirable-what in linguistics is called deontic modality (Fairclough, 2003). Used 318 times in the DeVos corpus, should appears 132 times in the cluster "should be" and 45 times in the cluster "we should," revealing DeVos's pattern of making explicit statements about the way things ought to be or what "we" (people in general) ought to be doing. In a speech from 2019, for example, DeVos tells parents about her vision for education freedom: "You should be free to make the decisions that work best for your children and your family" (September 16, 2019). DeVos also uses the cluster "should be" to make larger statements about how the notion of public education ought to be understood: "if every student is part of 'the public,' then every way and every place a student learns is ultimately of benefit to 'the public.' That should be the new definition of public education" (May 17, 2019).
When using the cluster "we should," DeVos is often making explicit statements about what "we" (people in general) ought to value. For example, in a speech advocating the expansion of school choice options, DeVos combines should with another keyword, embrace, to make a more general point about what we should desire: "we should embrace the promise of new solutions that challenge the way things currently are" (May 22, 2017). Fairclough (2003, p. 173) notes that deontic modal verbs like should are "relatively transparent markers of evaluation," and this is rather clear in DeVos's use of the word in arguments about what is good or right. After sharing the personal stories of students who have benefited from forms of educational choice, DeVos claims, "All of these students' experiences suggest that we should value and appreciate the individual trees rather than see a monolithic forest" (March 20, 2017). In this case, we have an overt statement about what we ought to value: the individual, not the collective.

Dichotomization
One of the more striking features of DeVos's discourse is a tendency to describe policy preferences in terms of how they are diametrically opposed to the policies of previous administrations or of those who disagree with her. As Laclau (2005) explains, one of the fundamental characteristics of populist discourse and logic is its use of dichotomies-particularly between "the people," on the one hand, and the institutional establishment, elites, or experts, on the other. DeVos often constructs such dichotomies through negation, as indicated by the prevalence of keywords such as nor and isn't. Of course, function words like these can be used in a wide variety of contexts; hence the importance of concordance analyses to determine how they tend to be used in the DeVos corpus. The results of these analyses revealed that 76% of the occurrences of nor and 51% of the occurrences of isn't served the purpose of introducing dichotomies between DeVos's policy preferences and those of the institutional establishment.
When using nor, DeVos tends to list two or more characteristics of the institutional approach to describe what is wrong with it. For example, in a 2019 speech, DeVos says that "an education that works for each student should not be determined by chance nor by government. Education-how and where students learn-should be determined by students and their families" (November 13, 2019). Here, the institutional approach is one in which lotteries or accidents of birth in certain cities or neighborhoods, or simply government assignments, determine where a student should go to school; DeVos then contrasts this with her approach to determining the where and how of a child's education, one that empowers students and families to make these decisions.
Similarly, DeVos uses the word isn't to distinguish her policy approach from an institutional approach: "This isn't about school 'systems.' This is about individual students, parents, and families. Schools are at the service of students, not the other way around" (July 20, 2017). In examples like this one, DeVos presents a dichotomous relationship between institutions ("systems") and individuals in much the same way as she did while describing her vision of choice, as discussed above. It is important to note that DeVos could have described her policy preferences without using such dichotomies. After all, she might have simply described a preference for meeting the needs of individuals. But as the keywords indicate, her tendency was to describe her policy preferences in terms of their antagonistic relationship with the institutional order.
DeVos also uses the word approach, a lexical word, to construct an antagonistic relationship between her policy agenda and those of the establishment. In 73% of the occurrences of approach, DeVos uses the word to set up a dichotomy between her policies and institutional policies.
Sometimes her references to the establishment are rather explicit, as indicated by the use of scare quotes in the transcript of a speech from 2018: "Let's question everything about the "established" approach to education to ensure what we do makes sense for the 21st century and that it actually works for students" (October 3, 2018). Calling attention to the word "established," DeVos adopts a populist logic that draws an antagonistic frontier between the establishment and those who believe students should be the focus of policy. In the following example, DeVos uses approach to introduce a stark dichotomy between current policy and her own vision for education: "Think for a minute about this: what descriptors sound more like our current approach to education? Adaptable, nimble, dynamic. Or structure, conformity, compliance? I think we all know the unfortunate reality, and it's leaving students unprepared" (January 25, 2018). In presenting the basic differences between her vision and current policy, DeVos juxtaposes three adjectives suggesting movement and change with three nouns suggesting immobility and lack of freedom.
In these dichotomies we have another mark of populist discourse: What is essential about DeVos's ideal vision is not merely its own positive content (freedom, choice), but its difference from, its incompatibility with, the establishment and status quo. Furthermore, the dichotomy between current and ideal are presented in rather general terms. On the one hand, we have "government," "'systems'," "the 'established' approach," and "structure, conformity, compliance"; on the other hand, we have "individual students, parents, and families," what "actually works for students," and what is "adaptable, nimble, dynamic." For Laclau (2005, p. 18), the dichotomies of populist discourses produce a simplification of the political space (all social singularities tend to group themselves around one or the other of the poles of the dichotomy), and the terms designating both poles have necessarily to be imprecise (otherwise they could not cover all the particularities that they are supposed to regroup).
Thus, the vagueness of the two poles of the dichotomy, much like the vagueness of the empty signifiers of freedom and choice, is crucial in that it can produce widespread appeal. The poles cannot be described as specificities, such as achievement gaps and proficiency in math and reading; rather, they must speak to certain general values with which political subjects can identify.

Imprecision
I have already discussed the functions of vague and imprecise language in the preceding sections, but the keywords in this final category are notable for their inherent indeterminacy. The salience of everything, something, and anyone-all indefinite pronouns-in the corpus suggests DeVos's tendency to offer general propositions or vague descriptions of a vision for education. The keyword something, for example, occurs 98 times in the corpus, and its log ratio of 2.01 indicates a relative frequency that is more than four times greater than that in the reference corpus. In 53% of the occurrences of something, it is used in a variety of phrases describing DeVos's vision for what is needed in education policy: "something bold," "something better," "something different," "something new." For example, in a passage similar to one quoted earlier, DeVos claims that it is "past time for something different. It is past time for something better in education. The thing that makes America great: freedom" (November 13, 2019). Similar to the use of dichotomies, something is used to introduce a policy vision that is distinctive not simply for its positive content, but for its difference from the old. Furthermore, as an indefinite pronoun, the word something adds another layer of imprecision to DeVos's characteristically vague descriptions of freedom and her appropriation of Trump's vague slogan, "Make America Great Again." This imprecision is also evident in the use of everything, a keyword that appears 82 times in the corpus, 21% of the time as part of the cluster, "question everything." For example, DeVos uses everything as part of her definition of rethink, which ranks fifth on the list of keywords: "'Rethink' means this: everyone question everything to ensure nothing limits students from being prepared for what comes next" (January 28, 2019). DeVos then provides her definition of everything: "And what is 'everything'? Everything is everything. Where, when, how, what, and why we do things today-and everything about what we could or should do differently." Here, we see how DeVos uses everything to make a call for radical change, for questioning all that we do-in essence, for changing the way we think about education. DeVos's tautological definition of everything ensures that she won't be tied to any specific policy item, only to a general overturning of the institutional status quo. Again, what is most prominent is the strong populist sentiment of rejecting the establishment.
The corollary of this anti-institutionalism, of course, is a call for empowering "the people." In the following example, DeVos's use of all three indefinite pronouns from the list of keywords enables her to make a general, sweeping statement about the status of "the people" in the U.S. Constitution: …if the Constitution does not explicitly permit the federal government to do something, it is supposed to be left to the States. To communities. To families.
Left to 'We the People.' Everything anyone needs to know about Americaabout who we are-is contained in those three little words.
'We the People' govern ourselves. We know what's best for ourselves, and for our own children. (December 5, 2019) As noted above, there is a certain "simplification of the political space" (Laclau, 2005, p. 18) that is characteristic of populist discourse. In this case, the simplification consists not only of the dichotomous relationship between the federal government and "the people," but also in vaguely defining the phrase, "We the People," as "everything anyone needs to know about America." America is the people, and the people know what is best for themselves.

Conclusion: A Populist Neoliberalism?
This corpus-driven analysis suggests that Betsy DeVos's education policy speeches represent a hybrid of neoliberal and populist discourses. Two of the highest-ranked keywords, freedom and choice, invoke basic values of neoliberalism: free markets, individual freedoms, freedom from government, and free choice among a set of competing alternatives (Harvey, 2005). Indeed, DeVos spends a great deal of time advocating more freedom, specifically from institutions that stand between individuals and their ability to choose what they believe is right for themselves. In describing her vision for the future of school choice, DeVos uses freedom to promote the neoliberal idea that individuals should go forth and find for themselves the sort of learning, growth, advancement-and ultimately happiness-that is best for them (Adams et al., 2019). This concept of freedom-of being an enterprise of oneself, of developing one's own human capital (Foucault, 2008)-is characteristically neoliberal in the responsibility it removes from the state, as provider, and heaps on the individual, as seeker.
Yet throughout the corpus, the calls for freedom and new forms of choice are embedded in a discourse that is also populistic in content and strategy. Freedom and choice, as empty signifiers, can be applied to an expansive series of demands, the common denominator of which is a demand for freedom from institutions, a notion of choice that is disengaged from any institutional arrangements. An anti-institutional orientation runs throughout the corpus, as DeVos draws an antagonistic frontier between the elite establishment and "the people"-that is, between those who claim to be experts (those who purport to know what kind of schooling is best for each individual) and those who are closest to the educational process. For DeVos, "the people" seem to consist of parents, students, and even teachers. Using a logic of equivalence typical in populist discourse (Laclau, 2005), DeVos suggests that what these groups have in common is their subjection to institutional arrangements that constrain them. While teachers want freedom from assigned professional development days and from others delineating what they must teach and how, parents and students have had enough of government experts telling them where and how they should be educated.
The rhetorical strategies of populist discourse are also evident across the texts. Keywords such as should, rethink, and embrace indicate DeVos's reliance on normative arguments in her policy discourse, while negation words and contractions such as nor and isn't tend to occur in dichotomous constructions pitting DeVos's policy approaches against those of the establishment. Furthermore, the salience of indefinite pronouns such as something and everything reveal a pattern of imprecise, sweeping statements of belief about abstract concepts such as freedom and the US Constitutional notion of government by "the people." The co-occurrence of neoliberal and populist ideas and discourse in DeVos's speeches may suggest a moment of ideological shift in education policy discourse, a disavowal of the neoliberal consensus of the last few decades concerning results-based accountability, tougher evaluations of teachers, high-stakes testing to close achievement gaps, and a notion of school choice centered largely on charter schools and public school options. DeVos's populist discourse can thus be interpreted as a reaction to a neoliberal order that, instead of limiting state power, ironically has concocted the expanding technocratic governing regime of New Public Management. In this view, DeVos's deployment of the neoliberal watchwords of freedom and choice may suggest that she functions as a sort of bridge between neoliberalism and a populist turn in education policy discourse, whether she is conscious of this or not.
Populism, however, is not an ideology; it is a "mode of political practice" that can be deployed by progressive as well as conservative movements and parties (Bonikowski & Gidron, 2016, p. 1597. Thus, while a populist turn may signify a shift from neoliberal hegemony, this does not tell us what we may be shifting toward. One possibility is that DeVos's hybrid of neoliberal and populist discourses signals only a call for a more pure or fundamentalist version of neoliberalism, one that does away with the institutional baggage of New Public Management and, instead, seeks a full privatization of education. In this case, the state's role would perhaps be reduced to dispensing tax dollars (e.g., vouchers) for parents to use as they wish-for private or religious schooling, home schooling, virtual schooling, or some combination thereof. This would be a neoliberalism shorn of institutions that ostensibly attempt to correct the failures and inequities of free markets, a return to the "libertarian promise that the market can displace the state" (Kiely, 2020, p. 404). In this view, DeVos's discourse suggests a return to Milton Friedman and Rose Friedman's (1980) vision of fully marketized education, one that dismantles public school systems and even compulsory education laws.
What is important, I want to suggest, is that a populist education discourse may signal a turning point. To return to Laclau (2005), populist politics arise when a series of political demands remain unfulfilled by the current institutional regime and, moreover, when it becomes clear that they cannot be addressed by the current regime. Indeed, there is considerable evidence that neoliberal hegemony in education policy has been losing legitimacy in recent years. Parents have resisted accountability regimes by opting their children out of standardized testing; teachers' unions in both predominantly Republican and Democratic states have called strikes to oppose fiscal austerity, choice programs, and pay-for-performance systems, often with support from parents and communities; and while Republicans still largely support school choice in the form of charter schools and sometimes vouchers, several 2020 Democratic presidential primary candidates' opposition to the Obama-era expansion of charter schools suggested growing fractures in the neoliberal consensus on school choice (Meckler, 2019;Strauss, 2016;Turkewitz & Goldstein, 2019;Turner et al., 2018).
But if we are at a turning point, what will come next? This question is particularly relevant at the time of this writing, shortly after the inauguration of President Joe Biden. The COVID-19 pandemic introduced new financial struggles to public education systems that were already vastly under-resourced, and Biden's administration and Democratic majorities in Congress have pushed for legislation that would fund extended learning opportunities for students from low-income families, protect educators and other school staff from layoffs, and upgrade school buildings in impoverished communities (Ujifusa, 2021). Support for such relief represents a stark contrast with the policies of the preceding administration and Secretary DeVos's anti-institutionalist vision. However, on other controversial issues such as school choice, President Biden and his nominee for secretary of education, Dr. Miguel Cardona, have been far less committal (Darville et al., 2021). In the coming years, will Biden and Cardona carry forward the kinds of policies enacted by the Obama-Biden administration during the Great Recession, when states and school districts had to enact a variety of neoliberal reforms in order to compete for much-needed funding (e.g., in the Race to the Top grant program)? And, to look further down the road, will future Republican presidents attempt to revive the sort of anti-institutionalist politics that DeVos supported during her tenure?
As discussed earlier in this paper, a number of scholars suggest that the rise of populist politics worldwide indicates that neoliberalism may be past its prime (Fraser, 2017;Kiely, 2020;Rose, 2017). Yet the road ahead remains unclear. Streeck (2017) uses the Gramscian term, "interregnum," to describe the current crisis of neoliberal hegemony in political economy, a time of uncertainty between a dying hegemonic order and one that is yet to be determined. Such a time is likely to include a battle of ideas concerning what should come next, "competing projects circulating with considerable backing from powerful forces in society" (Stahl, 2019, p. 344). In the arena of education policy, DeVos's discourse reveals one competing project, and as the findings of this study suggest, the very definition of a public education is at stake. If we are, in fact, in the midst of a turning point, it will be important for scholars of education policy, or, really, for anyone concerned about the future of public educational institutions, to understand the various forces seeking to fill the hegemonic void.

About the Author Michael Ian Cohen
University of Northern Colorado michael.cohen@unco.edu ORCID https://orcid.org/0000-0002-1288-4718 Michael Ian Cohen is an Associate Professor in the Department of Leadership, Policy, and Development, College of Education and Behavioral Sciences, University of Northern Colorado. His research focuses on neoliberal education reforms, New Public Management, and educator professionalism. He has served as a public high school teacher and an administrator at the school and district levels in New Jersey and Colorado. education policy analysis archives Volume 29 Number 16 February 15, 2021ISSN 1068-2341 Readers are free to copy, display, distribute, and adapt this article, as long as the work is attributed to the author(s) and Education Policy Analysis Archives, the changes are identified, and the same license applies to the derivative work. More details of this Creative Commons license are available at https://creativecommons. Please send errata notes to Audrey Amrein-Beardsley at audrey.beardsley@asu.edu Join EPAA's Facebook community at https://www.facebook.com/EPAAAAPE and Twitter feed @epaa_aape.