Classic,Segmented-,or Neo-Assimilation,Which Theory to Use? A Scientific-Method Investigation

Deductive and Inductive Hypothesizing

In the process of knowledge production, a deductive hypothesis uses an already-established theory to contemplate and predict the emergence of new outcomes. Simply, if C₁+ C₂+ … = E₁ represents the hypothesis, a deductive hypothesis has the cause C₁ and C₂ variables and searches for an unknown effect E₁ or outcome. For instance, assume a comprehensive theory on the relationship between smoking and cancer. A curious researcher can then use this theory to proffer a deductive hypothesis on the relation between smoking with other unobserved health outcomes such as infertility. In sociology, this type of deductive hypothesizing remains marginal (but see Kiser and Hechter 1991). ⁴ The closest is the Weberian-Simmelian ideal-type method (see Abend 2019 on “thick concepts”). As with deductive hypothesizing, ideal-type hypothesizing seeks to turn the “unstructured” empirical world into “structured” concepts with a priori boundaries and representations (Bradburn, Cartwright and Fuller 2017).

The difference between pure deductive and ideal-type hypothesizing is that, while the former emphasizes direct deduction from a previous theory, the latter combines deduction with broad empirical observations (Hempel 1965; Chapter 7). This is achieved through “constructing provisional ‘transitional’ types and ‘combinations’ of them” (Lazarsfeld and Oberschall 1965, p. 191). These ideal-type or pure concepts are abstract “ideal” images which are not found in the “real” empirical world (Weber 1978). Scholars use these ideal types to deduce and test hypotheses about the empirical world. Deductive and ideal-type hypothesizing does not imply that observations and empirical experiments are dispensable, but that theories give priority to empirical observations since it is the theory that makes sense of any observations and experiments. Throughout this process, concepts and theories come as close as possible to explaining the empirical world. ⁵

In contrast, the inductive method starts with immersion in, and/or the collection of empirical data. A pattern is observed within a context. Next, there is an attempt to find the causes or predictors of the pattern. Simply, if C₁+ C₂ + … = E₁ represents the hypothesis, an inductive hypothesis has the outcome E₁ and hypothesizes about the C₁ and C₂ variables. For instance, a researcher working with a particular population observes the emergence and rise of infertility rates in the said population. The researcher can then use this observation to formulate an inductive hypothesis about the potential causes of the observed infertility outcome — this outcome has a priori observable boundaries. To help identify the predictors, awareness of previous studies and theories is a must (Popper 2002 [1956], p. 124; Burawoy 2019; see Glaser and Strauss 1967 on grounded-theory method).

Once researchers hypothesize the predictors, they compile more data over time. More in-depth and longitudinal data play the role of internal empirical tests to weed out the impertinent predictors. If further observations support the hypothesized predictors, the researchers can present their findings to the scientific community for further external tests. Overall, the inductive method entails a continuous process of finding the correlation between outcomes and predictors (Timmermans and Tavory 2007, p. 8). In sociology, the works of Durkheim (1912 [1961]) on religion and Bourdieu (1984) on class and classification show that an in-depth case study approach suffices for concept and theory development insofar as it “accounts for much of the relevant behavior” (Glaser and Strauss 1967, p. 30). ⁶

Falsification

Inductive and deductive methods offer hypotheses but a hypothesis “cannot be asserted and then assumed as law” or theory as if its validity is a given (Dooley and Goodison 2020, p. 26). To be accepted as a theory by the scientific community, a hypothesis must be capable of “being tested by a confrontation with experimental findings” so to distinguish the empirically and logically valid ideas from prophecies (Hempel 1965, p. 3). As such, scientific tests’ goal is to falsify the correlation that a hypothesis has suggested. The difference is that the tests falsify a deductive hypothesis’ predicted outcome and falsify an inductive hypothesis’ proposed predictors. This is why it is essential to understand how a theory was initially developed. The knowledge about the “how,” or the deductive versus inductive method, helps researchers identify whether their tests falsify a hypothesized outcome, the impact of a predictor, or both. Researchers outline their hypotheses, including all the concepts and methods, a priori or before data collection and tests.

Researchers break down the correlation into four testable statements to test the hypothesized correlation. For instance, classic assimilation hypothesizes a correlation between language skills and making outgroup ties or language ↔ outgroup ties (Gordon 1964). Accordingly, one can hypothesize four possibilities: if language skills exist, outgroup ties exist (H₁ test hypothesis is what the theory predicts); if language skills exist, outgroup ties cease or if language skills cease, outgroup ties exist (H₀ null); if language skills cease, outgroup ties cease (neither). Evidence for the first hypothesis would corroborate the theory, evidence for the nulls would reject it, and evidence for the last does not say much about the theory.

When a hypothesis withstands empirical tests, the said hypothesis is corroborated but not verified indefinitely. Deductive or inductive, “all theories are trials; they are tentative hypotheses, tried out to see whether they work; and all experimental corroboration is simply the result of tests undertaken in a critical spirit, in an attempt to find out where our theories err” (Popper 2002 [1956], p. 80). With each passing test the assertive power of a theory, increases but never becomes bulletproof against falsification. The next test may falsify the hypothesis and theory. Therefore, there is an asymmetrical relation between corroboration and falsification. As such, falsification is different from statistical probability or confirmation models such as the Bayesian approach, which takes the “[statistical] degree of corroboration, or of confirmation, or of acceptability, with probability” of the hypothesis (Popper 2011 [1945], p. 408). ⁷ Instead, falsification requires knowing “what tests, what trials, it [hypothesis] has withstood” before the conclusion is accepted (ibid, 248). ⁸

To summarize, Figure 1 outlines the scientific methods of hypothesizing and falsifying.

OPEN IN VIEWER

Figure 1.

Scientific method of hypothesizing-falsification.

As Figure 1 shows, the scientific methods of hypothesizing and testing entail a process of offering hypotheses from theory or data and, through falsification, selecting the best hypothesis. The scientific community then admits this hypothesis as theory until a better, more accurate hypothesis (and theory) emerges. In this process, the function of scientific tests is to falsify and limit the range of hypothesized outcomes and predictors. A hypothesis and, later, a theory that explains any and all outcomes and predictors is not scientific. The improbability of falsification “is not a virtue of a theory (as people often think), but a vice” (Popper 2014 [1965], p. 48). This is because, without the possibility of falsifiability and, hence, replaceability, researchers cannot offer new ideas to move the knowledge production process forward. Simply, unfalsifiability means scientific stagnancy.

Classic Assimilation

Deductive, but more precisely, ideal-type hypothesizing, à la Simmel and Weber, informs classic assimilation theory's explanations of group relations (Park 1915; Gordon 1964, p. 69; see Kemple 2019 and Kivisto 2022 on Park's intellectual relation to Simmel and Weber). Classic assimilation's focus is on hypothesizing the outcome of assimilation. Park (1915, p. 578) imagines group relations as unfolding within an ideal-type institution, “a psychophysical mechanism … through which private and political interests find corporate expression”. This abstract institution, when defined as a nation, encompasses social groups or “national types which exhibit definite [linguistic and moral] uniformities” (Park 1914, p. 606). Having conceptualized the context of group relations as such, Park contemplated the outcome of such relations. He borrowed from other sciences, such as the theories of physiology, to outline how people interact and potentially become part of a community or an entity larger than the total sum of the individuals (ibid, p. 611). He then defined assimilation as a “symbiosis” process where “peoples and races who live together, sharing in the same economy, inevitably interbreed, and in this way if in no other, the relations which were merely cooperative and economic become social and cultural” (ibid). Park (1928b, p. 891) hypothesizes that, in the long run, when national types meet, they achieve a degree of cultural pluralism and solidarity sufficient for national existence.

Gordon (1964, pp. 69–71) slightly updated the anticipated outcome as ethnoracial plurality within the national culture. He then hypothesized that the assimilation process takes place in a series of seven steps, each of which, in sequential order, can be considered an independent cause or predictor for ethnoracial pluralism outcome. He takes several precautions. He distinguishes between the “cultural patterns” of a society —language, law, political associations, etc. — and different groups’ “overwhelmingly favorable … contributions made by way of cultural patterns” (ibid, p. 73 italics in original). Then, he argues that, in the United States of his time, these institutionalized cultural patterns closely resembled those of the “middle-class White Protestant Americans” (Gordon 1964, p. 74). Such patterns differ across societies according to their local norms and institutions — hence Favell's (2016, 2022) critique that contemporary researchers pay insufficient attention to the context of assimilation research.

In classic assimilation, the first step, acculturation, or behavioral assimilation, entails adapting norms such as language, workplace etiquette, respect for the rule of law, and individual liberties. Next, structural assimilation — as opposed to economic structures — implies that, with acculturation, individuals can go beyond their immediate family and ethnic group to make friendships and social relations with other groups. These relations lead to the pivotal element of intergroup marriages and, therefore, to the decline of exclusive ethnic affiliations. As a result, in the fourth step, the spread of group relations leads to wholesale identification as members of a national society. This sense of peoplehood has a snowball effect on the assimilation process. From here on, majority groups recognize immigrants’ assimilation attempts and gradually offer attitudinal, behavioral, and civic inclusion. Racial prejudice declines and, consequently, de jure discrimination is hypothesized to subside (see Abascal 2020 on majority groups’ response). The process unfolds over a timeframe of two or more generations. Gordon (1964, p. 71) qualifies this ideal-type model, “not only is the assimilation process mainly a matter of degree [for each group], but, obviously, each of the stages or subprocesses distinguished above may take place in varying degrees.” That, for instance, means “acculturation of the minority group may take place even when none of the other types of assimilation occurs simultaneously or later, and this condition of ‘acculturation only’ may continue indefinitely” (ibid, p. 77; see Telles and Sue 2019 on durable ethnicity, and Williams and Ortega 1990 on structural assimilation as the starting point of assimilation).

Classic theory delimits the context of such an ideal-type assimilation process according to the specific national contexts and subgroup conditions. Regarding the former, Park and Gordon are aware of the practical limitations of their ideal-type models, particularly the ethnic and racial barriers to coexistence as a nation. Park (1928a, p. 20) observes that industrialization and the development of professional echelons within various racial groups close racial distances and lead to the development of “biracial organizations” across the country. He (1914, p. 611) readily admits that racial markers put “between the races the invisible but very real gulf of self-consciousness” and limit the possibilities of becoming “a mere individual, indistinguishable in the cosmopolitan mass of the population.” Similarly, Gordon (1964, p. 60) expands on the cultural and institutional conditions of the “meeting of peoples” in the United States, “the attempted incorporation of the American Indian on the White conqueror's terms, and the massive immigration into this land of over 41 million people, largely from Europe but also from the other Americas and to a smaller extent from the Orient.” Again, he underlines that despite the “modern industrialism and urbanization and operating under a democratic political system” (ibid, p. 18), “there is no good reason to believe that white Protestant America has ever extended a firm and cordial invitation to its minorities” (Gordon 1961, p. 281).

To explain the role of subgroups, classic assimilation uses the concept of “ethclass,” the “social space created by the intersection of the ethnic group with social class” (Gordon 1964, p. 51; see also Warner and Srole 1945, p. 71). This ideal-type concept resembles Weber's (1978, p. 155) Stände or status group. In contrast to the Marxian horizontal economic class divides — i.e., ethnicity within the class — the Stände define themselves and others in terms of social estimation of honor, the combination of material and ideal values — i.e., class within ethnicity. The result is the economically unequal “societies who live along the lines of ‘ethnicity’” with each ethnicity comprising different social classes (ibid, p. 157; see also Du Bois 2015, p. 35). Gordon (1964, pp. 44–45) argues that, in the United States of his time, there were six social classes ranked from the “old family” aristocrats, to mid-level college-educated professionals and semi-skilled workers, to lower-level “unskilled laborers and the frequently unemployed.” The mid-level social classes, due to demographic and institutional prominence, shaped the core culture. The intersection of social class lines with ethnic boundaries creates the “ethclass.” Individuals can gain entry to different levels of social class within their ethnicity, but assimilation depends primarily on crossing ethnic lines. Gordon argues that class upward mobility does not necessarily entail leaving and crossing the primary ethnic group boundaries (ibid, pp. 46–47). If upwardly mobile individuals leave their ethnic groups, they may risk alienation and loss of social capital (Gordon 1961, p. 280).

According to these theoretical and contextual observations, classic assimilation theorists hypothesized several outcomes for the future of assimilation in the United States, including the biracial organization of society (Park 1914, 1928). Park had observed first-hand the rapid social transformations in Europe. At the time, the old multiethnic empires were falling apart, giving way to ethnic nation-states. With this in mind, and observing the tensions between whites and African Americans, Park (1914, p. 628) hypothesized the possibility of a federalist system “based upon the autonomy of the different races; a bi-racial organization of society.” Over a decade later, however, he observed that racial distances had somewhat dwindled due to economic prosperity and industrialization. He then predicted that there would be a time when racial groups will have equal footing in society, “the distances which separate the races are maintained, but the attitudes involved are different. The races no longer look up and down: they look across” (Park 1928a, p. 20).

Later, Gordon (1981, p. 182) hypothesized that assimilation could lead to a monolingual ethnically plural society, a bilingual pluralism such as Canada, or a multilingual pluralism such as Switzerland. Gordon observed that intermarriage and identification with Anglo-Saxons did not follow what the Anglo-conformists had speculated in the 20th century and interpreted this group pluralism as a sign that “Anglo-conformity and the melting pot” ideas were misleading (1961, p. 283). ⁹ He emphasized that monolingual ethnic pluralism would be the future of American life (Gordon 1964, p. 114). He (1981, p. 181) believed that, under American democratic values, different national groups could maintain a healthy degree of group separation while creating shared cultural patterns sufficient for national coexistence. For group pluralism not to lead to inequality, Gordon (1964, p. 240) underlined the role of the intellectuals and policymakers who must legitimize such plurality “in the public consciousness.”

To test the classic assimilation theory, researchers have used empirical observations and experiments to falsify the hypothesized correlations in the ideal-type model (inter alia see Brown 2006; Telles and Ortiz 2008; Schachter 2016). Some studies corroborate the theory, and some challenge the theory's optimism about intergroup relations. Indeed, in hindsight, it is clear that the assimilation process in the United States has not led to a biracial organization of society or to a Canadian-style pluralism. These two hypothesized outcomes have been empirically falsified. But classic assimilation's prediction about monolingual ethnoracial pluralism is not far from the empirical reality of US life today. Again, such empirical effectiveness does not mean this theory is inherently correct but that, for the time being, it stands as a falsifiable theory that explains US group relations. ¹⁰

Segmented-Assimilation

Inductive hypothesizing partially informed the development of segmented assimilation. To build the hypothesis, scholars observed three segmented assimilation outcomes along class lines: some immigrants join the mainstream upper classes, some the “white middle class,” and some the poverty-ridden “underclass” (Portes and Zhou 1993). ¹¹ The authors, in keeping with the inductive method, ask how and why “the process has become segmented” (ibid, p. 82). To find explanations for these outcomes, segmented-assimilation theorists use “selected ethnographic material and survey data from a recent survey of children of immigrants” (ibid, p. 75). As such, segmented-assimilation takes upon itself to discover the predictors that lead to the segmented outcomes.

To delineate the context of the theory, segmented assimilationists point to the de-industrialized hour-glass economy and racial prejudices in the United States. Portes and Zhou argue that immigrants accept entry-level menial jobs while the second generations shun the same jobs and this “disjuncture” creates a gap between first- and second-generations’ “material conditions and career prospects” (ibid, p. 76). In addition, the “rapid process of national deindustrialization … has left entrants to the American labor force confronting a widening gap between the minimally paid menial jobs that immigrants commonly accept and the high-tech and professional occupations requiring college degrees that native elites occupy” (ibid). Further, the racialized national context differs from the 20th century's environment wherein, for descendants of European immigrants, white “skin color reduced a major barrier to entry into the American mainstream” (ibid; cf. Karimi, Thompson and Bucerius 2023 on the cross-national contexts of racial barriers).

Then, within such a context, one can glean the predictors of segmented outcomes “from the standpoint of immigrant youths”: “The first is color, the second is location, and the third is the absence of mobility ladders” (ibid, p. 83). Later on, segmented-assimilationists refine these a: family or human capital, family structure, and the modes of incorporation. The latter itself comprises another three elements of government policy (positive, neutral, hostile to immigrants), ethnic capital (high, mid-range, low), and racial prejudice in the labor market (positive or negative). Segmented assimilation outcomes emerge it is argued, based on the different ways that these predictors combine and impact different immigrant groups (Portes, Fernandez-Kelly and Haller 2005, p. 1011; Haller, Portes and Lynch 2011, p. 755; see Portes and Rumbaut 2001, p. 283 for a table outlining the predictors, and Appendix 1 here for our adaption of the table).

Assimilation into upper classes results from consonant acculturation when both the first-generation parents and their second-generation offspring adapt to American culture and support one another's upward mobility in the mainstream. Assimilation into the middle classes happens when, through selective acculturation, first- and second-generations adapt to a new culture but retain their ethnic ties for mobility and protection against discrimination. Assimilation into the underclass results from dissonant acculturation whereby second generations adapt too quickly and abandon their ethnic communities. These second generations lose access to ethnic communities and fall prey to discriminatory job markets and inner-city cultures and, thereby, experience downward mobility. They join a “rainbow underclass” culture along with Black and Puerto Rican groups (Portes and Rumbaut 2001, p. 59; see Warner and Srole 1945, p. 147 for a precursor to the segmented outcomes).

In sum, scholars observed three segmented class outcomes and inductively hypothesized how numerous predictors combine to engender these outcomes. Researchers have tested for the correlation between various combinations of the hypothesized predictors and segmented assimilation outcomes. These combined predictors do not correlate with the outcomes (e.g., Luthra, Soehl and Waldinger 2018). One reason is that segmented assimilation deviates from the inductive method of hypothesizing in three ways. These deviations mean that segmented assimilation is not readily testable and is, hence, unfalsifiable.

First, the theory's key hypothesis suffers from conceptual boundary ambiguity (see Bradburn, Cartwright and Fuller 2017 on a priori concepts). Scholars observed three segmented assimilation patterns, but the theory does not define these patterns or outcomes as a set of empirical concepts (Jung 2009, p. 385; see Wacquant 2022 on the problematic roots and use of underclass). Simply, what economic and social criteria demarcate the underclass status? Similarly, concerning the predictors’ meanings, segmented assimilation hypothesizes acculturation as language skills and knowledge of local customs, the “first step towards assimilation” (Portes and Rumbaut 2001, p. 53), yet operationalizes it as language skills (ibid, p. 145). It is not immediately clear what the conceptual difference is between (consonant, selective, and dissonant) acculturation or language skills versus education; is one the outcome of the other, or are they collinear outcomes of other preceding predictors such as modes of incorporation? Further, despite listing several predictors, the hypothesis does not delineate the causal sequence of how each predictor impacts other predictors to arrive at either one of the segmented outcomes. For instance, it is argued that some “minorities are underprivileged in terms of economic rewards,” but instead of outlining how this underprivileged status emerges for each group, it is said that the “causal sequence leading to these outcomes” is immaterial since “regardless of how it comes about” these groups exhibit the same pattern or outcome (Portes, Fernandez-Kelly and Haller 2005, p. 1030; see Turner 2022, p. 184 on intermediate variables that have no effect in the seemingly causal sequences). In addition to conceptual and processual ambiguities, there is no indication in the hypothesis regarding the kinds of empirical evidence on intergenerational (im)mobility that could falsify the hypothesized predictors.

Second, as per the inductive method, a hypothesis outlines the predictors of an observed pattern from the same dataset. However, in segmented assimilation, the source of the hypothesized predictors remains undefined. Canonical publications start with qualitative stories of first- and second-generations coming from the world's four corners (Portes and Rumbaut 2001, 2006). Some nationalities’ stories point to the role of immigration policies, some to generational disconnect, and some to the role of ethnic economies. Ideally, in inductive research, predictors, when extracted from one nationality's assimilation outcome, would initially remain applicable to that specific nationality. However, instead of such a case-by-case analysis, the segmented-assimilation hypothesis blends and presents a bundle of predictors as if one group's experiences (read: predictors) would be readily applicable to other groups. ¹² This means that one cannot tell which predictors impact which groups or from which stories some of these predictors emerged. ¹³

Third, a scientific approach to hypothesizing accepts the falsifying evidence to provide space for the hypotheses with higher explanatory power. Yet, segmented-assimilation researchers do not address the empirical tests, which do not corroborate the hypothesized predictors. For instance, some studies test for a causal link between acculturation types and socioeconomic outcomes and conclude that “types of acculturation do not seem to matter much for socioeconomic outcomes” (Waters et al. 2010, p. 1185). ¹⁴ Others, skipping the role of acculturation types, measure the link between modes-of-incorporation predictors and educational outcomes. For precision, they replace national groups with skin color categories and find a negligible correlation between the predictors and education outcomes — the exception being “darker group-level skin color is associated with … a 0.18 decrease … in years of education” (Luthra, Soehl and Waldinger 2018, p. 27; see also Waldinger and Catron 2016). That these studies do not corroborate the hypothesis is not a surprise, given that the hypothesized predictors of the segmented outcomes do not directly emerge from the same data but are pulled from elsewhere.

Instead, such falsifying evidence is treated as exception cases (see critiques by Sakamoto, Takei and Woo 2012; Sakamoto and Hsu 2020). For instance, despite being destined for downward mobility according to segmented assimilation, Asian Americans demonstrate high rates of socioeconomic mobility (Pew Research Center 2020). Achievements are made into the model minority myth: Asian Americans are upwardly mobile but have had a long history of exclusion and, currently, have to work much harder than the rest, experience discrimination, and remain at the margins of the nation (e.g., Walton and Truong 2022). Further, it is argued that immigrants’ success is not due to the US opportunities, but an outcome of the human capital that they had amassed in their home countries and brought to the United States (e.g., Lee and Sheng 2023). Another case is the upwardly mobile groups who, instead of upper-class mainstream identities, still advocate ethnic identification (e.g., Mora 2014). This evidence is explained as an exception, as a symptom of “reactive ethnicity” which pushes these upper- and middle-classes into the ethnic communities (Portes and Rumbaut 2001, p. 148). Instead of abandoning its hypotheses of upward or downward mobility when faced with adverse evidence, segmented assimilation creates a chain of exceptions and has, thereby, become an unfalsifiable theory that wants to explain all outcomes (Gratton 2002; see Karimi and Wilkes 2022 on fitting the data into erroneous models). ¹⁵

Neo-Assimilation

A mixture of deductive-inductive hypothesizing informed the formation of the neo-assimilation hypothesis. Neo-assimilation observes racial diversity in the mainstream and explains this diversity as resulting from a process of assimilation. Here, the mainstream is the core set of social institutions where the native majority groups feel most at home (Alba and Nee 2003, p. 12; Alba, Beck and Basaran Sahin 2018, p. 101). In the mainstream, “individualistic and collectivist adaptation” (Alba and Nee 2003, p. 63) strategies are more important than race or skin color as barriers to assimilation (ibid, p. 48). Accordingly, a key question for neo-assimilation is “how then should assimilation be defined, given the prospects for a more racially diverse society arising from large-scale immigration of non-Europeans?” (ibid, p. 10). The answer is to define assimilation, in light of the mainstream outcome, as “the gradual erosion of ethnic, racial, religious, and other differences as determinants of life chances for immigrants and their children” (Drouhot and Nee 2019, p. 178). The result is a framework in which the mainstream is the outcome and assimilation is the predictor.

According to scientific methods, a hypothesis starts with a theory and hypothesizes a new outcome (deductive) or with empirical data and hypothesizes new predictors (inductive). Indeed, as per the deductive method, neo-assimilation relies on neo-institutionalism theory to understand social processes such as assimilation. Neo-assimilation, however, does not hypothesize a new outcome different from the already-observed mainstream outcome. Comparatively, as per the inductive method, the mainstream is neo-assimilation's outcome or “observed pattern [of] unprecedented ethnic, racial and religious diversity” (Drouhot and Nee 2019, p. 178). Neo-assimilation does not hypothesize new predictors different from the already-observed assimilation process. As a result, it is apparent that neo-assimilation combines and partially follows the deductive and inductive methods. The theory departs from the scientific methods of hypothesizing in three ways, making the theory not readily testable.

First, neo-assimilation starts with neo-institutionalism theory to make sense of the already-observed pattern or outcome, the mainstream. This is as if neo-assimilation is using neo-institutionalism as a predictor-provider tool. Neo-institutionalism theory maintains that individual actions and various policies layer up to shape the institutionalized ways of life in each country. Then, a feedback cycle between individual behaviors and institutions reproduces a normative consensus that sustains the institutions and prevailing life patterns (Mahoney and Thelen 2010). Accordingly, neo-assimilation explains the interactions between individuals’ actions and the institutions. In neo-assimilation, these interactions entail “the purposive action of individuals and the informal (i.e., networks and norms) and formal (i.e., rules and laws) constraints” (Alba and Nee 2003, p. 53).

Still, instead of extracting predictors from neo-institutionalism, neo-assimilation departs from neo-institutionalism to offer its own argument: as a result of the increasing numbers of nonwhite immigrants and the purposive actions of individuals, “ethnic stratification orders ultimately tend to become undone” (Alba and Nee 1997, p. 839) and, after that, the mainstream breaks away from Whiteness (Alba and Nee 2003, p. 13). This hypothesis is somewhat different from what neo-institutionalism, the underlying theory, would predict. The latter theory maintains that institutional consensus breaks on rare occasions when a sudden and irreversible divergence emerges between social institutions and individuals. The divergence disrupts the feedback cycle and stalls institutional functions (Ellermann 2015). Because these are rare moments, neo-institutionalism emphasizes that the institutionalized ways of life are modifiable but not easily replaceable. Therefore, if neo-assimilation hypothesizes a break from the institutionalized way of life in the United States, it would be desirable to delineate the specific predictors that shape such a break. To date, such predictors are missing from neo-assimilation writings (Alba and Nee 2003, p. 54).

Second, as with segmented assimilation's underclass, the social and economic definition or boundaries of the mainstream are lacking; hence the absence of clarity as to what kinds of data could falsify the existence of the mainstream. It is stated that “even in segregated neighborhoods, minority citizens know very well what the mainstream is” (Alba and Duyvendak 2019, p. 111); in each society, there are several “mainstreams” each of which is “internally diverse, differentiated by region, religion, and social class” but not necessarily “ethno-racially exclusive” (ibid, p. 110). Because neo-assimilation accepts that racial differences have already dissolved in the mainstream, it is unclear who the native-born majority is or what the prevailing life patterns in such mainstreams are. As such, the inclusion–exclusion criteria for the mainstream remain undefined. There is no way to “measure assimilation against any fixed racial characteristics” (Favell 2022, p. 37). Most recently, Alba (2020, p. 147) has argued that the mainstream still “reflects the social power of the White majority, which effectively controls the institutional and social spaces.” Regardless, assimilation research has invariably operationalized the “mainstream” as non-Hispanic whites’ ways of life.

Third, there is a circular reasoning between predictor and outcome in neo-assimilation. As mentioned above, neo-assimilation takes the assimilation process as the predictor but a predictor contingent on the mainstream outcome. The outcome entails the predictor, and the predictor leads to the outcome. For instance, a core component (or a predictor?) of the assimilation process is boundary blurring. Boundary blurring presumes two groups. If individuals from one group want to join the other, they adopt the cultural behaviors of the target group and “cross” the boundary onto the other group. The process ultimately leads to blurred indistinctive boundaries (Alba and Nee 2003, p. 60), “to hyphenated, if not hybrid, identities, which allow individuals to feel that they remain part of the group of origin” (Alba 2006, p. 356). ¹⁶ In this process, boundary-blurring is simultaneously the individuals’ actions, or predictors, “that can affect assimilation” (ibid, p. 348), a “blurred-boundary context” of assimilation (ibid, p. 356), and a “blurred boundary idea” of assimilation (ibid, p. 354 italics added). That is, boundary-blurring is a predictor, a context, and an outcome. It is not entirely clear whether the individuals are assimilating or whether the mainstream is blurred and, thereby, expanding (see Popper 2002 [1956], p. 76). As the outcome and predictors are not separable, they cannot be tested as such (see Goldthorpe 2001, p. 5). There seems to exist no way out but to abandon assimilation and, instead, adopt “the generic process of incorporation and prefer this term to the more one-sided and race-related (at least in the United States) concept of integration” (Drouhot and Nee 2019, p. 179 italics added).