A portion of this blog is consistently devoted to “What can we learn?” type of posts, discussing research that uses history to understand timeless political, economic and social questions – from the effects of immigration, to those of small government and lessons for the spread of pandemics. Besides the concrete historical insights that such research generates, it also aims at generalization of knowledge outside the specific historical context. Given that empirical work – in HPE and outside of it – does not yet have a rigorous way of credibly generalizing causal effects out of the local contexts in which they are identified, theory becomes very important for interpreting causal findings and aggregating them into a coherent body of knowledge. Yet the connection of empirical HPE and abstract, generalizable theory is weak, at best. Sean Gailmard and Arturas Rozenas have made these points in thought-provoking posts.
Arturas has highlighted the fact that current work in HPE is not abstract enough to enable generalization. A slightly different formulation of the same idea, which is the focus of this post, is that HPE researchers, in their majority empiricists, do not put enough effort in engaging with existing abstract theory in designing their studies and interpreting their findings. What currently tends to happen in a lot of HPE work is either total absence of theory or the proliferation of ad-hoc “theories” generated by researchers to explain the empirical findings in the context that they study. These theories are usually narrow hypotheses with little generalizability. Often, disciplinary incentives in some fields in which HPE scholars operate push them to generate overarching “big” theories out of very context-dependent cases, because new theorizing is valued more than empirical testing of existing theories.
My preference is for theoretical parsimony and some degree of specialization, in either empirics or theory, so that scholars can focus on their comparative advantage and increase gains from intellectual trade. In that model of knowledge accumulation, empirical researchers rely on the insights of established theory, at the most abstract level possible, to design and interpret their work. By so doing, individual empirical studies, both descriptive and causal in nature, can begin to form a coherent body of knowledge that provides evidence consistent with some theories and discredits or leads to the amendment of others.
Assimilation as a case at hand
I think about these issues a lot in my own work. I study assimilation, a process which is uniquely suited to be studied over the long run and using historical data. It is also a process that is worth understanding in more general terms. Lack of assimilation in some dimensions, such as economic outcomes, can be diagnostic of persistent minority disadvantages and discrimination. But complete assimilation may also be undesirable, as people value their cultural identities. Majorities and governments across the world demand or pursue assimilation in various ways, from aggressive immigrant integration policies to ethnonationalist citizenship laws. It is important to understand what these pressures mean for social cohesion and the outcomes of vulnerable minority groups.
HPE researchers can look to history for answers, for example by examining how assimilationist laws affected minority identity in the past. But credibly identifying the effect of a law is not enough to be able to answer the general question “How do assimilationist laws affect minorities?” or predict how an assimilationist law will affect a minority in a contemporary context. A credible causal design can give us some answers – for example, whether a particular effect of a law is at all possible, something like a proof of existence – but not a general answer to the questions above. For that we need many credible causal designs and a theory for to choose them and make sense of their results.
How can empirical work in HPE that studies assimilation profitably engage with abstract generalizable theory? In a chapter prepared for the Oxford Handbook of Historical Political Economy, I discuss how a rational choice-based framework that incorporates some well accepted psychological mechanisms like social identity can explain a range of findings on assimilation in HPE. That assimilation is consistent with rational choices on the part of multiple actors – minority members, majority members, the state – is not a new claim; scholars across the social sciences have made it before. But empiricists in HPE (and outside of it) don’t frequently use existing rational choice models of assimilation, many of which make rich predictions, to guide their work. Yet there is ample room to do so. One example that illustrates this is the effect of assimilationist laws on minority identity.
Assimilation policies and backlash
State laws aimed at promoting a majority identity and assimilating minorities sometimes succeed in their goals. Dippel and Frye show that land allotment policies enacted in the US in the early 20th century were successful in incentivizing assimilation among Native Americans, which federal authorities measured by rates of education. Assimilation policies that impose education in the language of the national elites have also sometimes been successful. Blanc and Kubo creatively use data on local dialects to show that the 19th century Guizot law that expanded primary education in France drove linguistic homogenization and increased national identification in the long run. But in other instances, assimilation campaigns lead to a backlash. In my own work I have found this to happen with language bans that strengthened the identity of immigrants in the US in the 1920s. In a similar vein, Seyhun Sakalli found secularization campaigns in Turkey to have led to a religious backlash by pious parents who reacted to assimilation of their children into secular values.
Empirical work identifies the possibility of backlash. But it cannot explain when backlash will occur. Each of these cases differs from the rest in countless contextual details. It is the work of theorists to aggregate the cases into a framework that predicts when backlash is more likely. But it is the job of the empirical researchers who want to answer the same generalizable question to look to existing theory that may provide guidance for their empirical work.
Theories of backlash
Several existing theories rooted in rational choice predict backlash. Without being exhaustive, one can broadly distinguish between individual-level and group-level mechanisms. At the individual level, backlash may be the result of resentment to policies perceived as coercive. Whenever incentives to assimilate are not too strong, such resentment may dominate in the decision-making of minorities and drive increased investments in their identity as a psychological response.
Simply assuming resentment may be unsatisfactory, but this response can also be microfounded in a rational framework that incorporates identity considerations. The seminal work of Moses Shayo, building on social identity theory, posits that individuals derive utility from group belonging, make choices on which groups to identify with, and then internalize the welfare of those chosen identities. Individuals desire to identify with groups that are viewed positively, or enjoy high status. One way to achieve high status is assimilation into a dominant majority. But there may be minority members who find this requires incommensurate effort – for instance culturally distant groups that struggle to reach majority standards, or groups subjected to policies hard to comply with, such as religious prescriptions which touch upon deeply held values. In those cases, doubling-down on minority identification is an alternative means to improving self-image and a substitute strategy to assimilation.
Backlash can also ensue as a group-level phenomenon. For instance, assimilation policies that increase the cost of minority behaviors lead to negative externalities for all members of a minority community. A club goods model of groups predicts that, in response to such externalities, the community or its leaders increase the stringency of prescriptions or otherwise require stronger signals of commitment to the minority group, in order to screen out community free-riders.
Back to the empirics
Both classes of mechanisms predict backlash against policies that raise the cost of minority identity. In both cases, the likelihood of backlash depends on the characteristics of the targeted group. Backlash is more likely among those whose costs of assimilation are higher and who are least able to “make it in” and be accepted by the majority if they increase their efforts to comply with assimilation policies – for instance the least integrated or most culturally distant individuals. Minority members with low assimilation costs have an incentive to leave the minority group under the club goods model and are more likely to assimilate than to retreat in a social identity model.
But the two frameworks also deliver different predictions. For instance, when group-level mechanisms are operational, minority behavior is affected by social sanctions or prescriptions at the community level. When the mechanism is individual-level and psychological in nature, such factors do not play a role. To distinguish between the two mechanisms, an empirical test might then look at whether minority members modify their behavior when observed by the group.
The case of backlash is an example for how guidance from sufficiently abstract and generalizable theoretical models is crucial for empirical work. Identifying a backlash effect to a policy could indicate that some behavioral – identity-related or other psychological – mechanism may be operative. By looking to existing theory and identifying what differentiates various behavioral mechanisms in terms of their predictions, the empirical researcher can then design their study with the purpose of distinguishing between them. For making progress on answering the general question of when backlash occurs, empirical researchers should look for cases that allow for credible identification not only of the first order effect of an assimilation policy, but also that of mechanisms that differentiate between different theories which can produce a backlash effect.
By all this I do not mean to say that empiricists cannot or should not contribute theoretical insights through their work. Sometimes, none of the existing mechanisms may fit the empirics, and new theoretical work is needed to make sense of the findings. An empirical HPE study can contribute that insight by amending existing theory. But we can’t get to that point without ruling out available theoretical explanations first. And we should strive for parsimony, by not straying too far away from existing theory. What is certain is that there cannot be as many theories as there are HPE papers, if it is for our field to generate any generalizable insights.
 Sometimes, evidence on multiple mechanisms may be found. For instance, Carvalho, Koyama and Sacks (2017) show that the emergence of ultra-Orthodox Jewish sects in 19th century Europe was a response to assimilationist policies and low returns to secular schooling and was driven by both individual and group-level channels.