Causal Status of the Main Claim

Establishing that the link between ethnic structures and policy outcomes is causal is clearly important to Lieberman. The argument for exogeneity is a simple one: Since the onset of AIDS is a relatively recent phenomenon and ethnic structures are slow moving, it is implausible that the former caused the latter; rather, the onset of HIV-AIDS can be thought of as a shock commonly applied across countries. Indeed, Lieberman argues that much previous work on the effects of ethnicity in other policy areas suffers from a “strong likelihood of endogeneity,” which he avoids. The argument is seductive but I think ultimately incorrect. Normally, when one seeks a “shock” to establish exogeneity, one looks for a shock to the independent variable; in this case, one would want some shock to the structure of ethnic boundaries that is uncorrelated with potential confounds. Here, the shock is to the timing of the choice situation, but it is not clear that this solves the endogeneity problem. To see why, let us say that for some third reason, a country has difficulty implementing coherent policies in some class of areas and that this has led to a strong internal ethnic boundary structure. Say now that policymaking in the area of HIV-AIDS falls within this class. In this case, boundary structures will be correlated with, but not necessarily causally related to, AIDS-HIV policymaking. Thus, even if ethnic boundaries are exogenous to the onset of AIDS, the author's estimates of the effect of boundaries on policy choices suffer from the same risks of failure of identification for which he criticizes other work.

Contribution of the Main Claim

How novel is the main claim? The clear focus on ethnic structures as an explanation for health policy choices is an important contribution. But it joins an already large literature suggesting that ethnic divisions lead to poor policy outcomes in general. Abhijit Banerjee and colleagues demonstrate, for example, that social divisions in India, similar to those examined by Lieberman, are associated with worse public goods provision outside of the context of HIV-AIDS.Footnote ¹ So is there really something particular about HIV-AIDS that merits specific explanation? There are two ways in which Lieberman's argument may differ from the general claim that diversity impedes public goods provision. One is to establish that the mechanisms through which diversity affects AIDS policy really are different from those in other arenas; another is to establish that the problem is particularly severe (or less severe) than in other domains. Lieberman does not address claims of the second type: There is no evidence given that the relationship between these indices and adverse policy outcomes is worse for AIDS than it is in other areas. In the following, I consider the evidence for a claim of the first type.

Rivals to the Explanatory Claims

Do these explanatory claims account for the patterns that support the main claim? While Lieberman entertains other explanations, he pays less attention to truly rival explanations—that is, alternative logics that could account for the same patterns. From the broader study of the effects of ethnic structures on policy outcomes, however, we know that there are many possible rivals. There could, for example, be a realistic assessment of the distribution of risk, but a reduction in compassion for the welfare of outgroup members who face higher risk than you do; or perhaps there could be a general inability to coordinate across group lines. Or perhaps strong ethnic boundaries structure the distribution of the disease in the first place, because of differences in behavioral patterns, because of patterns of between-group inequality, or because of greater fragmentation of social arrangements, introducing greater variation in the distribution of risk. Such unevenness in exposure may account directly for the weakness of responses.Footnote ²

There are, in addition, other political possibilities open to groups even if they operate within the general logic suggested by Lieberman. The most obvious is for affected groups to counteract the stigma by arguing precisely that AIDS knows no boundaries, and rather than seeking no action, lobbying for broad action targeted not only at their group. A second possibility is that affected groups engage energetically in blame politics, but use this to argue more vociferously for robust responses (“They made this problem; now they should pay to fix it.”).

Evidence on the Explanatory Claims

Now let us consider the evidence for the explanatory claims. Let me begin by noting two concerns related to the types of observable implications that can be used to assess these claims.

The first relates to information on perceived and actual distributions of risk. Both of the explanatory claims predict, for different reasons, a disconnect between the actual and the perceived distribution of exposure to HIV-AIDS. Understanding the relation between the two seems critical. At multiple points, however, Lieberman suggests that the actual distribution is not what is politically important, that what matters is perceptions. There is clearly some truth to this, and in some sense all data is perception. But there are risks run when the actual epidemiology is ignored. First perceptions surely reflect the actual epidemiology to some extent. But the core idea of the second explanatory claim is, I believe, precisely a claim about how perceived risks depart from some notion of actual risks. We are provided with information that some groups believed that risk was higher than others, but without understanding how these beliefs correspond to the actual epidemiology, we cannot assess whether group boundaries are really producing biases in perception or whether the language of group risk is a shorthand for describing relevant epidemiological patterns.

The second relates to the structure of the propositions. Lieberman argues that variation in ethnic boundaries creates variation in the extent to which relevant groups push for policy action. But this proposition is trickier than it might seem at first blush, since here the outcome is defined as a function of the treatment. A hypothesis of the form “the existence of a multiplicity of groups causes relatively underexposed groups to underestimate their risks” faces difficulties because in cases in which there is not a multiplicity of groups, there are no relatively underexposed groups; one cannot even begin to assess the evidence for the claim. The sort of statement that could be examined requires units that are independent of the treatment: “individuals that face relatively low risks underestimate their risks when they share group membership with other relatively low-risk individuals.” To engage in that kind of analysis, however, would require focusing more on the distribution of actual risk than is presently undertaken.

Let us now turn to the cases. The core evidence for these explanatory claims comes from a case study of India and, to a lesser extent, the South African and Brazilian cases.

What are the observable implications for the Brazil case? It turns out that an implication of the concerns I raised earlier regarding the structure of the causal proposition is that it is hard to know what to look for in the Brazil case. If indeed there are no relevant group boundaries, then how can one begin to assess whether more highly affected groups underestimate risks while less affected groups make weak policy demands? Evidence of such effects would violate the premise of the case. Indeed, the Brazilian discussion is relatively brief and consists largely in confirming that group boundaries are weak and that there was a national response in Brazil.

The South African case is richer. The observable implications here are that less affected groups should underestimate their risks and not support action, while more affected groups should be concerned about status and not demand action. Lieberman finds instances of both types of behavior. But South Africa remains puzzling. One would expect that if there were any place where the effects Lieberman describes would not be determinative, it would be in a case in which the affected group was both a large majority and in control of policymaking. More puzzling is why even when there were responses in South Africa they were so misguided, ranging from the naive (advocating better diets and vitamins) to the misguided (promoting alternative medications without the basic research behind it) to the criminal (raping babies). Perhaps this reflects the challenges facing attempts to construct a well-informed and effective lobby for serious HIV-AIDS responses in the country, but the fact that responses were “divorced from scientific best practice” (to use Lieberman's phrase) and not simply absent sits uncomfortably with the explanation that the driving force was fear of a loss in status that is associated with recognition of HIV-AIDS. President Thabo Mbeki's desire to find an African cure does of course reflect race politics in some way, but not in the way suggested by Lieberman's theory. A final puzzle is that although there is ample evidence of a politics of blame in South Africa, much of the blame that Lieberman discusses is in fact blame about the origins of HIV-AIDS. But such a politics of blame (unlike the politics of pure denial, which also exists) does not necessarily imply inaction, particularly by a dominant majority. Indeed, the blame could in principle provide a rhetorical basis for strong action. The African National Congress government had an opportunity to blame past governments for inaction and reorient policy in the area in favor of their constituencies.

The India discussion is broadly consistent with the theory, but some of the questions from the South Africa analysis arise here also. While there is evidence of denial, much of this is actually from the Hindu nationalists; while there is blame, much of this is blaming foreigners; while there appears to be a lack of national-level response, the national distribution of the infection rates does in fact appear to be very uneven (in the author's data, prevalence rates vary across regions from 0.05% to 2.27%, a rate almost 40 times as great).

Other cases discussed more briefly also provide mixed evidence. In Nigeria, according to Lieberman, groups in AIDS hotspot areas are the ones that criticize the lackluster response of governments as being part of a plot to destroy them, while the low-prevalence groups describe AIDS as a Western gimmick intended to destroy indigenous values. Neither of these patterns seems consistent with the theory. Gambia is cited as a case in which weak group boundaries have not led to adverse effects, but the government responses to AIDS in that country seem no better than the patterns seen under Mbeki in South Africa (President Yahya Jammeh is reportedly running a program based on an herbal cure for AIDS that was revealed to him by ancestors in a dream). Finally, the Uganda case also continues to puzzle. Uganda has strong ethnic divisions on multiple dimensions and has had weak responses in many policy areas. But it has been a leader in combating HIV-AIDS. Lieberman discounts the case because of the early rapid diffusion of the disease there and the fact that multiple groups were affected. Yet there was considerable variation in rates across Uganda, with still greater variance in early years. Moreover, his focus is on perceptions of prevalence, which he expects to be endogenous to group structures.

Do the cases convince? It depends on what one hopes to get from a small set of cases. Lieberman's breadth of research is exceptional—he undertakes masterful analysis of three cases on three continents. But three cases are few for assessing a general relationship and, as a result, too much gets asked of the cases; he tries to show not just that the dynamics he describes are at work but also that his explanation is the correct one in all cases. To do so, he tries to discount all other explanations (including the unsatisfying but possibly correct explanation that leadership matters). But establishing that Lieberman's story is the story in this small set of cases is neither necessary nor sufficient for the claim that the logics he identifies are universally critical. In the end, the evidence from the cases is mixed. But we should probably expect mixed evidence in particular cases even if the theory is right on average. A more modest ambition for the small set of cases is to demonstrate that the logic Lieberman identifies has face validity. In this, he is successful.