Guest Author: Dr. Thomas Millington is a professor emeritus of political science at Hobart and William Smith Colleges.


In a seminal book that compared the Jewish and Hispanic immigration experiences (Hattam 2007), Victoria Hattam announced the stakes that are involved in how Hispanics identify racially in the United States.  “If Latinos continue to identify as whites, or as white ethnics . . . the central cleavage will continue to fall between Blacks, on the one hand, and whites, allied with white ethnics, on the other.” However,  “if large numbers of Latinos break with the past (European immigrant practices)  by racializing their ethnicity, they may well shift the balance of power by forging a sizable coalition between Blacks  and Latinos identifying as ‘people of color’. ” (Hattam, 2007, 130; parenthesis added; the term “Hispanic” will be used in this article in preference to the term “Latino.”)

Hattam thus makes the  formation of a black/Hispanic coalition the basis for Hispanics to “racialize their ethnicity” in a manner that will displace the binary.  Racializing their ethnicity in this manner for this purpose, however, has proved highly problematic. Extensive empirical  evidence of interactions between blacks and Hispanics in all manner of venues, from labor markets to political coalition-building, reveals as much conflict as cooperation. (Telles, Sawyer, Rivera-Salgado, 2011; Lee and Bean, 2010). The underlying stumbling block is the fact that  most lighter-skinned Hispanics harbor the desire to assimilate by becoming white rather than risk becoming a racialized, disadvantaged minority by joining forces with blacks. This posture simply insures that that those darker-skinned Hispanics unable to become white will end up as just such a racialized minority.

Hattam overestimates the appeal to Hispanics of racializing their ethnicity by joining  forces with blacks to fight against white racism. Stephen Steinberg notes in a sympathetic but critical review of Hattam’s book (Cf. Skerry, 2009) that the author is right that the concept of immigrant ethnicity, articulated mainly by Jewish scholars, arose in opposition to the “stigma of race in the United States (read blackness).” But for Steinberg that does not mean, as Hattam contends, that immigrant ethnicity has the inherent potential to challenge white racism. On the contrary, Steinberg argues, immigrant ethnicity served in Jewish thought mainly as a way to become American without giving up what it is to be Jewish; without “assimilating into a bland Americanism,” on the one hand, or being lumped together with blacks in a “racial pariahdom,” on the other hand.  As such,  Jewish immigrant ethnicity was a way to not become involved in issues of racial discrimination against blacks; to become involved was seen as undermining their Jewishness, and sub rosa, undermining  objective of becoming white ethnics. Carving out a Jewish identity simply took the white/black binary for granted instead of challenging it.  For Steinberg, immigrant ethnicity, Jewish and non-Jewish, is inherently  a “stepping stone on the way to whiteness.” (Steinberg, 2009, 189-191; parenthesis in the original.) Steinberg is therefore dismissive of Hattam’s proposition that Hispanics may ever choose to racialize their ethnicity in a manner that blocks their path to becoming whites, or white ethnics, in the United States.

Notwithstanding the prognostications of Telles et al., Lee and Bean and Steinberg, the present article  will adopt Hattam’s perspective that by racializing their ethnicity Hispanics can successfully displace the white/black binary as the site of the central cleavage in American society. The argument offered here will, however, shift the site of Hispanics racializing their ethnicity from the problematic and unpredictable terrain of coalition-building with blacks to the terrain of Hispanics’ relations with the Census Bureau regarding the question of how Hispanic race is to be identified. At this site, Hispanic ethnicity will be taken to mean their country of origin. Racializing that country of origin ethnicity will be taken to mean consulting the racial structures in their countries of origin as a basis for determining their race in the United States.

The advantage of the Census form as the site for Hispanics to racialize their ethnicity along the preceding line is the fact that the Census Bureau has itself developed an interest in seeing this happen. Until Census 2000, the formatting of the racial choice question for Hispanics on the  Census form was deeply biased against Hispanics taking any account of their racial backgrounds in Latin America.  Census 2000 however indicates a change, albeit tacit, in the Bureau’s thinking about Hispanic race. By allowing respondents for the first time to choose more than one race the Bureau was implicitly  inviting Hispanics to consider their lines of descent from  the racial-mixing patterns in their countries of origin.  It cannot be presumed that the Bureau believes that all of Hispanic race-mixing began after they arrived in the United States!

In Census 2000 and 2010, however, Hispanics largely failed to exercise the multiracial option, preferring instead to stick to their favorite single race categories: White and “Some Other Race” (SOR). The White category is obviously a vehicle for expressing a Hispanic preference for becoming white or white ethnics in the United States. The SOR category expresses for Hispanics a protest that the official race categories on the Census form that Hispanics are being asked to choose from do not embrace  the reality of the Hispanic racial makeup. More specifically, perhaps, it reflects the view that Hispanics choose “other” because they are  are “beyond” the white or black question which so obsesses Americans (Navarro, 2003). A wide divergence has therefore developed between the Census Bureau’s preference for a multiracial turn in Hispanic preferences for their race, one that necessarily would bring their racial backgrounds in Latin America into the picture, and the Hispanic preferences for exclusive reliance the White and SOR categories, a bipolar projection of Hispanic “race” that has nothing to do with their actual racial backgrounds.

The question then becomes whether there is currently room in the Census Bureau /Hispanic relationship for successfully encouraging Hispanics to take more account of their country of origin racial structures in making their choices of race on the Census form.  How can the Census Bureau format the racial question so as to make this happen? Why did  the “more than one race” initiative  fail to generate significant Hispanic responses? What incentives can the Bureau introduce that will realistically cause Hispanics to voluntarily move away from exclusive reliance on the White and SOR categories? Is it possible for the Census Bureau to delegate to Hispanics in a compelling manner the discretionary authority to establish their race on a basis that aligns it with  the preferences of the Census Bureau?

Questions of this sort can be translated into the “principal’s problem” in principal-agent theory: creating an environment in which the agents (heretofore Hispanics) have incentives to align their preferences with those of the principal (heretofore the Census Bureau). By casting the Census Bureau’s relations with Hispanics in principal-agent terms it is possible to structure a prescriptive analysis that points the way to the achievement of alignment between Census Bureau and Hispanic preferences on the question of how to define Hispanic race. After delineating this analysis, and paying particular attention to Holmstrom’s theory of delegation in principal-agent theory, the article telescopes the  considerations and expedients that arise from the analysis into a concrete proposal for changes in the way the question on race on the Census form  is posed for Hispanics. If this proposal is adopted by the Census Bureau and its supervisory agency, the Office of Management and Budget (OMB), there is a reasonable expectation that the Hispanic manner of selecting among the racial categories offered to them will reflect their respective racial backgrounds to a significant degree. This would represent a vast improvement over the present Hispanic distribution of their “race” into the White and SOR categories as far as accuracy relating to actual Hispanic racial makeup is concerned. Furthermore, their distribution among a set of categories specifically tailored to reflect their multiracial background in Latin America would deter Hispanics from assimilating in the United States along racial lines laid down by the binary: light-skinned Hispanics joining Asians on the white side of the binary and dark-skinned Hispanics, including Afro-Hispanics, joining non-Hispanic immigrants on the black side of the binary.


Principal-Agent Theory


In the field of legal contracts the concept of principal-agent relationships was known originally with a different terminology. One party to the contract, the “promissor,”  promised to pay another party, the “promisee,” a sum of money for undertaking in good faith the performance of a task on behalf of the promissor. This concept of a legally binding monetary contract parallels, legal scholar Eric Posner observes, (Posner, 2000) the wider and more flexible concept of  an “agency relationship.”  Here “one person, the ‘principal,’ benefits when another person the ‘agent,’ performs some task with care and effort.” Problems arise in the agency relationship  when this care and effort is not forthcoming from the agent. In addition, preferences between principal and agent may diverge especially if agents develop agendas based on information that they possess but the principal does not.  But above all, Posner observes, the principal-agent relationship presupposes a  cooperative effort that is endowed with potential resources to overcome agency problems in a manner that  safeguards the interest of the principal.  For example, when an employer arranges for high quality training of a shirking employee. The knowledge acquired by the employee enables her to put more skillful effort into her work. As a result her preferences begin to be aligned with the expectations of the employer. In effect, the employer would “become” the agent, as Posner puts it.  By the same token the employee could be regarded as becoming “principled.” (Besley, 2007).

Can the same kind of alignment be brought about in the relationship of the Census Bureau to Hispanics on the question of the latter’s race? What would ”training” of Hispanics in this respect consist of? What would Hispanics becoming “principled”  mean in the matter of choosing their race?


*    *    *    *    *    *    *    *    *


An iconic example of the principal-agent relationship is the landlord-tenant case. The risk for the landlord is that he puts the tenant in a position of responsibility for the well being of the apartment but he does not have the capacity  to monitor the actions of the tenant which may result in him “trashing” it.  The principal’s risk here is called moral hazard (hidden agent actions). There is also an element of information asymmetryin that the tenant has more day to day knowledge of what’s going on in the apartment building than the owner. The landlord’s concern is that tenant behavior be such that new and equally well behaved tenants may rent his other apartments on the same floor. The actual  tenant’s  preference for neighbors may not parallel the landlord’s. He has a friend, let’s say, who is drug dealer and is currently under consideration by the landlord for tenancy in the adjacent apartment.  The actual tenant withholds this information from the landlord who proceeds to rent to the drug dealer. This is a case of the principal making an adverse selection based on hidden information on the part of the agent.  It is similar to the case of the used car salesman (agent) who hides  information about a car’s mechanical faults from a buyer (principal) who ends up getting a “lemon.” (Akerlof, George, 1970; Spence, 1975.)  Adverse selection. Both moral hazard and adverse selection problems usually result from the agent having an informational advantage over the principal, and using that information, or not using it, in ways that are prejudicial to the interest of the principal.

The agency problem in the landlord-tenant example is exemplified by the particular combinations of adverse selection and moral hazard that can occur. The agency problem, thus defined, gives rise to “agency costs.” These costs refer to the investments made by the principal in an information system that enables him  to monitor tenant behavior.  The principal can also invest in closer screening systems for future tenants. Monitoring and screening are informational systems that  principals can acquire at a cost for the purpose of mitigating information asymmetry and the problems of moral hazard and adverse selection that derive from it.

The tenancy contract can be behavior-based or outcome-based. These are staples in principal agent theory (Eisenhardt, 1989;  cf. Shapiro, 2005). In the behavioral case the acquisition by the principal of an informationally based monitoring system provides leverage for the landlord to modify tenant behavior in his preferred direction. An outcome-based contract would be based on the resulting condition of the apartment after the period of  tenancy and will involve the determination of whether a security deposit shall be returned to the tenant or not. In the employer-employee field, salary  is the basis of a behavioral contract. Commissions are the basis for an outcome contract.

Eisenhardt  argues that the “heart of agency” is a “trade-off” between the costs to the principal of measuring agent behavior and the costs of measuring outcomes of agent behavior (cf. Jensen and Meckling, 1976).  The approach in this article is not quite so strict. It will attempt  to see that, notwithstanding problems of moral hazard and adverse selection,  there is ample room in the Census Bureau’s approach to Hispanics for behavioral contracting as well as outcome contracting in order to bring about alignment of the preferences of both parties on the question of Hispanic race. We will notice that one crucial ingredient for bringing this about  lies outside the compass of the landlord-tenant model but exists in many of the politically or bureaucratically oriented examples of principal-agent relations: delegation.

In light of the foregoing, it is possible to begin characterizing the relationship between Hispanics and the Census Bureau in principal-agent terms. It is a fundamentally cooperative effort. This effort consists in the Census Bureau engaging Hispanics to supply current information on their race in response to the question on race on the Census form. There are two implicit “contracts” historically underpinning the relationship between the two parties on the question of Hispanic race. One is outcome-oriented and the other is behaviorally-oriented. They have worked against each other.

The initiative for the first, outcome-oriented contract came from the Nixon Administration. President Nixon is generally credited with giving currency to the term “Hispanic” as an identifier for people of Spanish descent in the United States. Nixon wished for them to be specifically identified on the Census form as a separate group so that the Republican Party could groom them as a constituency that could, along with “model minority “ Asians, counterbalance the blacks as the  chosen constituency of the Democratic Party. The opening gun in this campaign was the Administration’s inauguration of Hispanic Heritage Week, followed by Administration pressure brought to bear on the Census Bureau to include a separate question on “Spanish origin” on the 1970 Census form. It was included only on the long form since the the much more widely disseminated Census short forms had already gone to press (Skerry, 2002, esp. 37-39).

President Nixon left office unceremoniously in 1974 but his Hispanic initiative lived on bureaucratically. The separate Spanish origin question was included in the 1980 Census short form. Prior to that, in 1977, the Office of Management and Budget issued  an administrative document, dubbed  Directive 15 (OMB, 1997). This may be regarded as the first contract made between the Census Bureau and Hispanics on the question of their race. Directive 15 was clearly in tune with Nixon’s intention to shape Hispanics into a Republican constituency. It did this by attempting to cast Hispanics in their majority as a white ethnic population. This first step in the Directive was to indicate that Hispanics were an ethnic group not a race. A Hispanic was “(a) person of Mexican, Puerto Rican, Cuban, Central or South American or other Spanish culture or origin, regardless of race.”  This can clearly be taken to mean that Hispanics do not have any determinate racial heritage of their own but as an ethnic group they are free to choose “their” race from any of the  four standard race categories instituted by Directive 15: White, Black,  American Indian or Alaska Native, Asian. Directive 15 makes it clear that not many Hispanics were expected to opt into the Native American Indian or Asian categories and that therefore their effective choice was between choosing to be white or black. In other words, the  white- black binary was being introduced, in effect, as the mechanism for sorting out data on Hispanic race. Thus, what was important  to know was the “number of White and Black persons who are Hispanic.” Hispanic race “must be identifiable, and capable of being reported” in these terms. Nixon could not have asked for more. This is an example of what would be identified in later principal-agent work: pandering. (Cranes-Wrone, et al., 2001.)  In this case, Directive 15 was pandering to the Hispanic assimilationist preference for white (as opposed to black) as the choice of their race on the Census form, at the expense of not taking any account whatsoever of the Hispanics actual racial makeup. This flagrant partisanship contrasts with the the OMB’s putative “neutral competence” as a bureaucracy (Heclo, 1999).

The Directive was in effect addressing two dimensions of Hispanic ethnicity. The formal one   concerned their country of origin. The other, informal one concerned race. Directive 15 structured Hispanic racial choice so as to disconnect it from the first one in order to connect it instead to the second one: assimilation of Hispanics as white ethnics.  In other words, the Directive was contracting with Hispanics to become white ethnics. This is outcome-based contracting.

By the 1990s the forces of multiracialism and multiculturalism, fueled by immigration and inter-racial marriages, were developing strength and political voice. Advocacy groups, including especially Hispanics, were pressuring for modification of Directive 15 so that the growing multiracial component of the American population could be represented on the Census (Prewitt 2013; Mora, 2014). In an attempt to accommodate this charged political environment, the OMB approved changes in  Directive 15, called the Revisions to Directive 15 and issued in 1997. (OMB, 1997). The key change allows respondents for the first time to  choose one or more of the single races presented to them. There would now be 64 possible choices, the six single races (Pacific Islander and Some Other Race were added to the original four) and the 57 different ways that the original six could be combined, beginning with two race combinations, three race combinations, four race combinations, etc.

The multiple choice response (MCR) was incorporated into Census 2000. By incorporating MCR into the Census question on race the Revisions in effect created a second contract between the Census Bureau and the Hispanics. It was behaviorally-oriented in the sense that it was intended to modify Hispanics thinking and behavior on their race in the multiracial direction. The expectation was that MCR would compete with the single race White and SOR responses and thereby generate accurate information on the the patterns and extent of Hispanic race-mixing. It did not turn out that way in Census 2000 and Census 2010. See Table1.

Three observations can be made regarding the data in Table 1.

First, White-Alone and SOR-Alone remain the dominant choices for Hispanics. 90.1% of the total Hispanic population opted into one or the other of these two categories in 2000. 89.7% did so in 2010.

Second, the White-Alone category is gaining ground among Hispanics apparently at the expense of SOR-Alone. White-Alone increased by 9,845,861 (5.1%) compared to an increase of 3,611,000 in SOR-Alone which represents a decline of 5.5%.

Third, the Multiple Choice Responses (MCR) , granted that it is the next largest of the categories after White-Alone and SOR-Alone, represented only 6.30% in 2000 and 6.03% in 2010. There is not much of a challenge to the two larger categories from the MCR source. In fact, the majority (55.9%) of the MCR combinations chosen by Hispanics are White in combination with SOR. (US Census Bureau,  2010) This suggests that even among Hispanics who are willing to combine the White category, the preferred combination is with SOR rather than with the other racial categories which are specifically non-white. Reciprocally, the Hispanics who are willing to combine SOR do so predominantly with the White category. Given this data, it is reasonable to expect that if the SOR category were eliminated, the great majority of Hispanics, who would otherwise choose SOR, would choose White-Alone [see endnote #1].

It was possible after Census 2000  to argue (Lopez, 2004, 2005) that SOR was well on its way to absorbing the White-Alone Hispanic population in the Census. Lopez cites data from the 1980, 1990 and 2000 censuses that indicate a relative decline in the number of the White Hispanics, coincident with an  increase in the SOR Hispanics. The data from Census 2010, however, seems to augur a reversal of this trend. A more likely read, however, is that White-Alone and SOR-Alone are operating as oscillating parts of the underlying white-ethnic strategy: enough White-Alone to insure positioning on the white side of the binary, enough SOR-Alone  to hedge against wholesale absorption of Hispanics into “whiteness”.

The white option is gaining particular traction among Hispanics in the third generation. A recent study (PEW Hispanic Center, 2011)  finds that among the first and second generations the two alternative ways to avoid the white choice were to choose SOR or the generic write-in term for race, Latino or Hispanic. These two categories combined dominated the choice of the white category by significant margins in the first and second generations. However, among the third generation Hispanics the combined non-white choice was equalled by the white choice which ascended from 36% in the first generation to 44% in the third. One question asked the respondents whether they considered themselves a “typical” American. 70% of the Hispanics, all making more than $70,000 a year, replied yes. 66% of Hispanics replying  yes were all English-dominant speakers. 64% of the Hispanics who replied yes all had  at least a high school education. Those Hispanics who are leading the way to seeing themselves as becoming “typical” Americans are thus predominantly third generation, relatively wealthy, relatively well educated, mainly English-speaking, upwardly mobile and white by choice.  Many younger Hispanics appear from these survey results to be on their way to repeating the successful assimilation into American life that their European immigrant predecessors achieved.

This situation creates a major problem for the second contract. It is behaviorally based on expectations of significant multiracial modification in Hispanic thinking and acting on their race, when that thinking and acting were heading in the opposite direction of those expectations. The rub for assimilation-bound Hispanics is that multiracial modification in Hispanic thinking on their race gets in the way of them thinking of themselves as whites or white ethnics. A major part  of the Hispanic intent is to be distinguished from whites ethnically but  not racially.  The Census Bureau obviously lacked the kind of monitoring informational system that could have clued it into the non-responsiveness of Hispanic behavior to any effort to modify their racial status in the multiracial direction.  As a result, the  Bureau got blind-sided by the anemic response of Hispanics to the MCR option. By virtually ignoring the MCR option Hispanics were in effect reasserting their  commitment to becoming white ethnics. No matter that race-mixing is deeply built into the Hispanics’ heritage. Contracting with Hispanics on their behavioral shift in the multiracial direction was therefore inherently subject to a agent “sabotage” or “subversion.” (Brem and Gates, 1997;  Gailmard, 2002). It was a morally hazardous undertaking from the beginning.

The choice of an open MCR field having 57 possible combinatorial choices represents,   in addition, an egregious case of adverse selection. Adverse selection usually results from  information being withheld by the agents from the principal because the agents prefer to use, or not use, that information in a way that is different from how the principal would use that information if she were in the agents’ shoes and had access to it. What the Census Bureau wanted, we can infer, was for Hispanics to supply information that would be more accurate as to their racial makeup, especially the patterns and extent of race mixing, than the information that Hispanics were supplying by piling into the White-Alone and SOR-Alone categories. That sought-after information on race-mixing would necessarily have to reflect  the racial background in the countries of origin as the source of some significant part of Hispanic mixed racial makeup.  The Bureau was expecting that Hispanics would therefore think more in a  genealogical way and less in a assimilationist way about their race. But it was precisely this genealogical information that was being “withheld” by Hispanics. If the Census Bureau had known more about the racial background of Hispanics, it would have been able to select categories for  Hispanic choices that reflected their background. It would, as such, be more likely to engage Hispanic genealogical information.  Confronting Hispanics with a mathematical maze of multiracial choices that were not pre-selected to reflect the unique Hispanic racial heritage simply insured adverse selection.

Ultimately moral hazard and adverse selection derive from an underlying information asymmetry between the parties. Hispanics have more information than the Census Bureau precisely in the area where the Census Bureau needs to know about it, namely the genealogical area. If follows that if the Census Bureau is going to mitigate information asymmetry in the genealogical area (it is not able to “challenge” Hispanic assimilationist information given its purely subjective basis) it is going to have to develop its own information system that will help Hispanics behaviorally to make more use of their genealogical information.  One logical way that the Census Bureau can develop such a system is to base it on diligent collection of data on the prevailing racial structures in the Hispanics’ countries of origin. This information system would  help Hispanics to organize their genealogical information, giving them the incentive to exercise more discretion in their choice of race as opposed to simply by opting into White-Alone and SOR-Alone. The latter two represent a case of the agent making a “slack” choices (Epstein and O’Halloran, 1994), ones that have not had  much “care and effort” into making them, in Posner’s phrase.

Using this enhanced genealogical information for reference value in deciding their  race might happen this way. Take a third generation Hispanic woman whose parents are from Guatemala and El Salvador. Knowledge of the racial structures extant in El Salvador and Guatemala would enable our Hispanic subject to sort out  how she might think about her own race. If the racial categories on the Census form that were  presented to her reflected the categories that  El Salvador and Guatemala had in common with regard to classifying  their populations, this would connect with the genealogical information she has at her disposal. The next step would be to take account of the  differences in the way that the Guatemalan and  Salvadoran populations  are distributed across the categories that they share.  The information system that the Census  Bureau might acquire could be designed to facilitate this kind of decision making on race for all Hispanics. Just such an information system is developed in the next section.

The acquisition of additional background information in order to facilitate making  decisions on their race would not impose  burdensome  agent  “decision costs” (Stephanson, 2007) for Hispanics because these costs would  have been largely transferred to the Census Bureau. The basic question, however,  is whether Hispanics can learn what they need to know about their race in order to become “principled agents”, as opposed to “slacking” or “shirking” agents (Brehm and Gates, 1997; Epstein and O’Halloran, 1994; Gailmard and Patty, 2007),  of the Census Bureau. Do they have the necessary motivation?

The development of the behavioral contract supported by the information system described above will obviously not in and of itself influence Hispanic behavior on race in the desired direction. It needs to be coupled with an outcome contract that is based on the reasonable expectation that Hispanics will choose their racial categories in a manner that reflects their country of origin racial structures.  Outcome-based contracts are more motivating for the agent than behavioral contracts (salary versus commission) but if there is high uncertainty about the outcome then contracting on outcome becomes doubtful. It can be demonstrated (Holmstrom, 1984) how delegation in the principal-agent relationship may be used by the principal to increase the certainty that the outcome of agents’ choices will  conform to a pre-existing strategic plan of the principal. Holmstrom’s  theory of delegation remains the model for subsequent work on delegation in principal-agent theory. (Alonso and Matouschek, 2008.)


Advantage of Delegation


Delegation is a specialized case of  decentralized decision-making. Open decentralized decision making (i.e. without delegation)  occurs for Holmstrom when the principal bases his policy decision on the relevant information he expects his agents to supply him with given that they possess more of it than the principal himself. Open decentralization decision making occurs under conditions of complexity where unscreened information received from all the agents has  to be coordinated by the principal. The manner of coordination that Holstrom  elucidates involves the principal making his best policy decision based on the selection of  those combinations of agents’ messages which are deemed to be more stable than all of the others. Holmstrom assumes that agents as rational choice makers will send a message that stands the best chance of being part of a particular combination of  messages that leads to a policy decision by the principal that is judged favorable to the agent’s self-interested expectations. (Agents are assumed to have their own policy preferences based on  subjective probability distributions across alternative outcomes.) It is assumed that the other agents are sending their own “best” messages and each agent knows which of all of the other agents best messages will be. Among the set of all possible combinations of agent messages there is a subset of them (one or more) that have the particular characteristic that no agent can improve his “payoff” from the resulting combination by substituting another message, provided the other agents also do not switch their messages. Holmstrom identifies these particular outcome combinations of agents’ messages as being in a Nash equilibrium and advises that the principal  will select  the one that most closely approximates his own policy preferences.

It is notable in open decentralization that the principal  is not steering agent messages toward his most preferred distribution. He is allowing the logic of the Nash “solution” to  make the decision for him  in the n–person, noncooperative game (Harsanyi, 1967-1968) that he has instituted among his  agents. The gains from coordination here are small, and the administrative and information transmission costs are high. Information is being “strategically transmitted” by the agents in the sense of Crawford and Sobel (1982) whose work influenced Holmstrom’s approach, but the information supplied  is not being deliberately  configured in order to realize a pre-existing strategic purpose on the part of the principal – one that includes the objective of creating an alignment of agent preferences with those of the principal before they send their messages.

There is a suggestive relationship between Holmstrom’s definition of open decentralized decision making (i.e., without delegation) in the principal-agent context and the engagement of the Census Bureau with the “more than one race” option in Census 2000. The decision of how Hispanics were to choose their messages on choice of race from among the 57 possibilities was left entirely up to them. There was no effort to screen the messages because there was no mutual understanding ex ante between the Bureau and Hispanics about  what criteria should govern the screening process.

This kind of mutual understanding between principal and agent is precisely what, for Holmstrom, makes delegation successful as a specialized form of decentralized decision-making. He avers that  “delegation is a special case where information does not get coordinated in the decision process. “ This allows “delegation to be viewed as a process where the agent is given authority to make the final decision, but subject to constraints (the control set) set by the principal.”  This sets the stage for communicative alignment of the preferences of agent and principal by creating  “dimensions of exchange” of  information between them. This  produces a  “movement of preferences in the same direction.” It is the absence of delegation in this sense that would account for the miscommunication between Hispanics and the Census Bureau regarding the orientation  which MRC was expected to impart to the definition of Hispanic race.

Holmstrom’s control set consists of the only messages, or categories  of messages, that the agent is free to  choose from. The delegation problem for the principal, viewed in the case of a single agent, is to select an optimal  control set  that maximizes his expected value E given his utility function Fp, the decision d based on the individual agent’s selection of a signal given the control set c, and the state of the world Z. Z summarizes the exogenous factors including random shock that will determine the efficacy of the policy decision in achieving its objectives. (Z in the  case at hand would be the factors that would determine  how efficacious Hispanic choice of their race in a manner that reflects their racial origins would be in displacing the binary in the “real world.” )  Formally, in Holmstrom’s notation (cf. however Demski and Sappington, 23),


E[Fp(d(y|c), Z)]


The optimal control set for the Census Bureau,  would consist  of that subset of  the 64 existing single race and multiple race choices that most efficiently reflected the racial structure extant in the counties of origin of Hispanics. These categories could be pre-selected by the Census Bureau based on the information it can acquire about the set of racial categories that are in fact most commonly used in the classification of these national populations. The existence of a control set efficiently oriented to the countries of origin racial categories would serve to encourage Hispanics to pay more attention to their genealogical information as this connects with to the categories being presented to them.

Would the control set effectively screen Hispanic choices of race? Since a significant number of Hispanics are actually white,  the white category would presumably be in the control set. Would not Hispanics be likely to continue in to pile into the white category given its assimilation value to them even though it is placed in a control set? The control set might not necessarily have the desired effect of preventing this except for one crucial factor: pre-commitment by the Census Bureau.

Pre-commitment is a crucial aspect of Holmstrom’s theory of delegation working successfully to create alignment between agent actions and principal’s expectations. It is potentially a powerful incentive for Hispanics.  By pre-selecting  categories that are tailored to Hispanic racial backgrounds the Census Bureau is conveying in advance (pre-committing) how the information is going to be used and interpreted. Holmstrom notes “that delegating authority to the agent as opposed to asking the agent for information and promising to act on the information in a certain way, is a more convincing form of precommitment.” He observes that, in recent testing of an apparently earlier version of his model by Crawford and Sobel (1982), it was shown that in the absence of pre-commitment by the principal, in ”the ensuing equilibrium the agent will disguise his information by reporting a coarser partition.”  It was precisely the absence of pre-commitment by the Bureau that characterized the presentation to Hispanics of the mathematical field of racial combinations. Hispanics had no assurance about what concept of Hispanic race, if any,  might emerge from a potentially wide dispersion of Hispanic messages in the combinatorial  field. For all that Hispanics knew  this dispersion might be seized upon  as proof that  Hispanics are so mixed (up) that they really do not have any racial coherence after all.  Without any pre-commitment by the Bureau, Hispanics  played it safe by disguising their genealogical information and reporting on the Census form a “coarse partition” consisting mainly  of White-Alone and SOR.  On the other hand, using the control set of pre-selected categories that is tailored specifically to the racial background of Hispanics pre-commits the Census Bureau in a manner that cannot but be compelling for Hispanics.

The Census Bureau in this latter scenario is delegating to Hispanics the authority to make a decision about how Hispanic race, as something Hispanics belong to by lines of descent, is to be articulated. Hispanics will provide that articulation by choosing how they distribute their “messages” across the pre-selected categories in the control set. This represents a  delegation to Hispanics of a secured right to claim their own racial structure in light of their geographic origins in Latin America. This is the right which has been denied to Hispanics by the  insistence in Directive 15 and the Revisions that Hispanics have no race that can be determinately connected to their countries of origin. “Countries of origin are not races.”  This insistence effectively converted Hispanics into  racial “orphans.”  It is that sense of being racially orphaned by the Census that drives Hispanics in such large number into the White-Alone and SOR categories. By eliminating their status as racial orphans it can be expected that  the White-Alone and SOR categories will  lose some of their magnetic attraction for Hispanics. At the same time, the Census Bureau would be getting accurate information on the actual racial makeup of Hispanics.

Holmstrom garners additional reinforcement for his theory of what makes delegation successful by  maintaining that if  information is endogenously acquired in  the functioning of the principal-agent relationship, and this increases and improves its quality [see endnote #2], then it can be expected that the preferences between the principal and the agent will become more “coherently” related to each other. Furthermore, with “more dimensions of exchange” between the principal and agent this tendency for alignment of preferences is enhanced. This suggests the possibility of a two-way flow of information between the principal and the agent. In this setting, the outcome of Hispanic messaging  on their race becomes less uncertain, compared to the open decentralization arrangement in Census 2000 and Census 2010, and hence outcome-contracting with Hispanics is more viable.

In effect then, we are looking at an outcome-based contract and a behavior-based contract becoming linked together by  delegation. Another way to look at this is to regard the behavioral contract, based on the  Census Bureau’s information system regarding Hispanics’ racial background, as the pivot around which the mechanism of delegation, based on the  outcome- based contract, turns. It is to the development of that crucial information system that we now turn.


Countries of Origin: Selected Race Categories and Data Sets


This section selects categories of race in the Hispanic countries that are based on positive data sets. They are intended to provide an information system on the basis of which the behavioral-and-outcome-based contracts described in the preceding section can function. The strategic purpose of both contracts is to have Hispanics take more account of their country of origin racial structures, and less account of the motives that currently drive them preponderantly into the White-Alone and SOR categories. These selected data-based categories are thus being “essentialized” but this is being done strategically in order to advance the progressive role for Hispanic race in curtailing the white/black binary in the United States [see endnote #3].

A logical place to start this undertaking is with census data for the Hispanic countries of origin. The Integrated Public Use Microdata Series  (IPUMS) is the world’s largest archive of publicly available census samples. The IPUMS-International data base includes 159 samples from 55 countries from 1960 onward. The data are supplied to IPUMS-International by participating national statistical agencies. The coverage of the Hispanic countries of origin is extensive: 16 countries (missing only Guatemala, Paraguay and Honduras) and 62 census year samples. To facilitate comparison across the different census samples the variables are “integrated.” This involves recoding the variable from each census into a “unified coding scheme.” Table 2 presents the IPUMS-International unified coding scheme for the race variable, and indicates the categories (marked by an X)  that are available for the most recent census samples from the Hispanic countries of origin.

It is evident that an alternative method for selecting categories for the race variable in Hispanic countries has to be employed if the categories are going to function as a coherent framework for constructing their race  in the United States.  The participation of the countries of origin in the categories of the IPUMS unified coding system is far too sketchy for this purpose. In the first place the alternative set of categories has to apply only to the countries of origin not the other not the additional 36 countries in the IPUMS data bank.

In the second place, the alternative approach needs to establish how whole national populations are distributed across the key categories of race once they are selected. This is the operative basis for strategically essentializing the categories. The  unit of observation is not the individual or the household but the group, i.e. the portion of a national population that is classified into one or the other of race categories. The key statistic is accordingly that of these categorical proportions of a population expressed as a percentage or a decimal value. As such, this  alternative approach is not easily adaptable to the microprocessing methodology employed by IPUMS.  Micro data consists of records of individual responses  to census questions that give information regarding individuals themselves and and households. The units of observation are the individual and the household. IPUMS therefore eschews the use of aggregate or summary statistics in order to exploit the advantages of micro processing.“Aggregate data are compiled statistics, such as (in) a table. There are no such tabular or  summary statistics in IPUMs data. Micro data are inherently flexible. One need not depend on published statistics from a census that compiled the data in a certain way, if at all. Users can generate their own statistics from the data by any means desired including individual level multivariate analysis.”

It would be possible to compile aggregate statistics from micro data but it is felt more efficient to start with  aggregate data compilations in the first place. In the case of micro data the numbers containing information on individuals and households are fixed and not in need of interrogation.  By starting with already compiled data the numbers can optionally be regarded as tentative and subject to a process of interrogation and progressive refinement. This creates a way to contextualize the numbers by going outside of census sources into sources of quantitative and non-quantitative information. The approach adopted here is to multi-source  those categories in this manner in order to create flexibility in the task of strategically designing categories and essentializing the data sets that correspond to them. Taking this approach requires having in place a methodology in place for adjudicating the discrepancies in the data that inevitably  will arise.

In order to develop this alternative approach, three survey compilations of race data in the countries of origin have been selected as starting places: the Central Intelligence Agency’s World Fact Book (CIA, 2012); the Britannica Book of the Year (BBY, 2007); the European Regional Survey (ERS, 2007).  These surveys employ their own research apparatuses, draw on data from census and non-census sources, are widely cited and are significantly independent of each other’s data. All three surveys present quantitative data in the form of  percentages that correspond to categories of national populations. In the case of the CIA and BBY, the percentages are presented without historical commentary. ERS provides a discursive historical  commentary on its percentages. This data is introduced under various rubrics. For ERS it is “Population.” For the CIA it is “Ethnic Groups.” For BBY it is “Demography,” with the subcategory of “Ethnic composition.” The data sets as such are not replicable, that is, they cannot be reproduced because it is almost without exception not clear how the percentages were derived and from where. The only exception is when the surveys, on rare occasions, state that their data is from a particular national census.

The categories that are most frequently used in the surveys to classify country of origin  populations are these, in descending order of frequency: white, mestizo, Native American Indian, black, mulatto. Table 3 presents the ratios between actual use and total possible use of the key categories.

In the Actual Use columns the denominator 60 represents all of the surveys using the category for all of the countries (3 surveys x 20 countries). The numerator is the actual times the category was used by all of the countries.  For example, the CIA used the white category for 16 countries instead of all of the 20 of them. BBY almost hit a home run with the white category (19).  In the Actual Use rows, the denominator (100) represents all of the times that each survey might have used the all of the (5 categories X 20 countries).  The actual use frequencies for the rows and columns converge at 193/300 (64.3%) which represents the aggregate percentage of the time that the categories are used.

These categories reflect the basic racial structure of the country of origins from which Hispanics are descended. This will be referred to as the 3R+2 structure. The “3” stands for the three primary ancestral races which originally peopled present day Latin America. These three ancestral races originated on separate continents:  Native American Indians (Siberia), White (Europe), Black (Africa). Slaves came from Africa. Colonists, conquerors and imperial agents came mainly from Spain. These quantitatively measured migrations (Curtin, 1961; Sanchez-Albornoz, 1994; Morner, 1976b; Daniels, 1990) began in the 1500s and used a trans-Atlantic migration route. The first Siberian migration into the Americas began 20,000 to 15,000 years ago. It was probably amphibious and shunted Siberians down the Pacific coasts of North and South America. This migration gave rise to tremendous urbanized population concentrations of the ancestors of  today’s Native American Indians in the territories of  Mexico, Central America and the Andean countries of South America. By contrast, the Native American Indian populations in North America, including the archaeologically well-known Clovis, were largely nomadic and thinly distributed. These populations arose from a probable later Siberian expansion which followed a multidirectional, interior route as opposed the basically unidirectional coastal route followed by the first migration. (Schurr, 2004; Balter, 2008; Bonnichen et al, 2006; Bryan, 1976; Dillehay, 2000; Fagundes et al. 2008;  Meltzer, 2009).

Over time, large-scale genetic mixing of these three source populations occurred.  Notwithstanding, the present day country of origin persons who are primarily of African, European or Siberian descent remain physically distinguishable from each other. The populations which are of mixed descent are also physically distinguishable in their own right and they principally fall into two groups: mestizos (descendants of the Native American Indian and white admixture), and mulattoes (descendants of the black and white admixture). These are represented by the “2” in the 3R+2 structure [see endnote #4]. An O category will represent segments of the population that do not fit into the 3R+2 categories.

Using the three surveys as starting places to construct replicable data sets corresponding to the 3R+2 categories requires some not inconsiderable adjustment of the survey data to the 3R+2 categories. Frequently a 3R+2 category is not used in the surveys because a surrogate term is used, for example, “of European descent” instead of white. A 3R+2 category may be “zeroed out” simply because not enough people are found to be classifiable into it. (The 3R+2 structure assumes that in all of the countries of origin, all of the 3R+2 categories have at least some countable portion of the population that fits into them.) The mulatto category is frequently not used in the surveys for this reason (See Table 3). Also, 3R+2 categories are frequently used in ways that blur or deviate from the strict definition of those categories on which the 3R+2 structure is based.  For example, a 3R+2 category may be used but there is no exact quantitative value given in the surveys  to it. Often this is because  it is grouped with other categories of people into a catch-all type of category to which a global quantitative value is assigned. A frequent example of this is when the white category is subsumed into the mestizo category because of extensive mixing of whites into the Native American Indian population. The category mestizo itself is sometimes used to cover all manner of racial mixtures, including even mulatto, as opposed to being confined, as in the 3R+2 structure, only to the white-Native American Indian admixture. Another example of fusion is the black category which is not broken down into black and mulatto categories as required by the 3R+2 structure.  Further, the black category in the surveys often includes individuals who are black-Native American Indian mixtures, sometimes called Zambo. This category lies beyond the 3R+2 structure because mixed racial categories are strategically confined to the  principal mestizo and mulatto categories. In the Black Carib (Garifuna) case,  the  cross between English-speaking escaped black slaves from the British Caribbean islands and Native American Indian populations in Central America. This case carries the additional complication from the 3R+2 perspective that individuals in this category are today not assimilated into the Hispanic culture. The 3R+2 structure is strategically attuned to Hispanics. If a Garifuna, for example, immigrated to the U.S. from, say Nicaragua, he or she would not be likely to assimilate into Hispanic culture in the United States, nor would  many Hispanics in the U.S. from Nicaragua draw their descent line to the Garifuna.

Given these kinds of problems is often necessary to infer 3R+2 values from the survey.  The linked to Appendix B presents the inferential 3R+2 adjustments made in the survey data for all of the Hispanic countries of origin.


The  cases of Guatemala and Puerto Rico provide illustrate the adjustment method.


Guatemala    problem: ERS says that “Guatemala has the largest surviving indigenous population in  Central America. The normally dominant mestizo population of mixed Spanish descent (including assimilated Amerindians) here largely known as Ladinos accounts for barely .5500 of the total. Often included with the Ladinos are the small population of more purely Spanish descent and of other non indigenous origins, such as Syrians and Lebanese, Asians and the Garifuna settled on the Caribbean coast.” (parentheses in the original). The adjustment problem here is that the 3R+2 category W along with the 3R+2 O category are mixed into the 3R+2 MZ (called Ladino) category.


solution: Ladino is equivalent to the 3R+2 MZ category except that the  non-indigenous population (non-European immigrants and Garifuna) and pure white European population are being mixed into it. Since ERS sees these pure whites as extremely marginal to the Ladino value, it would be misleading, to state a ERS W value as .5500. The same is true for the Syrians, Lebanese, Asian and Garifuna populations which are all non-3R+2 categories. The solution chosen for adjusting ERS data to the 3R+2 categories is to add the given Ladino value of .5500 to the given Amerindian value of .4300 to get a total of .9800. The remaining .0200 population can be taken to signify the  extra-Ladino non-indigenous and white populations so that W.0200 and O=0200.  O includes the Syrians, Lebanese, Asians and Garifuna.



Puerto Rico

problem: A “mixed” population value of .0440 and a large “other” value of .1200 are given in the CIA survey. An Asian value of .0030 is also given.  The problem is to adjust  the two unidentified categories of “mixed” and “other” to the 3R+2 ML and MX categories, and incorporate Asian value into the 3R+2 O category.

Solution: “Mixed” in Puerto Rico can mean mestizo or mulatto. The latter is vastly preponderant given the relatively marginal Amerindian  presence. Therefore, the other value of .1200 can reasonably be applied to the 3R+2 ML  category, and the .0440 to the 3R+2 MZ  category. This would make the 3R+2 Other category equivalent to the  the Asian component. Thus, ML=.1200, MZ=.0440, O=.0030.


Appendix C , located in the Tables section following the tables, a spreadsheet of the survey data for all of the countries as  adjusted to the 3R+2 categories. It provides a preliminary view of the 3R+2 landscape in Spanish America.

Having adjusted  all of the survey data in to the 3R+2 categories and the  Other category, the problem now is to scrutinize, or what will be termed “adjudicate,” the adjusted data in the surveys in order to resolve issues of discrepancies and reliability.  The adjudication format  has five components. First, the range of the adjusted decimal values in the surveys for each 3R+2 category is stipulated. Second,  the adjudicated value is stipulated.  Third, the context is indicated in which the adjudicated value is determined including the specific operations, such as computations and/or using  proxy values. The adjudication context may use one of the survey values or an average of them. More frequently it will refer to sources of quantitative data that do not coincide with survey values, and also to non-quantitative scholarly articles and books that can be related to the 3R+2 categories in the Hispanic countries of origin.  Indispensable to researching the  latter was the old standby of Latin americanists, the Library of Congress Handbook of Latin American Studies (HLAS), now available online. The outside- of-the survey sources used are cited in the Specific Bibliography. Fourth, the distribution factor. This represents the amount by which the sum of the adjudicated decimal values for a category falls short of 1 or exceeds it. Since these adjudicated values will be treated as probabilities in the next section they must sum to 1. To even the decimal sums as much as possible to 1, the distribution factor is distributed across the individual decimal  values according to their respective “weights.” Thus, the category that has the largest decimal value is therefore going to absorb most of the distribution effect. All of the computations involved in the derivation of distribution factors and weighting the 3R+2 decimal values for each country of origin contained in the linked Replication Data File 1.


Following is an illustration of the adjudication method applied to the white population in Puerto Rico. Puerto Rico is selected because the analysis of the “whitening process” reflects the efforts of the U.S. to reduce multi-race relations on the island to the question of how many are white and how many are black.  This is the pattern that Directive 15 originally put in place for sorting out Hispanic race in the United States.


                                  PUERTO RICO


Distribution Factor: .0738

Marginal: .0057


*     *     *     *     *     *


W (ADV: .5742)

Survey range: .7210-.8050

Pre-Distribution Value (PDV) .6200

Context: The U.S. Census reports a .7586 W value. The W PDV overrides this and instead uses the 1899 War Department population survey W value of .6200. This is a radical step but it is done to reflect the findings or Loveman and Muñiz’s analysis of “How Puerto Became Whiter”especially in the first two decades after the island had come under U.S. occupation. Driving this whitening process were the expansionary “changes in the social definition of whiteness” due to  “the specter of what might become of their society were their colonizers to see Puerto Rico as predominantly non-white. The shadow of Jim Crow hung over the island of Puerto Rico in the early twentieth century, a constant reminder of what it means to be non-white under the rule of the United States.” From 1899 to 1950 Loveman and Muñiz observe  that the white share of the population increased from 60 to 80 percent thereby making the population statistically “much more racially homogeneous, much whiter than the population of the mainland United States.”They observe that even with the revision of the race question in the 2000 census allowing respondents to choose more than one race only 4 percent of island Puerto Ricans marked on the U.S. Census more than one race, comparable to Puerto Rican Hispanics in California and New York. (Loveman and Muniz, 1-3, 5). A related argument holds that  “… the U.S. Government attempted to divide the  Puerto Rican population neatly into ‘two main classes, pure whites and those who are not.’. . . In short, the U.S. government sought to apply a binary race model to a fluid multi racial situation in Puerto Rico.” By dropping the category of mulatto the 1930 U.S Census “accelerated the movement from nonwhite to white categories on the island.”(Duany, 2002; internal quote is from the Department of War, 1900).

Given the inflationary effect of U.S. rule on the W value in Puerto Rico,  the War Department’s original estimate is taken as  a more accurate, objective indicator than the subsequent census indicators..


The application and result of the  adjudication method, as described,  to each of the the Hispanic  countries for each of the 3R+2 categories can be found in the linked Replication file 2.
The tabulation of the adjudicated 3R+2 values for all of the Hispanic countries is  given  in Table 4.


Monitoring by Probability


The information system developed in the preceding section is intended to play a role in a facilitating new interlocking behavioral and outcome contracts between the Bureau of the Census and Hispanics, ones that will  influence the latter to take fuller account of Hispanics’  racial backgrounds in their countries of origin. The purpose of this section is to explore how this system may be used to monitor the functioning of the contracts by measuring Hispanics’ actual outcomes in choosing their race against probable outcomes given their countries of origin. The latter are generated by using the 3R+2 distributions in countries of origin to create expected (probable) distributions by Hispanics from those countries. This assumes, of course, that the Census Bureau does make available to Hispanics the 3R+2  categories recommended in this article. The actual data on Hispanic distributions would then become available in Census 2020 but also in the run up to the Census by testing Hispanic samples. This could be conducted by the Census Bureau’s ongoing statistical survey, the American Community Survey.

Comparing Hispanic 3R+2 probability distributions with actual Hispanic 3R+2 distributions will permit measurement of the deviations. These deviations are  to be expected. First and foremost is the fact that immigrant selection in the countries of origin is not a random (i.e., probabilistic) process but a deterministic one, i.e., one driven by causal factors that are biased in the way they draw on the racial categories of the national populations. In other words, we do not expect the racial makeup of Hispanic immigration flows to the United States  to be a function of the racial makeup of the sending country’s population. Over time, however,  random variables will affect the racial composition of the  immigrant flow, tending to bring it into line with the sending country racial structure. Cuba illustrates this. Until 1980 the immigration flow to the United States consisted of upper and middle class whites fleeing the Castro Revolution. In 1980 the Mariel boatlift produced a huge surge in the non-white (black and mulatto) participation in the immigrant flow. The racial composition of the immigrants successfully making it to the United States in subsequent boat lifts was randomly determined by the vicissitudes of storms, high seas and ever-present sharks. An additional random selection variable was the so called “wet/feet dry/feet” Coast Guard policy which granted visas to rafters who made it to dry land and denied it to the ones intercepted at sea. In 1994 the U.S. Interests Section in Havana instituted a lottery system for the distribution of visas to Cubans. This created another random draw on Cuba’s population for immigration to the United States. All of these randomizing factors suggest that the immigrant racial composition changed from the original white-only basis to a basis that began to mirror Cuba’s own multi racial structure. (Aguirre, 1976, 2002 [see endnote #5]; cf: Fulger, nd; Zovadny, 2003; Johnson, 2001).

As these random changes are registered in the various national immigrant flows, the Hispanics from one sending country are probably going to have 3R+2 distributions that resemble those of that country at least more than they resemble the 3R+2 distribution in another sending country. A randomly selected Mexican Hispanic is going to have a greater probability of being a mestizo than a randomly selected Dominican Hispanic; a randomly selected Costa Rican Hispanic will have a greater probability of being white than a randomly selected Salvadoran. etc. The main point, however, is that the country of origin decimal distributions are being employed as probabilities in this section not in order to predict what actual Hispanic 3R+2 distributions will be but to provide a standard for monitoring actual Hispanic distributions. New layers of interpretation can be developed from the monitoring  process about where Hispanics are going racially against the background of their  own racial heritage in Latin America instead of against the background of the magnetic attraction of the White-Alone and SOR categories of Hispanic “race.”

Among the monitoring questions that are raised by using Hispanic probability distributions  are these: Is the 3R+2 “control set” actually pulling Hispanics away from the white ethnic option in the direction of paying more attention to their genealogical information? Or, do the detected deviations indicate that Hispanics are in fact downgrading the white ethnic option but  not necessarily in a ways that would have them reproducing country of origin distributions?  Do the deviations occur more sharply among Hispanics from certain countries of origin but not others? Do deviations occur among Hispanics in certain of the American states but not others? Do the deviations occur more in certain of the 3R+2 categories than others? Do deviations, finally, suggest that the  Hispanic distributions in the American states and sending country distributions are not really independent of each other [see endnote #6]. For example, the probability of being black in Cuba may be less than the probability of a Cuban being black in Florida, but more than the probability of a Cuban being black in New York.  The Cuban black percentages in New York are likely to be higher than those in Florida where the Cuban white exile community is known for its racism against blacks generally and Cuban blacks in particular.  In other words, it is not probabilities that entirely determine the relative percentage of Cuban blacks in the 3R+2  Hispanic populations in New York and Florida, but differences in perceptions of racism.

The point remains, however, that without a standard against which to identify and measure deviations, the monitoring function of the Census Bureau’s information system could not be performed efficiently. The deviations would not even be recognizable as such.  The agency relationship between Hispanics and the Census Bureau would be weakened.

The first step in the procedure to be used to develop  Hispanic 3R+2 distributions given the countries of origin probabilities is to identify from Census 2010 data that measures how  the Hispanic populations in each state in the United State breaks down in terms of countries of origin. The Census 2000 data on Hispanic origins  was incomplete due to a misleading question on the form (U.S. Bureau of the Census, n.d., 2003) which was corrected in Census 2010.  The probable 3R+2 distribution for of each country of origin subgroup of the instate Hispanic populations is calculated by taking the number (n) of the Hispanics in each subgroup and multiplying it  by the probabilities (p_5) represented by the 3R+2 distribution in the corresponding country of origin. For example,  we can take the case of Mexican Hispanics in Alabama, where n=122,911. In order to estimate  how many of them are expected to be white, the probability of being white in Mexico (p=.1500) is  multiplied by 122, 911 to get an expected white value of 18,437.  The number of Mexicans in Alabama who are expected to be black is obtained by multiplying 122,911 by the probability of being black in Mexico (p=.0014) to get an expected black value of 172.  This procedure is repeated for the  NI, MZ, ML, and O subgroups.  In Table 5 the column totals represent  the expected 3R+2 distribution for all Hispanics in the state of Alabama. The computations for these expected Hispanic 3R+2 distributions are given in the SPSS replication data files corresponding to each state. These files are available at the following link to Replication File 3 (Thanks for computational help to my William Smith research assistant Emma Lowenberg).

The “Alabama” procedure is applied to all of the states in the United States.  Table 6 presents the results.

The expected overall  3R+2 distribution for Hispanics in all of the United States in Table 6  is somewhat different from the overall 3R+2 percentage distribution in the countries of origin.   Table 7  compares the differences in terms of percentages that are computed from the data in Table 6 here (United States) and Table 4 (Countries of Origin) in  the preceding section. Table 7 shows that the expected Hispanic 3R+2 population draws substantially less on the W , and substantially more on the MZ category. Mexico with its dominant MZ population, supplies the overwhelming portion of the Hispanic population in the United States, while significant  MZ sources are also provided by Honduras, Guatemala, Nicaragua, and El Salvador, Colombia and Ecuador. The three principal ML sources for Hispanics are Cuba, Dominican Republic, Puerto Rico. The main W sources for Hispanics are Mexico, Puerto Rico and Cuba.




This brief section will compare:  the current Census 2010 question on ethnicity and race; a version of a key proposed change currently under review by OMB;  the proposed change that is based on the current article.  The key questions on Census 2010 are 8 and 9 as presented in Figure 1.

Figure 1: Census 2010 Origin and Race Questions


Note. Please answer both questions 5 about Hispanic origin and question 6 about race. For this Census Hispanic origins are not races.


Question 5. Is person 1 of Hispanic, Latino or Spanish Origin?


⬜ No. Not of Hispanic, Latino or Spanish origin

⬜ Yes. Mexican, Mexican Am., Chicano

⬜ Yes. Puerto Rican

⬜ Yes. Cuban

⬜ Yes. Another Hispanic, Latino or Spanish Origin. Print origin, for example,

Argentinean, Colombian, Dominican, Nicaraguan, Salvadoran, Spanish

and so on.



Question 6.. What is person 1’s race. Mark x for  one or more races.


⬜  White

⬜  Black, African Am., or Negro

⬜  American Indian or Alaskan Native. Print name of enrolled or principal


⬜⬜⬜ ⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜


⬜  Asian Indian     ⬜  Japanese           ⬜  Native Hawaiian

⬜  Chinese            ⬜  Korean

⬜  Filipino             ⬜  Vietnamese       ⬜  Guamanian or Chamorro

⬜  Other Asian – Print race,  for           ⬜  Yes. Samoan

example, Hmong, Laotian, Thai,           ⬜  Yes. Other Pacific Islander –

Pakistani, Cambodian, and so on.                   Print race, Fijian, Tongan,

and so on.



Some Other Race. Print race.





Figure 2  presents one of of the six versions that is currently being studied by the Census according to Prewitt (Prewitt, 2013; Prewitt substitutes 8 and 9 for the Census 2010 questions 5 and 6 which are adhered to here)


Figure 2: Current Proposed Change to Census Questions 5 and 6


Please answer both question 5 and 6 about race and origin


Question 5. What is person 1’s race or origin? Mark x for one or more boxes.


⬜ White

⬜ Black, African American, or Negro

⬜ Hispanic, Latino, or Spanish origin

⬜ American Indian or Alaska Native

⬜ Native Hawaiian or Pacific Islander

⬜ Asian

⬜ Some Other Race or origin


Question 6. Write in Person 1’s specific race, origin or enrolled or principal tribe. For example African American, Argentinean, Chinese, Egyptian, German, Marshallese, Mexican, Mongolian, Native Hawaiian, Navajo, Nigerian , Tlingit, and so on. Write in specific race(s), origin(2)  or tribes.


The proposed change in Question 5 combines the previously separate questions on   race an ethnicity into one question by placing the category “Hispanic” into the lineup of the official OMB races. The Bureau is leaning strongly toward effecting this combination. (U.S. Bureau of the Census, 2014). This will require a clarification of  long-standing position of the OMB that Hispanics are an ethnic group not a race. The stem tries to fudge this by stating  that the official  lineup is of races “or” origins, in order to give the OMB an out of sorts. But the five official races in the lineup are OMB races and it certainly seems that the category “Hispanic” is taking its place as an OMB race. Prewitt observes that Question 5 treats the six groups “as they are treated in law and policy, as they appear on many official forms, are described in the media, studied in public health and the social sciences, and understood by the general public” (Prewitt, 2013, 303). Certainly the category Hispanic is widely treated as a de facto race. Prewitt therefore supports giving the official status of race to the category Hispanic.

Various problems arise from placing Hispanic race in the official lineup. If a Hispanic Dominican chose Hispanic as her race in Question 5 and then Dominican Republic as her origin in Question 6  this  would seemingly make this Dominican belong to the Hispanic race. But is Hispanic race generic to all Hispanics? Do Argentineans and Dominicans belong to the same Hispanic race?  If our Dominican subject chose race in Question 5 as Hispanic combined with white does that help specify Hispanic race?  In this case,  Hispanic race borrows its racial content from white since it has none of its own. Hispanic  race then becomes sort of a  “wild card” in the official deck of races. What if the Domincan subject chose just the white category in Question 5? Then we would have a Dominican who is racially white,  but in the United States sense in not in any Hispanic sense because the option of Hispanic race was not chosen in Question 5. This Dominican subject  would have in effect traded membership in a Hispanic race for becoming white in the United States, preferring to ignore the fact  that being white in the Dominican Republic, with its largely mulatto population, is a different matter from being white in the United States where the mulatto category does not exist and the white choice is heavily tied to the maintenance of the white/binary.

Unless the content of Hispanic race is specified it is meaningless as an identifier for Hispanics. Putting Hispanic as a race in the official lineup may attract some Hispanics out of the SOR category but would it do so meaningfully? Basically, Hispanic race in the official lineup would be simply a surrogate for the Some Other race category, i.e.,  Hispanic race remains some unidentifiable other, not something in itself.



Figure  3 presents this article’s proposal for change in the census questions on Hispanic race and origin. This would replace question 5 on Census 2010


Figure 3: A Content-Based Proposed Change in Hispanic Race


For persons of Hispanic origin. Answer both questions A and B below. Note:  Hispanic race and origin are interconnected. Hispanic race is linked by descent from the primary racial structures in their countries of origin. The racial categories here have been carefully selected as the ones that are most commonly used by all of the Hispanic countries of origin. The category of mestizo refers to a combination of white and Native American Indian. The category of mulatto refers to a combination of white and Afro-American.


A., What is this person’s’ country of origin or descent? Write in specific country or countries?



  1.     What is this person’s race? Mark x for one Hispanic race.


⬜ White

⬜ Black

⬜ Native American Indian

⬜ Mestizo

⬜ Mulatto

⬜ Some other race. For example, Asian or Pacific Islander. Write in some other race.




This proposed change departs  from OMB rules in ways A and B.

A)Hispanics do have a determinate racial structure tied to their origins. Directive 15  language that Hispanics “can be of any race” and Census Bureau language that “Hispanic origins are not races“ would have to be clarified to mean that Hispanic countries of origin are not single races but do represent a unique  combination of races.

B) It is basic to OMB procedure that all federal agencies which report racial statistics must at a minimum employ the five official races and their multiracial combinations. Agencies can introduce subdivisions of the primary categories but they must be aggregatable into the primary categories. They cannot present groups with a number of categories that is less  than the  official ones. In this proposal, Hispanics  are being presented with only three of the official  categories.  The “new” categories of mestizo and mulatto, however, already exist as two possibilities in combinatorial set of  57 possibilities. They are not new in that sense although  terminologically  they are. The main difficulty of the proposal is that Hispanics  are being presented with a number of categories which is less than the official number. The two  “absent” official  races (Asian and Pacific Islander) are however  being made available to Hispanics in the  “some other race” category. Asian and Pacific Islander categories, however, are not considered as belonging to the Hispanic race.

The ultimate value of this proposal is that it adds substantive content to Hispanic race. It proposes a genuinely multiracial basis on which Hispanics can coherently construct their race in accordance with their racial heritage. This will deter Hispanics from assimilating into the white and black sides of the binary thereby weakening its hegemony over race in the United States.



The Specific Bibliography includes just the sources used in deriving the country of origin data sets. This General Bibliography includes all other sources used.

Aguirre, Benigno and E. Bonilla Silva. 2002 .“Does Race Matter Among Cuban Immigrants? An Analysis of the Racial Characteristics of Recent Cuban Immigrants.” Journal of Latin American Studies 34(2): 311-324.

____________. 1976 “Differential Migration of Social Races from Cuba.” Latin American research         Review 11(1):103-124.

Akerlof, George, 1970. “The Market for Lemons: Quality Uncertainty and the Market Mechanism.“           Quarterly Journal of Economics (MIT Press) 84:(3):488-500

Alonso, Ricardo and Niko Matouschek. 2008. “Optimal Delegation.” Review of Economic Studies  75:259-293.

Bashi, Vilna. 1997. “A Theory of Immigration and Racial Stratification.” Journal of Black Studies. 27(5):668-82

Bendor, Jonathan and Adam Meirowitz. 2004. “Spatial Models of Delegation.” American Political Science Review 98(2):293-310

______________, A. Glazer,T. Hammond. 2001. “Theories  of Delegation.” Annual Review of Political Science 4: 235-239.

Besley, Timothy. 2007. Principled Agents? The Political Economy of Good Government. New York: Oxford University Press

Brehm, John and Scott Gates. 1997. Working, Shirking and Sabotage. Bureaucratic Response to a Democratic Public.  Ann Arbor: University of Michigan Press.

Crawford, Vincent P., and Tod Sobel. 1982. ”Strategic Information Transmission.” Econometrica 50(6):1431-1451.

Demski, J. S. and David E. Sappington. 1999. “Summarization with Errors: A Perspective on Empirical Investigations of Agency Relationships.” Management Accounting Research 10: 21-37.

Dessen, Wouter. 2002.”Authority and Communication in Organizations.” Review of Economic Studies. 69: 811-838

Eisenhardt, Kathleen.  1988. “Agency Theory: An Assessment and Review.” The Academy of Management Review 14(1): 57:74.

_________________. 1985. “Control: Organizational and Economic Approaches.” Management Science 31: 134-149.

Epstein, David and Sharyn O’Halloran. 1994. “Administrative Procedures, Information, and Agency Discretion: Slack Versus Flexibility.” American Journal of Political Science 38:697-722.

_______________________________. 1999. Delegating Powers. New York: Cambridge University Press.

Gailmard, Sean and John Patty, 2013. Learning While Governing: Expertise and Accountability in the Executive Branch. Chicago: University of Chicago Press

Gailmard, Sean. 2002. “Expertise, Subversion and Bureaucratic Discretion.” Journal of Law, Economics and Organization 18: 536-555.

Glover, Thomas and Kevin Mitchell. 2008. An Introduction to Biostatistics. Second edition. Long Grove, IL: Waveland Press.

Harsanyi, John C. 1967-1968. “Games of Incomplete Information Played by Bayesian Players.” Management Science 14(3): 159-182. Reprinted, ibid. 50(12): 1804-2004.

Hattam, Victoria. 2005. “Ethnicity and the Boundaries of Race: Reading Directive 15. Daedalus 134(1): 61-70.

_____________.  2007. In the Shadow of Race: Jews, Latinos and Immigrant Politics in the United States. Chicago: Chicago: University of Chicago Press.

Haney-López, Ian. 2004. ”The Birth of a Latino Race.” Los Angeles Times, December 29.

______________. 2005. “Race on the 2010 Census: Hispanics and the Shrinking White Majority.” Daedalus 134(1):42-53

Heclo, Hugh, 1999. “OMB and Neutral Competence.” In James Pfiffer, ed. The Managerial Presidency, 123-143. College Station, TX: Texas A&M University Press.

Holmstrom, Bengt. 1984. “On the Theory of Delegation.” In Marcel Boyer and Richard Khilstrom, eds., Bayesian Models in Economic Theory. New York: North Holland

_______________. 1979. “Moral Hazard and Observability.” Bell Journal of Economics 10: 74-91

Jensen, Michael and William Meckling.  1976. “Theory of the firm: Managerial Behavior, Agency Costs, and Ownership Structure.” Journal of Financial Economics 3: 305-360

Johnson, Kevin R. 2001.”Comparative Racialization: Cultures and National Origins in Latina/o Communities.” Denver University Law Review 78:633-655.

Laffont, Jean-Jacques and David Martimort. 2009 The Theory of Incentives: The Principal-Agent Model. Princeton, N.J.: Princeton University Press

Landry, Gerald  and Donna Maclean. 1996. The Spivak Reader. New York: Routledge

Lee, Jennifer and Frank Bean. 2010. The Diversity Paradox: Immigration and the Color Line in the Twenty First Century America. New York: Russell Sage Foundation.

Lupia, Arthur and Matthew McCubbins. 1998. The Democratic Dilemma. Can Citizens Learn What They Need to Know? New York: Cambridge University Press.

Mora, Christina G. 2014. Making Hispanics: How Activists, Bureaucrats and Media Constructed a New American Category.  Chicago: University of Chicago Press.

Miller, Gary. 2005. “The Political Evolution of  Principal-Agent Models.” Annual Review of Political Science 79: 1094-1116.

Navarro, Mireya. 2003. “Hispanics Choose ‘Other.’ “ Beyond Black and White.” New York Times, November 9.

Office of Management and Budget. “Revisions to the Standards for Classification of Federal Data on Race and Ethnicity.” Federal Registry Notice, October 20, 1997.

____________________________. Directive 15, May 12, 1977. Text, Federal Register July, 9, 1977.

PEW Hispanic Center. 2011. “National Survey of Latinos.”

Posner, Eric A. 2000. “Agency Models in Law and Economics.” The Coase Lecture. University of Chicago Law School.

Prewitt, Kenneth H.  2005, “Racial Classification: Where Do We Go from Here?” Daedalus. 134(1):5-18.

_______________. 2013a.”Fix the Census’s Archaic Racial Categories.” New York Times, Op-Ed, August 22.

_______________. 2013b. What Is Your Race? The Census and Our Flawed Efforts to Classify Americans. Princeton: Princeton University Press.

Ross, Stephen. 1973. “The Economic Theory of Agency: The Principal’s Problem.” American Economic Review 63: 134-139.

Sappington, David E.M. “Incentives in Principal-Agent Relationships.” Journal of Economic Perspectives 5(2):45-66.

Shapiro, Susan P. 2005.  “Agency Theory.” American Review of Sociology 31: 363:-84

Skerry, Peter. 2002. Counting on the Census: Race, Group Identity, and the Evasion of Politics. Washington, DC: Brookings Institute Press.

____________. 2009. Review of Hattam, In the Shadow of Race. Journal of Interdisciplinary History 39(4):612-614

Spence, A. M.. 1971. ”Insurance, Information, and Individual Action.” American Economic Review 61: 380-387.

Spivak, Gayatri. 1985. “Subaltern Studies: Deconstructing Historiography.” In The Spivak Reader, Donna Landry and Gerald Maclean, eds, 203-219.

____________. “Can the Subaltern Speak?” In Patrick Williams and Laura Chrisman, eds. Colonial Discourse and Post-Colonial Theory. New York: Columbia University Press, 66-111.

Steinberg, Stephen. 2009a. Race Relations: A Critique. Stanford, CA: Stanford University Press.

________________ 2009b. Review of Hattam, In the Shadow of Race. Perspectives on Politics 7(1): 189-191.

Stephenson, Matthew. 2007. “Bureaucratic Decision Costs and Endogenous Agency Expertise.” Journal of Law , Economics, and Organization 23(2): 469-498 .

Telles, Edward, Mark Q. Sawyer and Gaspar Rivera-Salgado. 2011. Just Neighbors? Research on African-American and Latino Relations in the United States. New York: Russell Sage Foundation.

United States Census Bureau. 2000. Redistricting Summary File. Tables Pl1 and Pl2.

Census 2000

______________________________. 2001. “Overview of Race and Hispanic Origin.” Census Brief. 2000. Elizabeth M. Grieco,  Rachel C. Cassidy

_________________________________,  2001 QT-P9. Hispanic or Latino by Type.  Census 2000. PCT 11. Summary File 1.

________________________. 2002. Modified Race Data Summary File.

_____________________________. n.d. 2003?. Population Division. “Analysis of General Hispanic Responses to Census 2000. Arthur R. Cresce and Roberto R. Ramirez. Working Paper 72.

_________________________________ 2010. Redistricting Summary File. Table 1 and 2


_______________________________. 2011a. “Overview of Race and Hispanic Origin.” Census Brief.  2010. Karen R. Humes, Nicholas A. Jones, Roberto R. Ramírez

_________________________________. 2011b. QT-P10. Hispanic or Latino by Type: Census 2010. Summary File 1.

____________________. 2014. “Race Reporting Among Hispanics: 2010.” Merarys Rios, Fabian Romero, Roberto Ramirez. Working Paper No. 102. Population Division.

Waterman, Richard and Kenneth Meier. 1998. “Principal-Agent Models: An Expansion.” Journal of Public Administration and Theory.8(2):173-202.

Zovodny, Madeline. 2003. “”Race, Wages, and Assimilation among Cuban Immigrants.” Population Research and Policy Review 22: 201-219.



Following are the sources used specifically in connection with derivation of racial structures in Latin America, i.e., the section on  country of origin data sets. .

Aguirre, Beltran. 1952. “Ethnohistoria de la población negra en Mexico.” In Sol Tax, ed. Accultration in the Americas. Proceedings and Selected Papers of the XXIXth International Congress of Americanists. Chicago: University of Chicago Press, 161-168.

_____________. 1946. La población negra en Mexico, 1519-1810. Mexico City: Editorial Fuente Cultural.

Aidi, Hisham. 2002. “Blacks in Argentina: Disappearing Acts.”  It is posted at

Allocco, D.J. et al. 2007. “Geography and Genography: Prediction of Continental Origin Using Randomly Selected Single Nucleotide Polymorphism.” BMC Genomics 8:68, March 10.

allRefer Reference. Columbia Encyclopedia. Country Studies

Andrews, George R. 2004. Afro-Latin America 1800-2000. New York: Oxford University Press.

______________. 1976. “Race Versus Class Association.” Journal of Latin American Studies. 11(1): 19-39,

______________. 1972. “The Afro-Argentine Officers of Buenos Aires Province, 1800-1860.” Journal of Negro History 64(2): 85-100, Spring.

Archibold, Randal C. 2014. “Negro? Prieto? Moreno? A Question of Identity for Black Americans.” New York Times. October 25.

Avena, Serio A. et al. 2006. “Mezcla genética en una muestra poblacional de la ciudad de Buenos Aires.” Medicina (Buenos Aires) 66(2): 113-118, March/April.

Balter, M. 2008. “Ancient Algae Suggest Sea Route for the First Americans.” Science 320(5877): 784-786..

Bamshad, M.J. et al. 2003 “Human Population Genetic Structure and Inference of Group Membership.” American Journal of Human Genetics 72: 578-589.

___________.  et al. 2004. “Deconstruction the Relationship between Genetics and Race.”  National Review of Genes 5:598-609.

Barriero, José. 2006-2007. “In Cuba, the Cry Was for Rain.” Issues in Caribbean Amerindian Studies 6(1), December.

Basañez, Miguel and Pablo Pars. 2001. “Color and Democracy  in Latin America.”  In Camp, 139-153.

Berg, Kate et al. 2005. “The Use of Racial, Ethnic, and Ancestral Categories in Genetics Research. American Journal of Human Genetics. 71: 519-532.

Bonnichsen, Robson et al. 2006. Paleoamerican Origins: Beyond Clovis. College Station, TX: Center for the Study of the First Americans

Bourgois, Phillipe. 1986. “The Miskito of Nicaragua: Politicized Ethnicity.” Anthropology Today. Royal Anthropological Institute, London. 2(2):4-9, April.

Boyd-Bowman, Peter. 1976. “Patterns of Spanish Emigration to the Indies until 1660. Hispanic American Historical Review 66(4):580-604.

Britannica Online Encyclopedia.

Bryan, Alan Lyle ed. Early Man in America; From a Circum-Pacific Perspective. Occasional Papers No. 1 of the Dept. of Anthropology. Edmonton, Canada: University of Alberta, 1978.

Bryc, Katarzyna et al. 2010. “Genome-wide Patterns of Population Structure and Admixture among Hispanic/Latino Populations.” Proceedings of the National Academy of Science (PNAS) 107 (Supplement 2): 8954-8961.

Bucheli, Marisa and Wanda Cabela. 2006. “Perfíl demográfico y socioeconomico de la población uruguaya segun su ascendendia racial.” Montevideo, Uruguay. Institución Nacional de Estadistica INE).

Camp, Roderic Ai, ed. 2001. Citizen Views of Latin America. Pittsburgh, PA: University of Pittsburgh Press

Comissión Economica Para America Latina (CEPAL). 2008. “Pueblas Indiígenas.”  Observatorio Demográfico No. 6, October.

Cerda-Flores, Norberto. 1987. “Gene Admixture and Distances between Populations from Monterrey, Nuevo León for Their Putative Ancestral Populations.”  Human Biology. 59(1): 31-50, February.

Chapin, Mac. 1991.“The Indian Population of El Salvador.” Mesoamerica 12(21): 1-40, June.

Columbia Encyclopedia. 6th Edition. Online.

Columbia Gazetteer of the World. 1998. Saul B. Cohen, ed. New York: Columbia University Press. 4 vols.

Congressional Report for Congress. 2008  Afro-Latinos in Latin America and Considerations for U.S. Policy. Washington: Congressional Research Service.

Cook, Howard. 2001. Coloring the Nation: Race and Ethnicity in the Dominican Republic. Boulder, CO: L. Rienner.

Coon, Carleton S.1973.  Origin of the Races. New York: Knopf. This advances the startling but now largely discredited thesis that human races are older than our species, being rooted in prior geographical separations of the homo erectus populations.

Coon, Carleton S. and Edward E. Hunt, Jr. 1966. The Living Races of Man. New York: Knopf

Crawford, Michael et al. 1981. “Black Caribs (Garifuna) of Livingston, Guatemala: Genetic Markers and Admixture Estimates.” Human Biology 53(1): 87-103, February.

Curtin, Philip D. 1961. The African Slave Trade: A Census. Madison, WI: University of Wisconsin Press.

Daniels, Roger. 1990. Coming to America. New York: Harper Collins.

De Palma, Anthony. 1998.“Cuban Site Casts Light on Extinct People.” New York Times, July 5.

Díaz, Edwin. 1992. “The Indigenous Population of Guatemala According to the 1981 and 1994 Censuses.” Santa Monica, CA: Rand. Also in Pebley. Chapter Six.

Dillehay, Thomas D. 2000. The Settlement of the Americas. New York: Basic Books,

Duany, Jorge. 1988. “Reconstructing Racial Identity: Ethnicity, Color, Class Among Dominicans in the United States and Puerto Rico.” Latin American Perspectives 25(3):147-172, May.

___________. 2002. The Puerto Rican Nation on the Move. Chapel Hill, NC: University of North Carolina Press.

Duthurbura, José Antonio del. 2001. Breve historia de los negros del Peru. Lima: Fondo Editorial del Congreso.

Early, John D. 1975. “The Changing Proportion  of Mayan Indian and Ladino in the Population of Guatemala, 1945-1969.” American Ethnologist.  American Anthropological Association. Washington, D.C. 2(2): 261-269, May.

EcoBrazil. “Another Brazil.” Eco Brazil Revised 1995 by Ricardo Barthem.

El-Haj, Nadia Abu. 2007. “The Genetic Reinscription of Race.” Annual Review of Anthropology 36: 283-300.

Encyclopedia Britannica Online.

Ethnologue. Languages of the World. .

Europa Regional Surveys of the World. 2007. South American, Central America and the Caribbean.. 15th edition. New York: Routledge.

Fagundes, N.R. et al. , 2008. “Mitochondrial Population Genomics Supports Pre-Clovis Origin with a Coastal Route for Peopling  of the America.” American Journal of Human Genetics  82:583-592

Farmer, Patricia. 2004. “Cracking the Racial Divide: Tracing the Historical Connection of the Castizo, the Modern Mulatto vs. the Colonial Mexican.” Online article, May.

Florestan, Fernandes. 1969, The Negro in Brazilian Society. ed. Phyllis B. Eveleth. New York: Columbia Univeristy Press.

Franch, José Alcina. 1974. “El problema de las poblaciones negroides de Esmeralda, Ecuador.” Anuario de Estudios Americanos 31:33-36.

Gallardo, Jorge Emilio. 1989. “Étnias africanas en el Rió de La Plata.” Buenos Aires: Centro de Estudios Latinoamericanos, Ultimo Reino. 1989.

Glen, David. 2007. “A Sociologist Offers a Harsh Assessment of How His Discipline Treats Race Relations.” The Chronicle of Higher Education. November 16.

Gott, Richard. 1999.“A Question of Black and White.” New Statesman, London, 12(549): 22-23, April 2.

Goyer, Doreen S. and Elaine Domschke. 1983. The Handbook of National Population Censuses: Latin American and the Caribbean, North America, and Oceania. Westport, CT: Greenwood Press.

Graham, Richard, ed. 1990. The Idea of Race in Latin America, 1870-1940. Austin, TX: University of Texas Press.

Halder, Indrani et al. 2008. “A Panel of Ancestry Informative Markers for Estimating Individual Biogeographical Ancestry and Admixture from Four Continents: Utility and Applications.” Human Mutation 29(5): 648-658.

Harris, Marvin. 1964. Patterns of Race in the Americas. New York: Walker and Co.

Helg, Aline. 1990. “Race in Argentina and Cuba, 1880-1930: Theory, Policies and Popular Reaction” in Graham, ed., 37-69.

Hoetink, Harry. 1967. The Two Variants in Caribbean Race Relations: A Contribution to the Sociology of Segmented Societies. London:  Oxford University Press.

Howard, David S. 2001. Coloring the Nation: Race and Ethnicity in the Dominican Republic. Boulder, CO: Lynne Rienner.

Interamerican Dialogue. n.d. Race in Latin America.

__________________. 2003. “Race Report. Afro Descendants in Latin America. How Many?” Washington D.C: Inter-American Dialogue.,

IPUMS. Integrated Public Use Micro Data Series. IPUMS-international converts census data for multiple countries, including Latin America,  into a consistent format. Census data used by IPUMS for  Latin America  is very sketchy with regard to the 3R+2 format that is used in the present paper. It therefore had to be supplemented by other quantitative and non quantitative sources and “adjudicated” into consistency with the 3R+2 format with an emphasis on replicability. The data sets yielded by this methodology are arrived at  independently of the IPUMS converted census microdata. It is however considered productive to compare the data offered here with IPUMS data.

Jackson, Robert. 1996. “Bolivian Mestizos as a Case of Changing Demographic Knowledge.” Cahiers Quebecois de Demographic 25(1) Spring.

James, Preston E.  Latin America. 1959. Third edition. New York: Odyssey Press.

_____________.  Latin America. 1942. New York: The Odyssey Press.

Jewish Population of the World (1882-Present).

Johnson, Lyman L. “The Racial limits of Guild Solidarity : an Example from Colonial Buenos Aires.” Revista de Historia de America, Rev. Hist. Am./Mexico, 99: 7—26, enero/junio, 1985

Johnstone, Patrick and Jason Mandryk. 2001. Operation World. Harrisburg, VA: RR Connolly and Sons. 2001

Jorde L.B. et al. 2000. “The Distribution of Human Genetic Diversity: a Comparison of Mitochondrial, Autosomal, and Y Chromosome Data.” American Journal of Human Genetics:       979-988.

Joshua Project.

Kaup, Monika and Debra Rosenthal eds.. 2002. Mixing Race, Mixing Culture. Austin: University of Texas Press.

Kertzer, David I. and Dominique Arel, eds.  2002. Census and Identity: The Politics of Race, Ethnicity and Language in National Censuses. New York: Cambridge University Press.

_____________________________. Censuses, Identity Formation, and the Struggle for Political Power.  In Kertzner and Arel, eds. 1-42.

Kluck, Patricia. 1989. “The Society and its Environment.” Library of Congress Country Study: Ecuador, Chapter Two.

Kosov, R. et al. 2009. “Ancestry Informative Markers Set for Determining Continental Origin and Admixture Proportions in Common Populations of America.” Human Mutations 30(1): 69-78, January.

Lancaster, Roger N. 1991. “Skin Color, Race and Racism in Nicaragua.” Ethnology 30(4):339-353, October.

Landale, Nancy S. and R.S. Oropesa. 2002. “White, Black or Puerto Rican? Racial Self-Identifiction Among Mainland and Island Puerto Ricans”. Social Forces 81(1):231-254, September.

Latinobarómetro. Providencia, Chile.

Leons, Madeline Barbara. 1978. “Race, Ethnicity and Political Mobilization in the Andes.” American Ethnologist 5(3): 484-494, August.

Lewis, Marvin A. Afro-Argentine Discourse: Another Dimension of the Black Diaspora. Columbia, MO: University of Missouri Press, 1996

Liboreiro. Cristina de. 1999. No hay negros argentinos? Buenos Aires: Editorial Dunken, 2nd  Edition.

Library of Congress. Federal Research Division. Since 1948. Country Studies.

Lisker, Rubin. 1988. “Gene Frequencies and Admixture Estimates in the State of Pueblo, Mexico.” American Journal of Physical Anthropology 76: 31-35, July.

___________ et al. 1986.“Gene Frequencies and Admixture Estimates in Mexico City.”  American Journal of Physical Anthropology 71(2): 203-208, October.

Liu, Yushi. 2013. “Softwares and Methods for Estimating Genetic Ancestry in Human Populations.” Open Access article online at

Lizcano, Francisco Fernandez. 2004. “Composición étnica de las tres areas culturales del continente americano al comienzo de siglo XXI.” Mexico. Universidad Autonoma. Convergencia 38:185-232, May-August

Loveman, Mara and Jerónimo Muñiz. “How Puerto Rico Became White: An Analysis of Racial Statistics in the 1910 and 1920 Censuses.” Paper Prepared for Presentation at the Center for Demography and Ecology, University of Wisconsin-Madison, February 7, 2006 (

_______________________________.  2007. “How Puerto Rico Became White: Boundary Dynamics and Intercensuses Reclassification.” American Sociological Review 72: 915-939, December.

Martínez-Cruzado, Juan C.  2002. “The Uses of Mitochrondrial DNA to Discover Pre-Colombian Migrations to the Caribbean: Results for Puerto Rico and Expectations for the  Dominican Republic.”  KACIKE: The Journal of Caribbean Amerindian History and Anthropology (On-line Journal). Special Issue, Lyne Guitar ed., 2002.

Mao, Xianyun, et al. 2007.“A Genome-wide Admixture Mapping Panel for Hispanic/Latino Populations.” The American Journal of Human Genetics. 80: 1171-1178.

Medina Lois, Ernesto and Ana Mara Kaempffer R. Universidad de Chile. 1978. Elementos de Salud Publica.”  Section 5.2.6 “Racial Structure.”

Meltzer, David J. 2009. First Peoples in a New World: Colonizing Ice Age America. Berkeley, CA: University of California Press, 2009

Miller, Marilyn Grace. 2004. Rise and Fall of the Cosmic Race: The Cult of Mestizaje in Latin America. Austin, TX: University of Texas Press, 2004.

Minorities at Risk (MAR).

Morner, Magnus. 1967. Race Mixture in the History of Latin America. 1967. New York: Little Brown.

_______________1976a. “Slavery and Race in the Evolution of Latin American Societies.” Journal of Latin American Studies, 8(1): 127-135, May.

__________________. 1976b. “Spanish Migration to the New World Prior to 1810. A Report on the State of Research.”. In First Images of America: the Impact of the New World on the Old. ed Fredi Chigelli, 737-82, 797-804.

Nobles, Melissa. 2002. Race Categorization and Censuses  in Kertzner and Arel, 43-70.

_____________  2005. “The Myth of Latin American Multiracialism.” Daedalus 134(1):82-88, Winter.

Operation World. 2001. Patrick Johnstone and Jason Mandryk eds. Harrisburg, VA:  R. Connolly and Sons.

Pebley, Ann ed. 1997. Demographic Diversity and Change in the Central American Isthmus. Santa Monica, CA: RAND.

Peoplegroups. .

Perez, Lisandro. 1984.  “The Political Contexts of Cuban Population Censuses, 1899-1981.” Latin American Research Review 19(2): 143-161.

Pia, Josephina. 1972. Humano negro: la esclavitud en el Paraguay. Madrid: Paraninfo.

Poe, Janita. 2003. “Being Latin and Black.” The Atlanta Journal-Constitution.  August, 6.

Porter, Eduardo. 2003. “Census Forms Work Hard to Find Proper Way to Identify Hispanics.” Wall Street Journal, January 21.

Pritchard, J. K. et al..2000. “Inference of Population Structure Using Multilocus Genotypes Data.” Genetics 155: 945-959.

Rahier, Jeqan Muteba. 2004.  “The Study of Latin American Racial Formations: Different Approaches and Different Contexts.” Review essay. Latin American Research Review 39(3):282-293.

Restall, ed. 2005. Beyond Black and Red: African-Native Relations in Colonial Latin America. Albuquerque: University of New Mexico Press.

Rocco, Paola P. 2002. “Genetic Composition of the Chilean population. Analysis of Mitochondrial DNA Polymorphisms.” Revista Medica de Chile. 130(2):125-131, February 2.

Rodas, Isabel. 1997. “En la busqueda de la diversidad ladino.” Cultura de Guatemala 18(1): 139-162, April.

Rodriguez, Clara E. 2005. “What it Means to be Latino.”

Roitman, Karem. nd. “Hybridity, Mestizaje, and Montubios in Ecuador.”

Rosenberg, Noah A. et al. 2002. “Genetic Structure of Human Populations.” Science 298(5602): 2381-2385.

Safa, Helen I. 1998.“Race and National Identity in the Americas.” Latin American Perspectives 25(3):3-20.

Sailer, Steve. Race Now. 2002. Part 3: “Where Did Mexico’s Blacks Go?. Online article.

Sánchez-Albornoz, Nicolás. La Población de la America Latina. 2nd ed. Madrid: Alianza Editorial, 1994.

_____________________. “The Population of Colonial Spanish America.” In The Cambridge History of Latin America, 3-35. Cambridge: Cambridge University press

Sans, Monica. 2000. “Admixture Studies in Latin America: From the 20th to the 21st Century. Human Biology 72: 155-177.

Sater, William R. 1974. “The Black Experience in Chile” in Toplin, Slavery and Race Relations in Latin America, 13-50.

Seely, H. and Mara Guadalupe Mirón. 1970. “Phenotype and Occupational Mobility in Guatemala City: A Preliminary Survey.” Science Education 54(1): 13-16, Jan/March.

Shapiro, Harry. 1945. ”Ethnic Patterns in Latin America.” Scientific Monthly. 61(5): 345-352, November.

Smith, T. Lynn Smith. 1966. “The Racial Composition of the Population of Colombia,” Journal of Inter-American Studies 8(2): 212-235, April.

Staples, Brent. 2007. “On Race and the Census: Struggling with Categories That No Longer Apply.” New York Times, February 5.

Susnick, Bratislava. 1989. “Ethnohistoria de Paraguay.” America Indgena 44(3): 432-490.

Swarns, Rachel L. 2004. “Hispanics Resist Racial Grouping by Census.” New York Times, October 24.

Tang, H. et al. 2005, “Estimation of Individual Admixture: Analytical and Study Design Considerations.” Genetic Epidemiology  28(4): 289-301.

Telles, Edward E. 2007. “Incorporating Race and Ethnicity into the U.S. Millennium Development Goals.” Inter-American Dialogue. Race Report. January

____________ . 1998. “Does it Matter Who Answers the Race Question? Racial Classification and Inequality in Brazil.” Demography 35(4): 465-74. November.

____________ . 2004. Race in Another America: The Significance of Skin Color in Brazil. Princeton, NJ: Princeton University Press.

Tilly, Virginia. 2005. Seeing Indians: A Study of Race, Nation, and Power in El Salvador. Albuquerque: University of New Mexico Press, 2005.

Toplin, Robert Brent, ed. 1974. Slavery and Race Relations in Latin America. Westport, CT: Greenwood Press.

United Nations Demographic Yearbook. 1948-2013.

Uslar-Pietri, Arturo. 1975. “Crucible of Races.” Americas 23(3): 28-35, March.

Van Cott, Donna Lee 1994. Indigenous Peoples and the State in Ecuador. New York: St. Martin’s Press.

Vandiver, Marylee Mason. 1949. “Racial Classifications in Latin American Censuses.” Social Forces 28(2): 138-146.

Wade, Nicholas. 2004. “Articles Highlight Different Views on Genetic Basis of Race” New York Times October 27.

Wade, Peter. 1993. “ ‘Race’, Nature and Culture.” Man. March 28, 1993: 17-34

__________. 1993. Blackness and Race Mixture. Baltimore: John Hopkins University Press, 1993

__________.  1997. Race and Ethnicity in Latin America. Chicago: University of Chicago Press.

Warnke, Georgia. 2002. “Race, Gender and Antiessentialist Politics.” American Journal of Sociology 108(2): 406-439, September.

Williams, Kent C. El Salvador. 2001.  “Afromestizo: The Third Root.”

_____________. Guatemala. 2001.  “Afromestizo: The Third Root.”

Wright, Winthrop. 1990. Café con Lelche: Race, Class and National Image in Venezuela. Austin: Texas: University of Texas Press, 1990

World in Figures. Boston, MA: GK Hall and Co, 1988. Compiled by Economist.

Zona Latina. Racial  Classifications in Latin America.