Document Sample
UC SANTA CRUZ PAPER Powered By Docstoc
					Practical Research Strategies for Mexican Indigenous Communities in California Seeking
                              to Assert Their Own Identity

                                   Ilene J. Jacobs
                                   Edward Kissam

                                     Presented At:

                       Indigenous Mexican Migrants in the US:
            Building Bridges between Researchers and Community Leaders

                                October 11-12, 2002
                           UCSC Inn and Conference Center
                            611 Ocean St., Santa Cruz, CA

                                     Sponsored by:
                  Latin American and Latino Studies Department (LALS)
                          University of California, Santa Cruz

                                     Co-sponsored by:
                    Frente Indígena Oaxaqueño Binacional (FIOB)
                        Chicano/Latino Research Center (UCSC)
                  Center for Justice, Tolerance and Community (UCSC)

    Practical Research Strategies for Mexican Indigenous Communities in California Seeking
                                  to Assert Their Own Identity


We examine in this paper the role that data play in the efforts of communities of indigenous
Mexicanos to forge and assert their own identity and describe several practical strategies which
we believe will enhance these communities’ abilities to prevail in a policy formation and program
planning environment which is increasingly ―technical‖. We believe that it will be important for
indigenous organizations to build their capacity to carry out applied research projects. We are
not, in any sense, apologists for the status quo which hails ethnocentricity in contemporary
policy analysis and program evaluation as the ―gold standard‖ for public dialogue and decision-
making about our collective future, as communities, as a state, or as a nation. We refer to ―the
tyranny of sociology‖ that determines the syntax of public dialogue because, in contemporary
America, ―technical‖ expertise often is professional narrow-mindedness based on tabular data
which characteristically displaces the old-fashioned processes of governance in which decision-
makers actually had to listen to their constituents and respect their insights.1

We have observed with dismay the spread of a bureaucratic culture which gives ritual homage to
―data‖ in almost magical terms without understanding how data are generated and then uses this
magical ―data‖ primarily as a negotiating tool in perpetuating information gaps between ―haves‖
and ―have-nots‖. These concerns escalate for practical, social policy and service planning issues
which affect indigenous communities in California because they are minorities among minorities,
―outliers‖ for better or worse, whose perspectives, concerns, priorities, and needs, are obscured by
the power of large numbers where regression to the mean rules supreme, making California’s
indigenous immigrant populations ―marginal‖ in every sense of the term.

We do not believe that new applied research capacity will necessarily make indigenous
community organizations ―better‖ ones; however, it will be useful for them and their advocates
to speak the lingua franca of bureaucracy, to have the analytic and communication tools to engage
in dialogue and debate, for example, to comment on public agencies’ rule-making, contest
misplaced priorities, correct thoughtless assumptions based on false stereotypes, suggest more
innovative and effective social programs, compete for funding, and, where necessary litigate to
assure equitable access to more responsive services.

We argue that, in order to do this, indigenous communities of Mexican and Guatemalan origin
will need to articulate their own research agenda and priorities, effectively critique and correct the
inaccuracies of conventional wisdom and standard datasets, initiate their own research initiatives
using appropriate and affordable research methodologies, and deploy the findings from more
authentic research programs in a strategic fashion in order to leverage change in inequitable
policies and ineffective social programs.2 We suggest initiatives through which California

  We are no more than interested observers of indigenous communities modes of governance, but we have
no doubt that the community processes of dialogue and decision-making under the system of ―Usos y
Costumbres‖ more closely approximate traditional ideals of democracy than the processes of contemporary
technocracy; for example, neither city managers nor school superintendents interviewed by Kissam in his
research on the Central Valley Partnership were very eager to hear from their constituents.
  Indigenous immigrant communities in the rural California communities we know best are primarily of
Mexican origin but these issues are important also for Guatemalan indigenous communities in urban
California and in other farmworker areas of the United States such as Florida.

indigenous immigrant communities can take the first steps toward research autonomy, an
important strand of self-liberation in a society and economy where information rules.

The Causes and Consequences of Distorted Data

We present here our analyses of the deficiencies of decennial census data and similar datasets as
an empirical foundation for systematically addressing the social program needs of communities of
indigenous-origin immigrants in California and the nation.

Why Focus on Census Data

Data from decennial censuses are, in many circles and in many official contexts, considered the
gold standard for applied social science research, the foundation for community needs assessment
and framework for social planning. We focus on the problematic ―case‖ of Census 2000 data in
five rural California communities where we have a fairly good understanding of the causes and
consequences of distorted census data on indigenous Mexican immigrants and their families.
There are several reasons why we focus on this case although there are many examples of flawed
, nevertheless widely-used data used as the basis for decision-making in the California context.3

The first reason we focus on census data is because they are the basis for developing a wide range
of other data series at the state, county, and local level. Census data have primary use in the
political representation and federal program planning and funding decisions, but the data are
used even more widely by the state, counties, and local communities for a variety of decisions.
These data, for example, provide the baseline for California Department of Finance population
projections throughout the decade, development of local housing elements, and planning related
to community development.

The second reason is that the census methodology and data are so respected yet so flawed.
Census methodology provides the paradigm for survey research methodology (because it is so
reliable in homogeneous communities and for traditional data needs) , yet it epitomizes the
ethnocentric assumptions and methodological limitations of standard applied research
techniques for generating data on non-standard populations such as indigenous Mexicanos living
in California. The goal for Census 2000 was, in the words of the National Academy of Science,
to be a ―mirror which reflects America‖, but the mirror is a distorted one. A critical problem is
that the mirror is widely-used as the basis for centralized, ―cookie cutter‖ approaches to social
planning. ―Minor‖ flaws in relation to the goal of generating an overall snapshot of America, i.e.,
the invisibility of a few hundred thousand farmworkers or a few hundred thousand indigenous
immigrants, are considered to be inconsequential at the national level, but they become huge and
serious barriers to forging reasonable responses to widespread community problems at the local

  Examples of fundamentally flawed data widely used but disastrously ethnocentric include standard data
on unemployment (which has been known for decades to have serious problems relating to definitions of
―labor force‖ and methods for measuring numbers of persons seeking employment) and STAR data on
educational achievement in California’s schools where there is a very high proportion of limited-English
students. We have for more than 5 years sought to convince the Census Bureau to address very basic
problems stemming from ethnocentricity—the assumption that employed persons have a single, stable
―job‖ (seldom true for immigrants or for that matter, researchers) and the assumption that few households
have more than six people living together.

The third reason is that census data are the ―800-lb gorilla‖ of funding allocation, driving
significant funding decisions by public agencies.4 A good example is the State CDBG program
where funding allocation includes a term about the relative number of persons in poverty and
another term about the number of overcrowded housing units. Other key data elements in
important social programs include number of children in poverty, persons with less than a high
school education and per capita income. Half (13 of 25) of the major programs analyzed by
GAO, including a variety of compensatory education, employment training, health care and social
service programs, rely on such population profile data in addition to population counts for
allocating funding.

Figure 1 lists several programs of particular importance to California communities of Latino
immigrants of indigenous origin which are affected due to sample bias resulting from differential
census undercount skewing California population profiles. This table also lists the specific
funding formula elements where undercount of immigrant farmworkers, and, particularly, of
indigenous immigrants would result in significant losses of funding for programs that provide
critical services for their well being. A full analysis of the social impact of greater or lesser
financing for each of these programs is beyond the scope of this paper. It is clear, however, that
many of these programs play a critical role in rural California immigrants’ ability to live in
anything better than marginal, sub-standard circumstances. It also should be understood that
census data drives state, regional, and local planning and resource allocation; accurate census
data is as important for these purposes as it is for allocation of federal funds.5
                                                Figure 1
          Federal Programs of Importance for Indigenous-Origin Latino Immigrants
                  Impacted by Differential Undercount in the Decennial Census

        Program Name                     National             Key Element(s) in Funding Formula
                                      Funding Level-
Title I Grants to Local              $7.5 billion           % of children 5-17 in poverty
Educational Agencies
MedicAid                             $104 billion           Per capita personal income

WIC                                  $3.0 billion           # children in families <185% of poverty
Maternal Child Health Services       $0.6 billion           # of children under 18 in poverty
Block Grant
Child Care and Development           $1.0 billion           Per capita income
Block Grant                                                 Population <5 yrs of age.
Community Development                $2.7 billion           Poverty

  Census data have a two-stage process impact on funding: first they determine the amount of funds
flowing from federal programs to state, county, and local community jurisdictions; and second they
determine the activities and projects funded by agencies with the discretion to grant the funds to
community-based organizations or various projects.
   A good example of the ways in which the census data affect state funding is in the county-administered
programs for families with pre-school age children that receive annually about $500 million in state
funding ($492 million in 2001-2002). County Commissions have funding discretion, but their decisions
about which groups to fund are in principle determined by data which provide indicators of the need to
serve various identified sub-populations, e.g. Mixtec-origin families in Madera County.

Block Grant-Entitlement                                # of overcrowded housing units
Community Development              $1.2 billion        Poverty
Block Grant-State Program                              # of overcrowded housing units
Home Investment Partnerships       $1.4 billion        # rental units occupied by poor
Program                                                # of problem rental housing units
                                                       # of families in poverty
                                                       Per capita income of 3-person family
JTPA Title II-A and B/WIA          $2.0 billion        # economically disadvantaged individuals
                                                       ages 22-72 w/ family income below
WIA/ EL/Civics                     $.07 billion        10-year growth in LPR immigrants
                                                       3-year growth in LPR immigrants
WIA/Adult Literacy Act             $.46 billion        # of persons 16+ without HS diploma

Vocational Education               $1.0 billion        Population between 15 and 19

* General Accounting Office, ―Formula Grants: Effects of Adjusted Population Counts on
Federal Funding to States‖, February, 1999. We have not included in Table 1 programs where the
only census-driven funding formula element is the state’s or local jurisdiction’s share of the
national population.

It is particularly important, in assessing the potential impact of census undercount on rural
communities with high proportions of immigrants and on the social integration of immigrants in
general, to recognize the importance of federal funding formulas which allocate funding to
impacted counties or local communities. This is because census undercount ―smooths out‖ in
larger units. The small communities where immigrants are concentrated have the highest relative
proportion of immigrants, thus they also are likely to have the deepest ―pockets‖ of what we call
―mega-undercount‖, that is undercount above 10% of total population.

The systematic biases which affect the reliability of census data for numbers of children and
characteristics of children in communities with high concentrations of indigenous-origin
immigrants is one of our primary concerns, in part because programs for children are so important
to immigrant families and to society as a whole, but, also, because the census data on children are
particularly vulnerable to systematic undercount (an issue we discuss in more detail below).

Title I funding is a particularly important concern, both because of the importance of
compensatory education as a mechanism for supporting educational services to immigrant
children (both Migrant Education and ―regular‖ Title I services) and because of the specific
provisions of the funding formula. Title I funding includes Basic Grants, plus a sub-program of
Concentration Grants and Targeted Grants which direct funding to the most seriously
disadvantaged counties in the nation. Virtually all California children of indigenous origin
should be served by programs which are supported entirely or primarily with Title I funding.

Funding for child and maternal health is another important issue There continue to be serious
constraints on indigenous immigrants’ access to Medicaid (medi-cal) due to eligibility
restrictions to citizens and legal permanent residents, nevertheless, issues related to federal
funding for this program and other maternal-child health services remain important because they
support, at least, provision of services for all California-born children.

Several programs with census-driven funding formulas (such as CDBG) directly fund
community infrastructure, in addition to programs which directly affect the well-being of affected
individuals and families in a particular population. The negative impact of systematic bias in
decennial census data in communities with high concentrations of indigenous-origin immigrants
is likely to be particularly severe because crowded sub-standard housing is so common in these
areas and because it plays a major role in census undercount.

We note that approximately $700 million in annual funding earmarked for programs serving
migrant and seasonal farmworkers (Migrant Education, Migrant Health, Workforce Investment
Act-funded employment training) canbe affected by census profile data, even though funding
formulas do not rely primarily on census data.

Causes of Differential Census Undercount—Theoretical Framework

It was not until the 1970’s that serious attention came to be given to understanding the causes of
census undercount. It is only in the past 15 years that adequate attention has been paid to what
might be termed the structural causes of undercount, i.e., ways in which the decennial census
research methodology interacts with contemporary social systems to give rise to undercount. Our
contribution to the field of applied policy research on census undercount has been to recognize
and demonstrate that in the context of extremely divergent social and economic conditions, the
interaction among these multiple factors would give rise to what we call ―mega-undercount‖ or
the under-representation of immigrant, particularly farmworker, communities which fall not into
the ―debatable‖ range of 3-7% undercount, rather into a range of 20-50%. This range of
undercount leads not only to problems in quantifying numbers of persons in a community –in the
areas where we have focused, migrant and seasonal farmworkers—it also grossly distorts the
population and community profile (Kissam, Gabbard, and Martin 1993; Kissam and Jacobs 1996;
Kissam and Jacobs 2001). This structural analysis of undercount is extremely important for
California indigenous immigrants, and the communities in which they live, because their census
participation is negatively affect by a broad range of social factors, including:

   quality of US Postal Service address lists/Census Bureau address lists. This
   housing conditions—low-visibility, illegal, subdivided, ―unusual‖ housing
   housing conditions—crowded housing
   housing segregation—living in ―linguistically-isolated‖ blocks
   limited-English ability and limited literacy
   prevalence of recent immigration
   prevailing patterns of employment and working conditions
   quality of census operations—recruitment/hiring and training
   quality of census operations—non-response followup (NRFU)
   quality of census operations—procedures for enumerating migrant labor camps

Growing Understanding of the Social Causes of Census Undercount—Immigrant
Communities as Extreme Cases of Undercount

Our initial work, in the late 1980’s, on census undercount of migrant and seasonal farmworkers
recognized that the structural causes of undercount which had been widely acknowledged in
urban areas also are present in rural areas. This recognition was perhaps easier in California
where it is clear that ―farming‖ is less a matter of American folklore about rural white picket
fences, and more about the factories in the fields described by Carey McWilliams more than 50
years ago, and less a matter of finding remote residences and more a matter of dealing with

communities described by Michael Kearney and others as part of a global phenomenon of
―peripheralization of the core‖.

We adopted a newly-available theoretical model of census undercount stemming from the work
of David Fein (Fein, 1989; Fein and West, 1990) and a team of Census Bureau researchers. It
was, based on data collected in the ―Causes of Undercount‖ Survey as part of the 1986 Los
Angeles TARO (Test of Adjustment-Related Operations), a large-scale test census which took
place in inner-city areas of Los Angeles. Fein’s research showed very deep pockets of undercount
of Hispanics in East Los Angeles, in the order of 20%, or four to five times greater than reported
at the national level. We recognized that the structural (social system) analysis of the causes of
undercount demonstrated by Fein had particular relevance for understanding the undercount of
immigrants in general and, specifically, the undercount of farmworkers. Kissam, Martin, and
Gabbard used Fein’s logistic regression model to generate the ―synthetic estimate‖ of
approximately 50% probability of census omission for the typical California farmworker
household. We then compared this synthetic estimate to actual farm labor data from the 1990
Census (PUMS) and labor market data (UI records) to demonstrate that, when adjusted for
MSFW migrants not in California on Census Day, the estimates and extant data sources were in
close agreement showing that less than half of California’s farmworkers were identified in the
decennial census.

Components of the Census Undercount of Immigrants in Rural Communities

We discuss below the distinct factors implicated in census undercount in both inner-city
neighborhoods with high concentrations of immigrants and the rural communities where
indigenous immigrants to California are concentrated. It provides a conceptual "map"
demonstrating the "cascade" of events which generate the ―mega-undercount‖.

Component 1 of rural California undercount of indigenous immigrants stems from reliance on a
March "snapshot" as the basis for generating a profile communities when a "movie" showing
migration patterns throughout the year is needed to provide a full portrait of the population.
Transnational migration patterns indicate that a significant number of back-and-forth migrants are
in Mexico at the time of the census snapshot. Ironically, among the Mixteco farmworker
populations we understand best, it is likely that the sub-group of transnational migrants most
likely to disappear from the census snapshot are earlier cohorts of immigrants who adjusted their
status under IRCA who are able to travel back-and-forth between California and their home
villages.6 We have limited data on this; however, our research in Parlier showed that 26% of the
farmworkers in this community were out of the country when the 1990 decennial census took
place (Griffith, Kissam, Garcia, et al 1995).

Component 2 of the undercount stems from Census Bureau problems in assuring that census
forms are received by persons living in ―low-visibility‖ crowded housing in rural California
communities and that delivered forms are returned. Specific reasons identified for missing entire
housing units include factors related to census implementation (address list omission, erroneous
deletion of real addresses) and factors related to the social context of hard-to-count populations (

 This is a good example of an area of research where indigenous communities are better prepared than
outsiders to investigate the proportions of individuals and families who are not in the U.S. during the
census time frame. The National Agricultural Worker Survey sample provides a good basis for estimates
of proportions of farmworkers who are not in the U.S. in March, but does not have a large enough sub-
sample of indigenous farmworkers to make such analyses possible.

physical characteristics which affect unit visibility—hidden housing, illegally sub-divided

Our efforts to improve Census 2000 and our census evaluation research taught us that these
problems vary greatly from community to community. Some communities with substantial
concentrations of Mixtecos, such as Arvin, have very little ―low visibility‖ housing, while
somewhat similar communities such as Parlier, with a significant population of Mixteco and
Triqui residents, has a good deal of ―low visibility‖ housing (Kissam and Jacobs 2001; Sherman
1995). An extreme case where low-visibility housing results in census undercount is that of a
North San Diego County encampment well-known to Oaxacan migrants who have come to
California over the past 15 years; although CRLA had called it to the attention of local census
office staff, it never was enumerated. The Component 2 ―social cause‖ of census undercount
distorts the sample of persons in the census so that itunder-represents the most economically and
socially marginal populations such as recently-arrived transnational migrants from minority
communities in indigenous areas of Mexico and Guatemala, e.g. Chatinos, Amuzgueños (one of
the groups in the hidden camp CRLA enumerated).

Traditional census research tends to attribute causes of census undercount to census
respondents, however, our research has emphasized systemic causes. One area where
respondents do contribute to census enumeration problems is where a lack of English speaking
ability and little schooling makes it difficult to complete the English-language highly-formatted
census form.7 This problem surely contributed to some households’ failure to complete and
return census forms, but this should have triggered non-response followup by a Census Bureau
enumerator.8 We were disappointed to discover that in Census 2000 in rural California, some
households apparently in the census sampling frame ( houses to which a census form had been
mailed), yet no one had returned the form were not enumerated because of a failure in the local
census office non-response followup (NRFU) procedures.9

Component 3 of the undercount stems from difficulties in listing everyone who lives in the
housing unit on the census form. Previous census researchers had suggested that the main
problem was that census respondents ―forgot‖ persons such as newly-born infants. Our research
in Census 2000 showed that the main problem was that the census form which secured
information up to 6 persons in the housing unit did not permit otherwise willing census
respondents to list all of their family members or members of another family who shared a house,
apartment, or trailer with them.

  Forms formatting conventions present special difficulties for low-literate respondents. Low-literate
readers have difficulty in understanding procedural directions for forms completions even when the terms
are the correct ones in Spanish, e.g. ―renglon‖.
  A good deal of Census Bureau research has shown a correlation between high rates of forms not being
returned by mail and census undercount, generally believed to stem from not fully understood modes of
system breakdown in local office operations.
  This was very surprising because the mailback rate was relatively good even in rural communities; we
believe this stems in part from the fact that so many of the ―hard to count‖ households were omitted from
the census address lists. In cases where there was a contact by a Census Bureau enumerator, most, but not
all, respondents we talked to were pleased with the help they had received in completing the form.
Problems with enumerators simply stemmed from cases where enumerators did not speak Spanish.

Component 4 of the undercount stems from census form design in which bureaucratically-defined
concepts of ―race‖, ―Hispanic origin‖, and ―ethnicity‖ are so muddled and gave rise to such
extraordinarily strange survey instrument design that tabulation for rural communities might be
worthless.10 We are not yet certain that we know the full dynamics of census efforts to represent
ethnic minorities among rural Latinos, but it appears that in most of the rural communities where
indigenous immigrants are settled, the census data make it impossible to recognize 4 out of 5
indigenous households (see discussion on p. ____.11

Census Undercount in Rural California Communities with High Concentrations of Indigenous

We focus next on findings in five communities examined as part of CRLA’s research on Census
2000 in rural California—Oxnard, Arvin, Madera, Parlier and Santa Maria—all of which are
known to have relatively high concentrations of Oaxacan indigenous immigrants, most of whom
are farmworkers. The 1991 California Institute of Rural Studies survey of Oaxacan village
networks in California, enumerated more Oaxacan immigrants in Madera than in any other
California community (2,444—more than one-third of all Oaxaquenos enumerated in rural
California).12 Santa Maria , at that point, was recorded as having 365 Oaxaquenos, Arvin, 240,
and Parlier, 33. Surprisingly, only 11 Oaxaquenos were found in Oxnard.13

The variations in undercount by community suggest that in some cases efforts by census
personnel and local partner organizations served to mitigate problems of undercount. Our finding
that undercount varies substantially from community to community underscores how problematic
census quality problems can be for data users. This problem stems from the recognition that a
single adjustment factor cannot be easily applied to all these communities, although they might be
considered to fall into a single stratum, i.e. ―rural farmworker communities‖.14 Table 1 reports
our estimate of undercount in the five case study communities with a high concentration of
indigenous immigrants.

  Hispanic respondents’ confusion about the census question for ―race‖ resulted in absolutely useless
breakdowns of census respondents listing their ―race‖ as ―White‖ or as ―other‖. Arvin, a community where
we know that almost all Hispanics are of Mexican origin, had 40% of Hispanic respondents list their race as
―White‖ while about 60% listed their race as ―other‖.
  We know that Arvin has a significant concentration of Mixtecos and a well-established network of
immigrants from San Juan Mixtepec, but census data suggest there are only 280 persons living in
households which speak Mixtec or another indigenous language—about 2.5% of the Hispanics ( we believe
the actual proportion is at least 10%).
 David Runsten and Michael Kearney, ―A Survey of Oaxacan Village Networks in California
Agriculture‖, California Institute of Rural Studies, 1994.
  Our interviews in 1999-2002 with Mixtecos and Triqui farmworkers working the ―strawberry‖ circuit
which includes Oxnard, Santa Maria, and the Willamette Valley suggest there were surely significant
numbers of Mixtecos in Oxnard by 1990.
  The strategy needed to yield ―best estimates‖ of the actual population and population characteristics of
persons in each community would be to generate synthetic estimates, using multivariate regression models
such as the one developed by David Fein on the basis of the Causes of Undercount Survey in the 1986
Census Test (Fein, 1989).

                                            Table 1
                              Extent and Type of Census Undercount
                                in CRLA Case Study Communities

Community             Study    Overall                     % not enumerated          % not enumerated
                      Block Undercount in                    -- Total HH               --Partial HH
                       Pop.  Study Block                      Omission                   Omission
                                 (N)                              (N)                       (N)
Oxnard             161          34.1%                            9.3%                     24.8%
                                 (55)                             (15)                      (40)
Santa Maria        128          10.9%                            8.6%                      2.3%
                                 (14)                             (11)                       (3)
Arvin              109          13.8%                            6.4%                      7.4%
                                 (15)                              (7)                       (8)
Parlier16          170          37.6%                           27.6%                     10.0%
                                 (64)                             (47)                      (17)
Madera17           157          19.8%                           15.3%                      4.5%
                                 (31)                             (24)                       (7)
Source: CRLA Census 2000 Community Studies

The extent of undercount varies substantially among the communities, as do the components of
undercount, i.e. the relative contribution of total and partial household omission to the overall
undercount. Varying patterns of undercount are an important consideration in assessing the
reliability of census data at the local community level because each gives rise to a different bias.
Low census quality in the Oxnard sample block, for example, stems entirely from the failure of
the census to secure information on persons # 7-15 in large, complex, crowded households. The
Parlier undercount, in contrast, stems primarily from omission of low-visibility housing because
there are more ―back houses‖ rented by long-term settled farmworkers to transnational migrants
and socio-economically peripheral residents, while in Oxnard, these peripheral sub-populations
share a crowded single-family dwelling with the ―primary‖ family unit.18 Arguably, the
variations in housing accommodations affecting the nature of census undercount stem in large
measure from differences in the mix of migration networks in each community as well as the
history of each network’s migration and its success in coming to control housing resources.

 Excludes 5 unresolved households, two due to refusals, three due to non-availability of residents after
multiple visits.
   Excludes four unresolved households due to refusals. The analysis also includes four ―ambiguous‖
households in which there were inconsistencies but where we considered it possible to impute census
status. We tabulated two of these households as having been enumerated and two as not having been
  Assumes that seven in-movers who arrived after the questionnaire had been returned had actually arrived
after April 1. Thus the analysis excludes these in-movers from the sample as they were not and should not
have been enumerated.
 In Parlier, the majority of the landlords are mestizo former farmworker families from Guanajuato or
Michoacan, while the majority of the transnational migrants are Oaxaquenos.

Partial Household Omission

The CRLA community case studies show 2 causes of partial household omission, neither of them
related to household residents’ ―forgetting‖ to list persons living there as some census researchers
have suggested.

The first, and most serious source of partial household omission is that the census form has
space to provide population profile information for only 6 persons living in the housing unit. This
ensured that in any household with more than 6 persons, in the study communities and other
communities with crowded housing, some persons were at risk of not being enumerated. We
specifically asked all respondents in the study blocks whether they had returned a census form
and, if they lived in a household of more than 6 persons, whether someone from the Census
Bureau had come to ask for information on the persons for whom no personal characteristics had
been reported. We heard of no case in which there was follow-up to determine the characteristics
of the missing persons.

The second type of household omission is related not only to census form design but to the
prevalence of ―complex‖ households occupied by several distantly-related families, unrelated
families, or ―unaccompanied male transnational migrants‖. The census form was filled out to
reflect the composition of the primary family unit, e.g., a couple and their 2 children, and the
secondary family unit was totally omitted, in some of these households, even if the primary
family unit numbered fewer than 6 persons. We are less certain about what happened in
households consisting entirely of unaccompanied male migrants, but it is possible that in these
households (which also have an internal social structure based on smaller social/economic units
of men clustered on the basis of extended family or village network ties) some of the residents
were omitted.19

Total Household Omission

 There are also 2 types of total household omission in the community case studies. The first and
more systematic pattern of total household omission occurred where the housing unit appeared
not to have been in the Master Address File in the first place. These dwellings included backyard
trailers, shacks used as ―back houses‖, housing quarters which were structurally part of a single-
family home, with their own entrance (―cuartitos‖), camper shells, and sheds. These omissions
systematically skew the community profile to make the most economically and socially marginal
portions of the overall community ―invisible‖ in census data. The second type of total household
omission arises in the situation where there is mail delivery and a census form is delivered to a
dwelling, but is not returned and there is no non-response follow-up. We know little about the
dynamics of this component of undercount except that it appears to be fairly random. Non-
response follow-up was much better in some communities (e.g., Arvin, Santa Maria) than in
others (e.g., Oxnard), but we know only that this type of census problem arises from some
breakdown within census operations.

Bias in Community Population Profile as A Result of Census Undercount

Our primary concerns about census data have centered on the problems of quality that arise
when differential undercount systematically skews the population profile of low-income

  We have tabulated these households as being partial household omissions only when there were more
than six persons residing in the dwelling.

communities. These concerns are real and practical. Flawed census-based population profiles,
therefore, create substantial problems for local jurisdictions seeking to secure equitable access to
the program funding targeted to provide services to the very sub-populations underrepresented in
the census-based population profile, as well as for those program managers who are seriously
attempting to adapt program designs to respond to community needs.

Figure 3 shows the types of sample bias associated with each of the types of census omission
observed in the CRLA case study communities. The most serious consequence that we have
observed is the serious under-representation of children, which results from householders in large
families systematically listing persons on the census form, beginning with the couple filling out
their form and then, usually, listing children in decreasing order of age.

                                       Figure 3
                        The Impacts of Differential Undercount
        on Indigenous Immigrant Community Profile and Social Program Funding

   Type of Omission       Types of Persons Typically      Impact on Social Program Funding
                                   Left Out
Total HH Omission-                                        Moderate downward bias in estimates
dwelling not in              transnational migrants-        of prevalence of households in
Master Address File           unaccompanied males                      poverty

                             transnational migrants       Strong downward bias in estimates of
                                 young families              extremely poor individuals and
                                                              families, e.g. <40% of poverty
                               very poor families                        guidelines

                             indigenous persons not       Strong downward bias in estimates of
                           linked to locally dominant      educationally disadvantaged, e.g. %
                               network (e.g. Maya)           of population w/ < High School

Total HH Omission-        settled immigrants with very     Slight downward bias in estimated
failure of NRFU             limited literacy or English     prevalence of crowded housing
                           ability and their dependents
                                                          Slight downward bias in estimates of
                                                                 educational attainment

Partial HH Omission-         underrepresentation of       Strong downward bias in estimates of
Large Nuclear Family       pre-school and elementary        # of low-income children, children
                              school-age children          targeted in compensatory education,

                                                          Moderate downward bias in estimates
                                                            of prevalence of crowded housing
Partial HH Omission-         under-representation of       substantial under-representation of
Complex Household              recent immigrants,         pre-school and elementary school-age
                              especially arrimados         children, especially limited English

                                                           downward bias in community-level
                                                           estimates of prevalence of crowded

Differential Undercount within Rural California Communities

All of the sample blocks in the CRLA case study communities of census undercount were chosen
because community researchers believed they were ―at risk‖ for possible census undercount.
There appears to be an even higher risk of undercount for farmworkers than for other types of
families in these study blocks in what are predominantly immigrant neighborhoods. Not all
farmworkers are of indigenous origin, but farmworker undercount remains a serious problem for
indigenous Mexican immigrants in rural California because they are concentrated in farmwork.

In the ―Causes of Undercount‖ Survey (CUS) in Los Angeles, neighborhoods with more recent
immigrants had the most serious undercount but, even among recent immigrants, undercount
probably varies greatly.20 Based on hundreds of discussions with recently-arrived teenage
migrants we know that different sending villages have more or less developed networks; and that
persons in established networks have better access to housing than others. This probably skews
the census profile of different indigenous networks, e.g. migrants from communities such as
Juxtlahuaca, San Juan Mixtepec, Tlapa, or Santa Maria Tindu with long migration histories being
better-enumerated, while migrants from more remote areas with more recent migration are less
well-enumerated, thereby skewing the cultural-linguistic profile of indigenous migrant
communities in California.

Table 2 reports the demographic composition of the community case study blocks and the extent
of farmworker undercount in each.

   It is assumed that the very high levels of census undercount in neighborhoods and communities with high
levels of recent immigration stem primarily from fear of government authorities. We believe that sub-
standard housing conditions, low literacy, and language problems have more to do with undercount in
these neighborhoods than respondent refusal. The Los Angeles CUS was not designed to determine which
immigrant networks were represented in the study area nor to explore the relationship between migration
network maturity and undercount.

                                            Table 2
                  Demographic Profile of Case Study Community Sample Blocks
                        and Extent of Farmworker Undercount in Each

Community                              Oxnard         Santa         Arvin        Parlier       Madera
% Hispanic                               94%           95%          98%           100%           91%

% LEP adults                             78%            43%         78%            65%           41%

Mean School (Adults 18+)                5.4 yrs.     8.1 yrs.      5.3 yrs.      8.2 yrs.       6.6 yrs.

Mean HH Size                              5.9           5.8          3.9            4.6           4.0

# Farmworker HH’s in Block        16         19                       17           (20)          (23)
and % of total                 (55%)       (73%)                    (61%)          49%           55%
# and % of Farmworker HH’s         6          5                        5            11             6
Totally or Partially Omitted   (37%)       (26%)                    (38%)         (55%)         (26%)
% of Problem HH’s in Block       6/9         5/6                      5/5         11/12           6/7
which are FW HH’s              (67%)       (83%)                   (100%)         (92%)         (86%)
Ratio of Problem FW HH’s to
Problem Non-FW HH’s             1.4:1       1.8:1                    ----         11.5:1         4.1:1
Source: CRLA Census 2000 Community Studies

Table 2 shows that even within communities at high risk of census undercount due to widespread
prevalence of low-visibility, crowded housing, recent immigration, limited-English speaking
ability, and low literacy, census undercount is concentrated primarily in the farmworker
population. It should be remembered that even in households where farmworker dependent
children are U.S. citizens and fluent in English, their parents’ status as farmworkers means that
these children will have a higher risk of undercount than other similar children, primarily because
they tend to live in more crowded, low-visibility housing.

The Case of Two Informal Labor Camps which are California Destinations for Indigenous

The two case study labor camps are illegal encampments, which are controlled by San Diego
County tomato producers known to rely primarily on transnational migrants to make up their
labor force.21 CRLA notified the Local Census Office about these camps, but they never were
properly enumerated with the procedures designed for ―special places-migrant camps‖; these
encampments could not be enumerated using standard procedures because they were hidden
camps without mail delivery. Failure to enumerate persons living in this sort of concealed
housing has very serious consequences for data users interested in the growing ethnic diversity of
California because the populations of the encampments are primarily recently-arrived indigenous

  CRLA consultant Anna Garcia first became aware of these encampments in the course of field research
she was conducting for the Center for Mexico-US Studies at the University of California, San Diego in the
mid-1980’s. Two of the CRLA community researchers had lived in these camps in the 1980’s when they
had first come as migrants from Oaxaca to California.

transnational migrants. Table 3 provides a summary profile of the ethnic/linguistic and
demographic composition of these two encampments.
                                          Table 3
                         Profile of the Non-Enumerated Residents
                     of Two Hidden Labor Camps in San Diego County

Characteristics                                        Camp 1         Camp 2          Overall
                                                       (N=69)         (N=39)          (N=108)
Primary Language
Spanish only                                              9%            18%             12%
Mixtec                                                   54%             --             35%
Triqui or Amuzgo                                          7%            82%             34%
Mixtec, Triqui, Amuzgo and Spanish                       29%            48%             19%
Limited-Spanish                                          72%            33%             58%
Educational Attainment
3 years of school or less                                70%            28%             55%
4-6 years of school                                      23%            67%             38%
6+                                                        7%             5%              7%
Source: CRLA Census 2000 Assessment

Census Long-Form Data on Mexican Immigrants of Indigenous Origin

The recently-released SF-3 tabulations of Census 2000 long-form data on the five CRLA case
study communities where we know there are high concentrations of Latinos of indigenous
(primarily Oaxacan) origin shows that the decennial census ‖mirror‖ provides only a blurred and
ambiguous outline of the true picture of these communities. This review shows that Component
4 of census undercount (Mis-Identification) provides an additional, and major, source of error in
profiles of indigenous Mexican populations in rural California Table 4 reports our tabulations of
relevant data for the case study communities from the CRLA study.
                                              Table 4
                      SF-3 Data from Census 2000 as a Basis for Profile of
                                 Mexican Indigenous Immigrants
Case Study         # Hispanics # Persons 5+ in         # Possible     # Recent        Recent
Community              5+             HH w/           Indigenous     Immigrants    Immigrants
                                    Amerindian       Ancestry and (1995-2000)        as % of
                                  Language and %      % of Total                    Hispanics
                                  of Hispanic 5+ *    Hispanic**
Arvin                10,046              281              1,717         1,508         13.2%
                                       (2.8%)           (13.3%)
Madera               25,667              471              6,838         3,469         11.8%
                                       (1.8%)           (15.8%)
Santa Maria          40,469              382             12,391         5,527         12.0%
                                       (0.9%)            (27%)
Oxnard               101,235             440             24,365        12,439         11.0%
                                       (0.4%)           (21.5%)
Parlier               9,725               93              1,722          933           8.6%
                                       (1.0%)           (15.9%)
* Tabulated based on “other” languages as tabulated in SF-3, i.e. non-Indo European, non-
Asian Pacific Islander, Non English

** Tabulated based on ancestry being listed as unclassified or not reported

The data problems which compromise the reliability of the census data presented in Table 4
provide a striking example of the problems inherent in ethnocentric research within the context of
California’s ethnically and linguistically diverse rural communities.

The tabulations of languages spoken at home list ―other‖ as the closest classification for any of
the Amerindian languages of Mexico or Guatemala. ―Other‖ language is defined as non-
European, non-Asian or Pacific Islander, and non-English so we can assume with some certainty
that most of the persons in households where some ―other‖ language is spoken are persons who
speak Mixteco or some other Oaxacan language.22 Language patterns in Oaxacan indigenous
households where the language spoken at home usually include both Spanish and a native
indigenous language such as Mixteco Alto or Mixteco Bajo, therefore we cannot consider
―language spoken at home‖ as a reliable indicator of the ethnicity of the household since it is very
likely that many of these households were classified as Spanish-speaking households.

The Census 2000 data on ―ancestry‖ (the Census Bureau’s closest approximation of ethnicity)
provide a similarly uncertain picture of indigenous Mexicanos in our rural case study
communities because no classification exists for indigenous Mexican ethnic groups and because
―unclassified and not reported are tabulated together‖ in reports on specific ancestry.
―Ancestries‖ which are classified include such esoteric ―ancestries‖ as ―Carpatho Rusyn‖,
―Celtic‖, ―Bermudan‖, and ―Pennsylvania German‖, but Mexican and Guatemalan ancestries are
not tabulated, neither at the level of linguistic-cultural identification (e.g. Triqui, Tzeltzal) nor at
the regional/pan-linguistic level (e.g. Oaxaqueño, Maya). Therefore, based on census data, we
have no certain idea how many Mixtecos, Triqui, or other Mexican indigenous immigrants there
might be in these communities.

In the case of Arvin, we know that (according to census data) the lower bound for numbers of
Mixtecos is 281, i.e. the count of persons 5+ years of age speaking a non-Asian, non-European,
non-English language and that the upper-bound is 1,717.23 Given what we know about the
proportions of Mixtecos in the farm labor force, the ratio of farmworker dependents to
farmworkers, and the local economy of Arvin it is very unlikely that there could be less than
1,500 persons of Mixtec ethnicity in Arvin.

The CIRS Mixteco survey identified 2,444 Oaxacans in Madera in 1991 and the Census 2000
data suggest a lower bound of 471 and an upper bound of 6,838, thus we consider the ambiguity
introduced by the Census Bureau’s ethnographic incompetence extremely unfortunate. Our
knowledge about international migration networks, increasing representation of farmworkers
from Oaxacan networks in the California farm labor market, and Madera’s key role as an
upstream migration node in Pacific Seaboard networks means that we consider it extremely
unlikely that there were less than about 4,000 Oaxacan-origin immigrants in Madera in 2000.

  Other possible speakers of Amerindian languages might be U.S. or Canadian American Indians but the
data on race and country of birth suggest that most of these ―other‖ language households are indeed
Mexicans of indigenous origin.
  While the California Department of Education language census generally provides sound, up-to-date
information often useful in profiling populations, there appear to be problems in identifying households
such as Mixtec households ( only 8 Mixteco-speaking children are identified in the latest data on Arvin
Elementary School District from the California Department of Education and only 62 speakers of Mixteco
are identified in Madera Unified School District).

Recent survey work by Kissam conducted in the Madera area suggests that at least 31% of low-
income Mexicanos in the Madera area are of Oaxacan indigenous origin.24 This would yield an
estimate of approximately 3,600 indigenous-origin Mexicanos in Madera. However, using
birthplace as an indicator of ethnicity, it would seem that even a higher proportion of Madera
Spanish-speaking households are of indigenous origin.25

Overall Assessment of Census 2000 Data On Mexican Immigrants of Indigenous Origin
living in Rural California Communities

Our research on Census 2000 operations in 5 communities in rural California with higher-than-
average concentrations of Mexicanos of indigenous origin, primarily Oaxaquenos, suggests that a
cascade of failures in research methodology (Components 1-3 of census undercount and reporting
(Component 4) make the available data on indigenous Mexicanos in these communities highly
unreliable. Overall undercount in the neighborhoods where most indigenous immigrant
individuals and families live ranged from 11% to 38%. The flawed methodology for identifying
indigenous-origin Mexican immigrants as such further compromises reliability of the official
data. Our review of Census 2000 long-form data in communities which we know to have
concentrations of indigenous-origin suggest that the most conservative estimates of numbers of
immigrants of indigenous origin if we were to trust SF-3 data fall midway between the flawed
possible indicators of indigenous Mexican ethnicity (i.e. household language classified as non-
European/non-Asian/non-English and ancestry not reported or classified). While Census 2000
data indicate that indigenous-origin persons constitute somewhere 3-13% of Arvin’s population
and 2-16% of Madera’s population, it is likely the most accurate estimates are that about 8% of
Arvin residents and 10% of Madera residents are of indigenous origin.

Total and partial household omission skew the resulting profile of the indigenous population who
are enumerated in different ways, therefore the divergent modes in which census methodology
fails in different rural communities makes it very difficult to generate reliable ―synthetic
estimates‖ to correct sample bias in the census profile of indigenous Mexicanos. It is possible to
make qualitative statements about the types of sample bias which result from ethnocentric census
research methodology (e.g. under-representation of younger, more recently-immigrated
transnational migrants, under-representation of the youngest children in the large, complex
households in which the poorest indigenous immigrants live). Review of the SF-3 data reveals
that what should have been easily-avoidable definitional problems actually present the greatest
threats to the integrity of census data on indigenous Mexican populations in rural California.

What we know about the case of Madera, suggests that the census data are so flawed that
improvement is clearly possible and necessary. Even imperfect, but affordable research
methodologies can be used to generate population profiles which, despite their technical
limitations, compete favorably with the grossly distorted profiles of immigrant populations of

  This tabulation is based on information about indigenous languages spoken at home. Listed languages
include Mixteco, Zapoteco, and Triqui. Data were collected as part of The California Endowment/Tri-
Valley Tobacco Survey conducted for Radio Bilingue. The analysis presented here is based on a sub-
sample of 85 survey respondents interviewed at the Madera remate in June-July, 2002.
  Listed birthplaces elicited in the survey which we know to be Oaxacan indigenous villages of origin
included San Juan Copala, Huajuapan de Leon, Juxtlahuaca, Nochixtlan, Magdalena Loxxicha, Miahuatlan,
Nochixtlan, Ometepec, Putla de Guerrero, San Aguistin Atenango, San MiguelTlacotepec, San Sebastian
del Monte, San Sebastian Ixtapa, Santa Catarina Yutandu, Santa Maria Tindu, Santa Maria Tlacotepec,
Santa Roxa Caxtlahuaca, Santiago Naranjas, Tlaxiaco, Tequesquitlan, and Tlaxiaco.

indigenous origin which emerge from the technically sophisticated research but theoretically
flawed research which gives rise to decennial census data. This, in turn, suggests that indigenous
community groups might be able to initiate their own research programs to generate a faithful
image of their own communities, an authentic mirror of their own lives. The next section of this
paper describes options and promising strategies to make this possibility a reality.

Repairing the Flawed Mirror of Census 2000-Next Steps

What are the next steps to improve the distorted current profile of indigenous communities of
Mexican and Guatemalan immigrants in California ? We believe the most challenging, yet most
important next step will be indigenous communities identifying their own research priorities and
defining their own research agenda. A key concern is that such an agenda should be, on the one
hand, cognizant of relevant ―mainstream‖ research agendas but, on the other hand, should not be
unduly influenced by externally-identified priorities. The resulting research agenda should be
―authentic‖, i.e., it should truly reflect concerns within the indigenous immigrant community.
An agenda of that nature should be attractive to the more sophisticated philanthropic institutions
and public sector agencies.26

Processes for Developing an Authentic Research Agenda

Applied research is too important to be simply left to ―experts‖ or to funders, which is not to say
that experts and potential funders should be ignored. They control key resources which are
absolutely necessary for the success of such a research agenda and they are legitimate
stakeholders in the outcomes. It would be wise to articulate and implement a structured and
broad-based process of community consultation to determine the agenda , its priorities and it’s
sequence .

We believe that the basis for a successful process of community consultation is to recognize that
―ordinary‖ people, i.e. non-experts, are very good at qualitative assessment of what’s going on in
their lives and in the social dynamics of their communities, although they would be hard-pressed
to provide quantitative estimates of even the most widely-observed ―issues‖, e.g. the rate of
arrival of new migrants coming to California, prevalence of substandard housing, extent of
underemployment in a given community, or problems faced in accessing competent health care.
There are real problems which must be confronted in structuring community consultation
processes so that the guidance provided by ―the community‖ is not simply generic ratification of
fairly evident priorities and the full, diverse, range of community interests and opinion can be

A Framework for Articulating A Practical, Community-Oriented Research Agenda

There are many practical questions about how best to implement a process of community
consultation to provide the foundation for indigenous organizations to further refine and ―market‖
such an agenda. We do not address these here, however, we recommend exploring five
conceptually-distinct research agendas They are the following:

  One policy perspective is that part of the value of investments in community-based organizations stems
not simply from the activities which such funding permits but from the ―value added‖ by such
organizations’ knowledge of the environment in which they are working and the difficult-to-measure
contribution of their commitment to their organizational mission.

1. Basic Population Research on Indigenous Immigrants.

This research is critically needed both to correct the distorted picture of California rural and urban
indigenous immigrant communities and to fill in key gaps in the picture. Based on our census
research, we can strongly recommend priorities such as the need to improve estimates of the
number of children in families of indigenous-origin and to determine key issues relevant to social
policy, e.g., the proportions of undocumented, legal permanent resident, and citizen children in
these families. The ultimate decisions on research priorities should be made by the indigenous

There is a broad spectrum of competing ―basic research‖ priorities, due in part to the gaping holes
in census research. It would be very useful to understand the ethnic diversity of indigenous
immigrants to California. There is consensus that the diversity of indigenous immigrants coming
to California is increasing, yet there is little sense about which village networks and languages
are represented These questions boil down to asking whether a community-defined basic
research agenda should include the full range of ―standard‖ data items collected in a survey such
as the census irrespective of immediate relevance to current issues (e.g. grandparents who care for
children).27 Or should indigenous community organizations’ research agenda focus on addressing
basic questions which are not asked by the census but are important in community life (e.g.
information on the multiple types of economic activities households carry out to survive in low-
income communities, information frequency and duration of return to Mexico—either to carry
out cargos or participate in the tequio or to deal with family problems)?. We think the indigenous
―basic research‖ program should be tailored to drop particularly American preoccupations (e.g.
specific sources of public assistance income, type of fuel used for cooking) and add new data
items and analyses (e.g. information on limited-Spanish/limited-English persons in indigenous-
origin immigrant households).

2. Targeted Research on Issues/Problems Faced by Indigenous Immigrants.

Here is where community research will be most critical because there are legitimate competing
agendas. For example, standard data sources do a terrible job of distinguishing the problems
faced by recently-arrived, predominantly male, predominantly young transnational migrants from
settled immigrants. Should understanding their needs be a priority? Are these young men really
willing to forego (as agricultural employers argue) decent housing to maximize their earnings and
remittances? How often do they migrate and what are the main problems they face in the course
of migrating? How often do they go without any access to health care and why? Is exposure to
HIV a major problem (as some researchers fear) or not? Might it be wiser to focus on the
problems faced by settled immigrants of indigenous origin who are in the process of settling in
California? Do indigenous-origin children and youth face ethnic discrimination in their schooling
and, if so, to what extent? To what extent is the health care of indigenous-origin women from
communities where Spanish is not spoken compromised by language barriers and, if so, what are
the most prevalent problems?

The answer is that multi-faceted research is needed and the process of collectively identifying,
articulating, and prioritizing the research questions is a potentially rewarding one. One of the

  The criterion for including or dropping questions from Census 2000 was articulated narrowly in terms of
legislatively-mandated questions and comparability with earlier data series rather than priorities based on
the utility of the resulting data for sound program planning. Thus, the questions on grandparents caring for
children (legislatively-mandated) and types of fuel used at home (comparability with previous data series).

promising possibilities of such a community-initiated program of targeted research is that it
provides a welcome antidote to the poorly-articulated ―cookie cutter‖ research agendas identified
in the research programs of mainstream institutions (often designed to be comparable with
questionable prior research). Perhaps more importantly, community research can provide the
practical basis for mounting social policy and program responses which are not narrowly-
confined within the guidelines of a single bureaucracy, which instead are designed as across-the-
board responses to problems which run across a broad spectrum of programmatically-defined
―populations‖ (e.g. parents with pre-school age children).28 It should be noted that the lines
between ―basic‖ and ―targeted‖ research are fuzzy ones and that this presents opportunities, for
example, to insist on special procedures to assure that indigenous communities are not left out of
the picture, or to insist on improved definitions of basic variables in the research (e.g. ethnicity)
so that indigenous communities do not blend into the background of research on ―Hispanics‖.

3. Opportunistic/Entrepreneurial Funder-Driven Community-Based Research

A fundamental reality of the world of applied policy research is that research funders define the
research agenda, often with a minimal level of consultation with those who best understand the
populations being studied or those who may ultimately use the research findings. We do not
recommend that community-based organizations adopt a ―purist‖ stance and forego the
opportunities of responding to funder invitations to explore issues which affect their
communities—even if this does, to some extent, shift their own priorities. We recommend that
community-based organizations engage in dialogue wherever the opportunity permits, nudge
funders’ research agendas and designs in ways which will be more productive from their
perspective. One of the interesting opportunities is that research driven by a fairly narrowly-
defined set of research interests can be broadened to address other critical issues.

The current boom in funders’ interest in program evaluation presents some opportunities for
community-based research on indigenous Mexican and Guatemalan immigrants in California—at
least to the extent that program evaluation can be shaped to incorporate sound research. Prior
emphases on process-oriented evaluation left little room for sound research, but new emphases
on evaluation and program outcomes makes sound research feasible— only if there is an
insistence on appropriate strategies to assure that indigenous immigrants are included in the
research population and adequately identified once they are included.

We are currently impressed by The California Endowment’s commitment to solid community
research as one component of its multi-year multi-million dollar Agricultural Worker Health
Program. Our perspective is that this multi-year applied research effort which is being designed
to track individual, family, neighborhood, community, and systemic outcomes presents many
opportunities to improve basic understanding of indigenous immigrants’ lives and most pressing
community needs, at the same time as it generates information on ―best practices‖. Indigenous
community organizations will need to persistently assert their own research interests, their
insights about the basis for sound program design, and oversight to assure that there will be
consistent attention to their experiences as a ―sub-population‖ among farmworkers.

  Research conducted by Kissam and his colleagues (Anna Garcia, Anna Rodriguez, Jo Ann Intili) for
Radio Bilingue makes it clear that in the minds of parents of pre-school age children in Fresno and Tulare
Counties the problems they faced related to family life in general not to the more narrowly defined ―target
population‖ for programs funded by Proposition 10, i.e. children 0-5 years of age.

4. Opportunistic/Entrepreneurial Issue-Driven Research

It will also be important for community-based organizations of indigenous immigrants and their
allies to shape their research agenda in part to respond to new policy opportunities, newly-passed
legislation, and newly-issued regulations. Here too, there are many legitimately competing
priorities. To what extent should research be designed to explore issues where a policy solution
or legislation may be urgently needed? To what extent should research seek to generate
information on ―best practices‖, i.e. on what works best for which group of potential
beneficiaries who currently might be poorly-served by existing program designs, or institutions?
Both potential uses of research findings are legitimate and might bring practical improvements in
community well-being. Our perspective is that an entrepreneurial approach which actively seeks
opportunities to advance certain areas or priority research questions as part of an overall long-
term agenda has great promise.

A good example of such opportunities is that the ―mainstream‖ policy and program planning
agenda shows an emerging interest in questions relating to the cultural and linguistic competency
of service providers. Recent regulations issued by the Department of Health and Human Services
regarding patient’s rights to health care services in their own language have given rise to a
burgeoning policy debate as to whether program designs mandated by the regulations really are
needed (e.g. a prominent physician, arguing tacitly that children’s translation for their parents is
adequate and that the costs of interpreters would lead to unaffordable increases in health care
delivery). The California Endowment’s overall strategy for improving health care in California
rests on a variety of initiatives to improve the cultural competency of health care providers. This,
of course, provides an opportunity and, arguably, an obligation for community-based
organizations serving indigenous immigrants to seriously assess the cultural and linguistic
barriers to adequate health care and to explore issues relating to possible solutions, e.g. what
would be required to recruit and train Mixtec and Triqui-speaking paraprofessionals to actually
function effectively as cultural interfaces between physicians and their patients?

Similarly, California’s passage of AB 540 which allows immigrant students to be considered
California residents for the purpose of determining tuition fees at state institutions of higher
education (UC, CSU, and CCC) presents an opportunity and obligation to assess how well the
current provisions work for youth from indigenous-origin families, whether they are graduating
from high school, have equitable access to courses which satisfy the A-G requirements to enroll
in the University of California system, will be able to afford even the fees paid by in-state
residents, or whether there is a pressing argument for passage of the federal legislation (DREAM
Act) which would provide a special program of immigration status adjustment for in-school

5. Collaborative Research on Improvement of Census Bureau Methodology

Community-based organizations advocating on behalf of indigenous immigrants cannot afford to
abandon efforts to improve Census Bureau Research. They should join in the collective efforts of
a wide range of organizations representing minority populations adversely affected by census
undercount. CRLA, MALDEF, Asian American and Native American tribal groups have all been
actively involved in efforts to improve census methodology and operations. While progress has
been slow, the voices of indigenous immigrants should be heard and their insights about how to
improve census methodology will be crucial in the coming decade as a new national data effort,
The American Community Survey, takes shape and ongoing efforts are made to improve census-
style national surveys such as the Current Population Survey. Indigenous immigrant

organizations should present themselves as collaborators with the federal government and be
recognized as such.

Final Reflections on Articulating a Community-Driven Research Agenda

The five types of research sub-agenda we articulate above are not irrevocably separate categories
of research; they represent, instead, strands which interact and should be woven together in
ongoing strategic planning to develop the indigenous immigrant community’s research agenda.
Advocacy from indigenous community-based organizations can serve to shape the emerging
agenda of ―mainstream‖ research agendas, blending what we call ―basic‖ research with what we
call ―targeted‖ research as the basic research agenda comes to include an increasing proportion of
research questions which reflect the interests and priorities of indigenous communities
themselves. Recognition of ―issue-oriented‖ research opportunities can provide the basis for
shaping funders’ interests so that the gap between funder-driven research agendas and
community-oriented research agendas diminishes.

It is entirely appropriate for community-based organizations seeking to influence mainstream
research agendas to look to potential allies as they begin to implement initiatives to make
mainstream research more responsive to community priorities and needs. Our sense is that a
number of important philanthropic institutions in California clearly understand the importance of
immigrants for California and the diversity of Latino immigrants. We believe they are potential
allies in efforts to forge a well-articulated research agenda for indigenous communities and by
indigenous communities. We believe the ―targets‖ of such a campaign to re-define research
should be the public institutions at the local, county, state, and federal level which, in their
preoccupation with disparities in equity among different racial groups, consistently ignore the
disparities which exist within racial groups (or pseudo-racial categories such as ―Hispanic‖).
Even marginal improvements will be worthwhile because the current situation could not be much

Appropriate, Affordable Research Methodologies for Research

It is crucial, both from a research perspective and from the pragmatic perspective of community
activists, to understand that no currently affordable research methodologies are ―ideal‖ nor
provide the ―perfect‖ research tool for applied research on populations of indigenous Mexican
immigrants. Pursuit of perfection and reliance on ―conventional wisdom‖ is definitely the arch-
enemy of sound research on minority populations. Our experience suggests that the best research
on ―hard-to-research‖ populations is that which asks relevant and useful questions and chooses
from a ―toolbox‖ of field research techniques the best approach for answering the central research
questions on a study by study basis.

We conclude this paper by briefly pointing to several research methodologies which have been
used to address problems inherent in conducting sound research in the rural communities where
so many of the indigenous immigrants to California live and work. We believe that all of these
methodologies can and should become part of the inventory of research tools which indigenous
immigrants’ organizations can use to explore their own research agenda. While high-quality
multi-stage random sampling is very expensive and, thus, the more affordable research strategies
we describe have the advantage of allowing community field researchers to draw on their stores
of social capital and funds of knowledge of community life to elicit more accurate information
than better-funded technically superior research which is ethnocentric in the modes it uses to
communicate with survey respondents.

Network-Based “Snowball” Sampling and Ethnographic Research

Some of the most useful research on California immigrants in rural communities has been based
on snowball sampling. A ground-breaking study of rural immigrants’ health care needs, ―The
Tulare County Health Study‖ was conducted by Richard Mines and Michael Kearney two
decades ago (in 1982) based on snowball sampling. A similar approach was used by Mines and
Martin in a 1986 EDD-sponsored study which provided the first profile of California
farmworkers recognizable as corresponding to the actual working conditions and population
profile of farmworkers.29 The CIRS research ―survey‖ of Oaxacan village networks in California
agriculture (Kearny and Runsten 1991) used this technique. This study is a good example of the
practical utility of affordable, albeit imperfect, research methodology as well as of the necessity
to provide appropriate explanations about the limitations of the research.30

Snowball sampling is a fundamental technique used in ethnographic research. The important
advantages of this approach relate not simply to reliability but to the researchers’ ability to elicit
important insights in the course of the research. Our sense is that the value of such
methodologies is evident in the utility of the research carried out using this approach—most
notably the work of Michael Kearney and Bonnie Bade but, also, the excellent research by Fred
Krissman. Bade’s ethnographic work has, for example, done a better job of highlighting critical
issues regarding indigenous immigrants’ access to health care than much more costly research
designs such as the California Agricultural Worker Health Survey (CAWHS) conducted by
CIRS—because her research was better designed to address priority issues than the ―generic‖
research approach of the CAWHS.31

Kissam is using ethnosurvey strategies in his community case study research on social dynamics
and civic life in two communities with concentrations of Mixtecos—Arvin, CA and Woodburn,
OR. What we have discovered in these communities, one (Arvin) with a population of about
8,000 immigrants, the other with a population of perhaps 6,000 immigrants (Woodburn) is that in
communities of this size, network-based snowball sampling is feasible—because the networks are
not so large nor are there so many networks to explore that costs skyrocket out of control.

Service Population Studies

In certain circumstances, service population studies have merit. Census Bureau researchers, for
example, have conducted excellent research on the potential of using school-based counts of
children to assess the extent to which the decennial census failed to enumerate school-age
children. The methodology is convincing and the selection bias is not serious when the

  Many of the interviewers in this study were former farmworkers employed by the California monitor-
advocate. The profile of farmworkers in the Mines and Martin report is likely to have much more closely
approximated the true profile of the 1985 California farmworker population than the 1980 census which
was the basis for grossly inequitable national distribution of funds for farmworker programs.
  The principal investigators, David Runsten and Michael Kearney, were, for example, clear that the study
did not seek to estimate numbers of Mixtecos from Guerrero or Puebla and that the original focus had been
on enumerating Mixtec and Triqui persons, resulting in an under-estimate of Zapotecos.
  Interestingly, the CAWHS ―rediscovered‖ a number of farmworker health issues which had been
highlighted more than two decades earlier by the California Raza health Alliance more than 20 years
before—in 1979.

population is limited to elementary school-aged children.32 If indigenous community-based
organizations were able to convince local school districts to conduct a ―special census‖ of
immigrant students developed specifically with the goal of determining the extent of ethnic and
linguistic diversity within the population and designed specifically to overcome the problems
with the existing language census, such research could have a positive impact on schools’ ability
to respond to their students’ needs as well as student’s educational experience.

It is clear that there would be sample bias affecting the reliability of certain research findings
from studies of service populations of indigenous immigrants to California, but appropriately-
conditioned studies might have utility. For example, it seems quite useful to consider using
hometown associations’ membership lists as the basis for a variety of exploratory research efforts,
while understanding that, by definition, the surveyed population represents only a self-selected

Household Samples in the Context of Community Case Studies

A commonly used methodology for applied research in contemporary contexts is random-digit
dialing phone surveys. We have consistently opposed such approaches to research on California
immigrants because we know that a significant proportion of the rural California immigrant
population who are the focus of our research (18-30% of the population universe) do not have
listed phone numbers. The type of bias in samples of persons with phone numbers is, from our
perspective, so serious as to make the exercise unproductive.

The intermediate-cost approach we have used is to conduct household-oriented community case
studies where the initial sampling frame is purposefully chosen, e.g. persons living in Arvin, or
Parlier. We believe this is a particularly responsible approach in rural California farmworker
communities because the diversity within a community can approximate the diversity between
communities. We would prefer to be able to present reliable information about the diversity of
immigrants within Farmersville or Arvin than to present an unreliable picture of rural immigrant
communities based on a phone survey of Spanish-surname persons, a universe which would
disproportionately represent U.S.-born Hispanics and long-term settled immigrants and
underrepresent recently-arrived, predominantly indigenous transnational migrants.33 Such studies
are not cheap and there is the possibility of sample bias due to refusals. However, our experience
using these techniques in Parlier (Farm Labor Supply Study), in Long Beach, Redwood City, and
Sanger (Survey of Limited-English Latino’s Adult Education Needs), as well as in Lindsay and
Winters (Central Valley Partnership Evaluation Community Survey) is that refusals did not
seriously jeopardize the research.34

The CRLA research on census undercount in rural farmworker communities used a community
case study approach. What worked well in this study was that CRLA community workers,
several of whom were Mixtecos, were able to rely on their local knowledge of the study

  Selection bias would become an issue if a school-based study of middle school and teenagers’
educational needs or overall perspectives were needed.
   Our mention of Farmersville is purposive since it is in this community that Fred Krissman’s dissertation
research showed that there were at least two major networks in the community, Mixtecos and Zacatecanos,
and that even within these networks there was a broad spectrum of migration patterns and living
circumstances. In Arvin, we know that a similar situations exists with major differences between long-term
migration networks of Guanajuatenses and more recent Mixteco networks.
     A critical practical issue relates to procedures for replacing non-responding households.

communities to identify study blocks which would be at risk of census undercount and,
subsequently, establish rapport with persons in the block to secure detailed information on their
experience in Census 2000.35 The drawbacks to the methodology in this context is that we did
not have the resources to draw a multi-stage random sample of study blocks in the case study
communities and, thus, cannot definitively demonstrate that the overall rate of undercount in the
community (although the at-risk blocks were not noticeably different than adjacent blocks).

Intercept Surveys

Intercept survey techniques are widely used in market research because they are affordable and
yield rapid turnaround of research. The sampling frame for intercept surveys is really a venue, a
place frequented by a cross-section of the study population. Typically, intercept studies have
interviewed people outside supermarkets, in shopping malls and such. There are, of course, two
key concerns regarding the reliability of this strategy. The first concern is to correctly understand
and describe the overall population frequenting the venue and to assess the extent to which this
population is representative of the desired study population or not. The second concern is to
assure that the survey techniques do not result in unacceptable sample bias so that significant sub-
groups in the study population are left out of the survey sample.

Anna Garcia explored use of intercept interviews in research to assess the prevalence of smoking
among transnational migrants as part of a research for Radio Bilingue (Kissam and Garcia 1998).
Garcia wisely chose public parks as the venue for her intercepts of transnational migrants and
conducted interviews during the later afternoon during the summer. This sampling approach
worked well we believe—because so few in this population have any other place to ―hang out‖
and because, with time on their hands, virtually all were willing to participate in an interview—at
least if it was not too formal or burdensome. In 2002 we moved this approach forward by
conducting one component of a large-scale audience research project for Radio Bilingue,36 The
TVT/TCE Survey, in a variety of remates in Fresno, Madera, and Tulare counties, relying heavily
on Mixtec-speaking interviewers.37 Our assessment is that this strategy worked well and is a
very promising strategy for community research because the population of immigrants who
frequent the remates seem to represent a very good cross-section of the low-income, immigrants
who make up the research population which interests most service providers who seek to improve
the lives of rural California immigrants.

Kissam, Garcia, Mullenax and their research team used a version of intercept survey techniques
in combination with ethnographic research as the basis for their study of the living and working
conditions of teenage farmworkers—most of whom turned out to be teenagers of indigenous
origin; while the study methodology constrained the team’s ability to generalize from the
research, a more formal approach would not have yielded the practical insights about the sorts of
interventions which would most immediately have a positive impact on their well-being.38 The

   Of the five communities we focus on here, research in three was conducted by Mixtec community
workers, Antonio Flores (in Oxnard), Jesus Estrada (in Santa Maria), and Fausto Sanchez (in Arvin).
   Preliminary findings are presented in Ed Kissam, ―Community Reflections on Radio Bilingue’s Service:
Alternative Approaches to Audience Research on Program Impact‖, Latino Public Radio Summit,
September, 2002.
  Both these interviewers, Rafael Flores and Jorge San Juan, are associated with FIOB. Our observation is
that their ability to establish rapport with interviewees and successfully elicit in-depth candid responses was
even important than their language abilities.

venues for ―intercepting‖ and interviewing the teenagers included several versions of street corner
labor markets—a drive-by bakery where pinqueros congregated to pick up workers and the main
―bus station‖ where troqueros picked up workers, a convenience store parking lot in Arizona used
extensively by transnational migrants, coyotes and raiteros, ―El Parque de los Negros‖ in Madera.
Each venue has its unique challenges but, taken together, they provide a more accurate picture
than formal multi-stage sampling would in exploring the living conditions of this population.

Final Reflections on Appropriate Research Methodologies

All of us who believe that policy decisions and program planning are best made on the basis of
empirical data need to push forward to articulate and implement research agendas which generate
reliable insights. The search for research perfection is a wild goose chase, as even a reasonably
superficial perusal of the history of science will demonstrate. The true value of empirical
research lies, in large measure, on articulating and systematically exploring issues of importance.
The idea that universities or other august research bodies are the only institutions with the
―capacity‖ to do sound research is a prevalent but dangerous one, particularly if community-based
organizations accept this notion and leave it to ―others‖ or ―outsiders‖ to do the research which
might be crucial in pointing the way to more effective or novel strategies to address the problems
their communities face. It is critical at the same time, that community-based organizations
concerned about the well-being of indigenous immigrant communities in California work
diligently to generate the systematic research findings which will complement their first-hand
knowledge and anecdotal evidence about what goes on in these communities, what are the most
pressing community problems and what solutions will be most promising.

Researchers and community activists alike need to work hard at building effective applied
research collaborations. Our experience in the work CRLA and The Aguirre Group/Aguirre
International have done on immigrants and farmworkers, as well as in the Kissam’s research as
part of Radio Bilingue’s audience research program, gives us first-hand knowledge of the benefits
of working in collaboration with community leaders and activists from California indigenous
communities to conduct relevant research on community issues. We face serious challenges in
pursuing a long-term research agenda because our support comes from ―soft money‖. It would be
naive to think that indigenous community organizations will not face similar challenges. This is
even more reason for advocacy and, ultimately, insistence on the principle that indigenous
communities are key stakeholders in social science research on their communities. We look
forward, individually and institutionally, to partnerships working together to better understand
indigenous immigrants’ experiences in their individual, family, and community lives in

Our hope is that improved understanding will result in improved responsiveness of social
programs and institutions. This entails confronting many challenges, analyzing the implications
of available research and advocating for increasingly ―customized‖ social programs and policies
in order to leverage significant impacts in Latino communities which are, despite their apparent
homogeneity tremendously diverse in terms of ethnic, cultural, and linguistic composition, not to
mention immigration status. There is much more work to be done. We will all need to be patient
and persevering.

 Ed Kissam et al, ―No Longer Children: Case Studies of the Living and Working Conditions of the Youth
who Harvest America’s Crops‖, Aguirre International, 2001.