A CONSUMER MAIL PANEL RESEARCH METHODOLOGY FOR SURVEYING ETHNIC POPULATIONS

Ethnic populations have always posed special challenges for survey research. A minor obstacle is that Blacks and Hispanics are severely underrepresented in consumer mail panels; but because panels are so large, this first problem is easily solved. Through a simple, commonplace function of "weighting" or "sample balancing", deficient groups are easily restored to their true values in the population.

The second problem is more complex. When compared with their "real world" counterparts, ethnic minorities recruited for mail panels tend to be more literate, consumer-savvy, upscale and in general, more "mainstream". In a word, Black and Hispanic mail panel members do not reflect their true ethnicity; because lower income, less educated inner city residents are not easily recruited as mail panel respondents. To compound the problem, a fundamental condition of mail panel participation is literacy in English — a requirement that automatically excludes significant numbers of both legal and undocumented immigrants of all nationalities. Indeed, while Asians are generally more upscale than even mainstream Whites, language impediments may hinder this group's mail panel membership — especially among older constituents. Among Hispanics, the problem is far more acute.

For all these reasons, American Sports Data, Inc. some years ago ceased the tabulation and publication of ethnic sports participation statistics. In 2005 however, when resources permitted a new methodology (a diverse, cherry-picked oversample of minority populations that within certain caveats would be nationally representative) ethnic projections of sports participation in the U.S. became feasible.

In the actual U.S. population aged six years and older, Blacks account for 11.9% of all individuals; in the TNS/NFO consumer mail panel, their presence is only 6.4%. Hispanics suffer an even greater deficiency: 14.4% in the population versus 3.8% for the panel. Asians represent only 1.5% of panel members, but 4% of the U.S. population.

Across the U.S., Black households have a median income of only $29,470, compared with $44,517 for Whites; 43% of all African-American households have incomes of less than $25,000, and nearly 1 in 5 (19%) is headed by a single woman with children. Although educational attainment is on the rise, only 4.8% of the Black population over age 25 held advanced degrees in the year 2000. For Whites, the comparable statistic for postgraduate education was 9.5%.

Descriptions vary widely, but the most convenient census definition of Gen Y is age 5-24. 34% of all Blacks are members of Generation Y — the much publicized psychographic which happens also to be an exceedingly important sports participation cohort. To the extent that young Blacks have been under represented in surveys, overall Black participation rates have been underestimated. This deficiency will be remedied in the present study.

Hispanics have a similar, even younger socioeconomic profile. Latino households registered a median income of only $33,565, 36% earning less than $25,000. Generation Y comprises 38% of this youthful minority, which also lags Blacks in academic achievement; only 3.8% of Latinos over 35 hold advanced degrees.

But a huge unknown percentage of the U.S. Latino population speaks only Spanish. And since a precondition of consumer mail panel research is literacy (if not fluency) in English, the language gap looms as the most obvious methodological flaw in a self-administered mail survey of the Hispanic population.

The most conservative estimates place the number of undocumented Hispanics in the U.S. at 6 million: a huge "unofficial" population unrecorded in Census tabulations — and by extension, not counted in surveys of consumer behavior. To this unknown degree, the current survey will understate projections of Hispanic sports participation.

With storied levels of educational achievement and a median household income of $55,521, Asian-Americans occupy the highest rung in the socioeconomic ladder. 17.3% hold advanced degrees, versus only 8.9% for the nation as a whole! On the other hand, a large percentage of older Asians are not fluent in English, suggesting that consumer mail panel membership (though not necessarily survey projections) may be skewed to some unknown youthful extent.

With a median household income of only $30,599 and low educational attainment, Native Americans rank near the bottom of the social hierarchy. Like other minorities, they have a substantial Gen Y contingent (37%) and a large corollary pool of sports participants; however, with only 2.5 million people (0.9% of the U.S. population) their small numbers cannot support detailed statistical segmentation in the current report.

In sum, the research challenge posed by ethnic and racial minorities is twofold: there are simply too few in consumer mail panels, and respondent profiles of successfully recruited minorities do not reflect true multicultural diversity. This first problem has a simple and straightforward solution: amass sufficient numbers of ethnic minority respondents necessary to support a minimal level of statistical analysis. With 500,000 member households, the TNS/NFO mail panel contains thousands of panelists who, when "oversampled", can easily satisfy the "hard-to-fill" minority quotas that collectively mirror the U.S. population.

Since existing Black and Hispanic mail panel members are generally far more upscale than their counterparts in the general population, the second remedy was to achieve the right "mix" of these population segments; adequate proportions of inner city residents with low income and education levels, large numbers of children, and in the case of Blacks — a large percentage of female-headed households.

It was decided that an augmentation of 3,000 minority questionnaires (targeted to 1,000 Blacks, 1,600 Hispanics and 400 Asians) could largely compensate for both the overall panel deficit in minorities and the specific shortfalls in racial mix for each group. With an oversample emphasizing youthful downscale ethnic minorities in urban settings (and rural areas), a new expanded SUPERSTUDY® would swell to a mailout of 28,000 questionnaires.

The final step was the most crucial: to ensure that panelists who responded to the survey would be properly weighted to reflect the true composition of minority sub-groups in the U.S. population. To derive sample balancing targets, 4 matrices were constructed. Based on year 2000 Census data and comprised of race, age, gender, household income, market size, and region, ASD was now able to weight each respondent in the raw sample (which had been divided into 170 sub-groups) according to the degree his or her particular segment deviated from the "true" national proportion. Theoretically — because her "type" was deficient in the sample — a female Hispanic aged 66 might be assigned a weight of 3.25. Conversely, a Black male with an income of $75,000 from an urban area might be over-represented, and therefore receive a weight of only .85. The sample balancing variables underlying the matrices were as follows:

A sufficiency of female-headed Black households (and their distribution) precluded the need for sample balancing on this variable.

Since children under 18 cannot be assigned an educational "attainment" level, a sample containing younger age groups cannot be balanced on this variable. However, household income generally serves as an excellent proxy for education, and the survey is not likely to suffer from this potential imbalance.

In sum, language barriers pose the far more serious issue in consumer surveys. Legions of undocumented, youthful Hispanic "aliens" who escape the Census may severely understate the absolute projected numbers of sports participants (but not participation rates) for Hispanics in general.