The United States Census Bureau produced and released Spanish surname products for 1950, 1960, 1970 and 1980. This 1990 version is another way station in an ongoing research journey. This paper, “Building a Spanish Surname List for the 1990’s—A New Approach to an Old Problem,” differs from its predecessors in two significant respects.

1. Until 1990, name has never been part of a permanent Census electronic record. Following the 1990 Census, the Census Bureau appended name to 7 million Census records for the purposes of determining undercount. The “List” is constructed by tabulating the responses (surname by surname) to the Spanish origin question for persons in that sample. Well over 90 percent of male householders with the surnames: GARCIA, MARTINEZ, RODRIGUEZ, and LOPEZ responded affirmatively to the Spanish origin question while less than 1.0 percent of male householders named SMITH, JOHNSON, and BROWN provided a positive response to the Spanish origin question.
2. In the past, a name was either on the list (e.g., Garcia) and was taken to be Spanish or it did not appear on the list. The assumption was that any name not on the list was not Spanish. Since neither BROWN nor SILVA appeared on the 1980 Spanish Surname list, one would naturally assume that neither name was Spanish. In the electronic version of the 1990 “List” we append auxiliary data for 25,000 surnames including both SILVA and BROWN that allow users to form their own lists. Almost 60 percent of the SILVA’s in our 1990 Census sample responded that they were Hispanic while less than 1 percent of BROWN’s claimed to be Hispanic. Moreover, another auxiliary item suggests that the letters S I L V A form a potentially Spanish word. That same statement cannot be made for B R O W N. From this data, some users might include SILVA on their own personal Spanish surname list, while others would justifiably arrive at an opposite conclusion.

We must emphasize that this product does not violate the confidentiality of Census responses. On average, each captured surname represents about 40 householders. Moreover, we provide no subnational geographic data nor is there any indication of first name or age of respondent. Given these conditions, we are confident that this file does not provide information that could identify any individual enumerated in the 1990 Census.