An error in data editing has caused some respondents who should have been coded as speaking a language other than English at home (ESPEAK = 1) to instead be coded as not speaking a language other than English at home (ESPEAK = 2). The error has also caused a smaller, separate group of respondents who should have been coded as not speaking a language other than English at home (ESPEAK = 2) to instead be coded as speaking a language other than English at home (ESPEAK = 1).

    The respondents mistakenly coded as not speaking another language at home should have received a value for the variable TLANG1 to indicate which language other than English they spoke at home, and they also should have received a value for EHOWWELL to indicate how well they spoke English; instead, these respondents were classified as not in universe for TLANG1 and EHOWWELL. For a subset of these respondents mistakenly coded as not speaking a language other than English, the misclassification caused the people in their households—including these respondents themselves—to be categorized as not part of a limited-English-speaking household (i.e., mistakenly categorized as RLNGISOL = 2 when the true value was RLNGISOL = 1).

    As for the respondents accidentally coded as speaking another language at home even though they did not (accidentally coded as ESPEAK = 1), values for TLANG1 and EHOWWELL were allocated for these individuals even though the individuals in fact were out of universe with respect to those variables. The miscoding of these particular individuals did not cause anyone from any household to be mistakenly classified as part of a limited-English-speaking household (i.e., misclassified as RLNGISOL = 1) when in reality the household was not limited English speaking.

    The data-editing error in question affects only a small percentage of TLANG1 and EHOWWELL observations, and it affects even smaller percentages of the ESPEAK and RLNGISOL observations. Data users should be aware of this issue when using ESPEAK, TLANG1, EHOWWELL, or RLNGISOL, particularly if comparing estimates over time.