Item Nonresponse, Allocation, and Data Editing of the Question on Hispanic Origin in the American Community Survey (ACS): 2000 to 2007

August 2010
Working Paper Number: POP-WP086
Roberto R. Ramirez and Sharon R. Ennis


The question on Hispanic origin was first introduced in the 1970 Census and subsequently has  been included in every census and major national household survey questionnaire since then (i.e.,  Current Population Survey (CPS), American Community Survey (ACS), and Survey of Income  and Program Participation (SIPP)). Over the last 35 years the question on Hispanic origin has  undergone numerous changes and modifications all with the aim to improve the quality of  Hispanic origin data in the United States and surrounding territories (Alberti 2006, Martin et al.  2004, 2002, Cresce and Ramirez 2003, Cresce et al. 2001). Modifications made to the Hispanic  origin question for the 2010 Census include reinstating the term “origin” and targeting the  Hispanic origin write-in line; and providing detailed Hispanic origin groups, such as Dominicans  and Colombians, as examples of “Other Hispanic or Latino.” Results from research studies  conducted by the Census Bureau in the last three decades show that the question on Hispanic  origin has performed relatively well in the last four population censuses (1970, 1980, 1990, and  2000) as measured by low item nonresponse and allocation rates; although this is not to say the  question itself is free of problems (Ramirez 2005, Martin 2002, Cresce and Ramirez 2003,  Cresce et al. 2001). One notable issue documented in Census 2000 was the increase of general  response reporting (i.e., “Hispanic,” “Latino”) among Hispanics at the expense of specific  detailed origin reporting. This was largely attributed to the absence of specific Hispanic origin  examples on the questionnaire (Martin, 2002). 

Much has been written about the quality of Hispanic origin data from censuses but few studies  have specifically examined the quality of Hispanic origin data from national household surveys.  Considering how important censuses are for the apportionment process and the distribution of  federal funds in the United States, it is no surprise that more attention has been paid to census  counts of the Hispanic population. This may change in the next few years as the ACS, the  largest national household survey in the United States, replaces the traditional census long form  as part of the Census Bureau’s 2010 Decennial Census redesign program. The ACS will gain  more importance as the sole provider of socioeconomic and housing data of the general U.S.  population and its minority groups at the local community level. The 2010 Census will be the  first modern census not to have a long form in 80 years. Given the prominent role the ACS will  play in the upcoming decades in providing official federal socioeconomic and housing statistics  of the nation, one question has come to mind among many researchers inside and outside the  Census Bureau – what is the quality of the Hispanic origin data from the ACS? This paper will  attempt to answer this question. 

April 24, 2009
November 2010