Skip Navigation
Managing an Identity Crisis: Forum Guide to Implementing New Federal Race and Ethnicity Categories
NFES 2008-802
October 2008

Appendix C: Bridging Basics

For our purposes, "bridging" refers to the process of making race data collected using the 1997 standards comparable to data collected using the previous 1977 standards to allow time trend analyses using those data. Following a shift to the 1997 standards for collecting race and ethnicity data, which include five race categories and offer respondents the opportunity to select multiple races, it may be necessary for agencies to use two sets of data for a finite length of time referred to as a "bridge period." To facilitate the study of historical trends in data collected before and after the shift to the new standards, during this bridge period agencies will not only collect new data along the 1997 guidelines, but may also consider creating a "bridging estimate," defined as a "prediction of how the responses would have been collected and coded under the 1977 standards."2 In other words, the bridge data set estimates how the newly identified multiracial populations would have identified themselves under the old single-race system.

bridging basics

Though bridging estimates will fail to give users completely accurate pictures of the racial and ethnic makeup of populations before and after the standard change, they will provide some approximation that will bridge the gap between old and new data and allow important analyses such as the AYP or other educational, social, or economic trend studies to be conducted.

In this report, for illustrative purposes only, bridging is sometimes described at the individual level. However, it should be noted that the bridging methodologies discussed should be used at the aggregate level in most cases for general education data purposes—that is, they should be applied at the aggregate level to divvy up groups of multiple-race individuals into single-race groups, rather than to assign each individual multirace respondent to a single race category or to divide each individual among the categories.

To whom does bridging apply?

Bridging estimates are only necessary where there are respondents who choose multiple races to describe themselves.3 It is assumed that students who report a single race under the new reporting scheme also chose the same single race under the old scheme. No bridging is necessary for such individuals. Even Hawaiian Natives and Other Pacific Islanders, who are separated from Asians under the new standards, can easily be recoded to align with previous-year standards simply by adding them to the 1977 category "Asian or Other Pacific Islander," which includes the new category. In cases like these, where a clear one-to-one relationship exists between old and new categories, no formal bridging methodology is necessary. However, because the new standards allow respondents to choose multiple races, the job of guessing which single race these multirace individuals would have chosen for themselves if presented with the 1977 standards grows more complicated and some formal bridging methodology, therefore, becomes necessary.

Spare the bridge, omit the child: Opting out of bridging

If an agency decides that a break in historical data is acceptable, it might decide to forgo the bridging process—a decision that may be justifiable in a number of scenarios. For instance, this decision may be acceptable if there is little change in the racial composition of the agency's population over time. In addition, an agency might not bridge its data if the proportion of multiracial students and staff in the agency's population is small enough so as to have only a negligible effect on the agency's race data overall. These agencies may choose to treat data on multirace individuals as missing when calculating time trend analyses rather than attempt to bridge these data.4 However, it should be noted that excluding multirace individuals from such calculations might significantly affect data on minority populations.5 For other agencies that have substantial multirace populations and would like to facilitate longitudinal studies using their data, a number of ways to create bridging estimates are available.

Different uses, different bridges

The choice of a bridging technique depends, in part, upon how the data will be used. Statistical reports that follow the characteristics of a group of students or staff over time can probably be supported by estimates of race and ethnicity distribution derived through bridging at the aggregate level. Civil rights data collections that are concerned with outcomes for individual students may have different requirements for assigning race and ethnicity status.6


2 OMB (2000), p. 85.
3 This is true, with the exception of multirace respondents who choose "Asian" and "Hawaiian or Other Pacific Islander" as their component races under the 1997 Standards.  For these individuals, no bridging is necessary as they can be simply recoded into the 1977 single-race category "Asian or Other Pacific Islander," which includes both 1997 categories.  Census 2000 found that 2.4 percent of the total U.S. population identified with two or more races—a rate that varies widely by state, racial-combination, and age group.  The percentage of the population reporting two or more races ranged from less than 1 percent in Mississippi and West Virginia to more than 21 percent in Hawaii.  White/American Indian or Alaska Native was the most prevalent multirace combination, being selected by more than a million respondents, followed by White/Asian, which was chosen by nearly 900,000 respondents (excluding additional race categories collected in the Census 2000, commonly combined as "some other race," which are not among the five 1997 standards race categories).  The frequency of multirace identity is clearly rising with each new generation. Among the 60- to 64- year age group, only 1.3 percent reported two or more races, while the younger age groups report consistently higher percents (3.1 percent among 15 to 19 year olds; 3.4 percent among 10 to 14 year olds; 4.0 percent among 5 to 9 year olds; and 4.9 percent among children under five years of age).
4 Agencies that do not create bridge estimate data should footnote the first occurrence of data collected under the 1997 standards so users know that the data are not comparable to those of previous years.
5 Ingram (2003), p. 4.
6 Guidelines on how multiple-race responses should be allocated for civil rights enforcement can be found in the OMB bulletin entitled, "Guidance on Aggregation and Allocation of Data on Race for Use in Civil Rights Monitoring and Enforcement." For a discussion on individual-level bridging, see "NCHS Procedures for Multiple-Race and Hispanic Origin Data: Collection, Coding, Editing, and Transmitting."