Primary family.--A primary family consists of two or more persons including the head of the useholds and all (one or more) persons in the household who are related to the head by blood, arriage, or adoption. It may include a subPily consisting of a married couple with or

rout children, or one parent with one or more ildren under 18 years of age, living in the usehold and related to but not including the ead of the household or his wife. The subfamily considered a part of the primary family for :ll tabulation purposes.

3. The amount of other money income received in 1949, such as interest, dividends, veterans' allowances, pensions, or rents.

Gross rent.--Monthly gross rent is contract rent plus the reported average monthly cost of water, electricity, gas, and other fuel paid for by the renter. If furniture is included in the contract rent, the reported estimated rent of the dwelling unit without furniture is used for the computation instead of the contract rent.

SOURCE AND RELIABILITY OF THE DATA

The data in tables 1 through 4 are based on complete counts, including all dwelling units and families with the specified characteristics. The distributions involving income in tables 48 and 5, however, were prepared from data collected on a sample basis. As part of the 1950 Census, family income data were obtained from about onefifth of all families. For these tabulations, additional interviews were made among nonwhite families to increase the income sample above the 20 percent level. This was accomplished by a subsequent field enumeration of a sample of nonwhite families who were not in the original sample but were living in substandard dwelling

units.

Although the figures shown in tables 1 through 4 are based on the same data as the forthcoming 1950 Census tabulations of these items, they may differ from those to be published as part of the Census. The present tables were obtained by tabulating directly the data as transcribed in the field offices of the Census Bureau and the figures, therefore, do not include all of the refinements that result from a careful examination of the schedules.

In addition, the data are subject to other biases due to errors of response or nonreporting which cannot be corrected in editing. Factors affecting the accuracy of enumeration include the respondent's knowledge of the facts and the ability of the enumerator to obtain accurate information on such items as income, rent, dilapidation, and plumbing facilities. The regular 1950 tabulations are also subject to these

biases.

Reliability of Estimates in Tables 48 and 5

Because of sampling variability, per ntage distributions shown in table 4a and all figures in table 5 may differ from those that would have been obtained from a complete count. (The

of 10 percent living in subthe sampling

To illustrate, for a figure based on white primary families standard renter dwelling units, variability is 3.2 percent; the chances are about 19 out of 20 that the percentage obtained by a complete census would have been between c.8 percent and 13.2 percent.

The sampling variability of a specified percentage of total primary families with designated characteristics will vary according to the proportion of white families and nonwhite families making up this percentage. For example, consider the sampling variability of a figure of 10 percent based on total primary families. The maximum sampling error to be expected of such a figure would occur when the entire 10 percent includes only white primary families and the chances are about 19 out of 20 that this sampling error would not exceed 2.5 percent. The minimum sampling error would occur when the entire 10 percent includes only nonwhite primary families and the chances are 19 out of 20 that this sampling error would not exceed 0.8 percent. For other specific characteristics com❤ posed of 10 percent of total primary families the sampling variability may assume any value