homes each. These clusters were numbered from 1 to N, where N
represented the number of clusters in a county. Clusters were
numbered consecutively beginning in the northeast corner of
the county and thereafter in a serpentine manner from east to
west, and reverse.2 By the use of a table of random numbers,
the clusters to be included in the study were chosen. When
the desired number of records was obtained in any given county,
interviewing was discontinued.

 Scope of Data Examined
 Certain dependent variables, or characteristics which are
dependent on others, were initially assumed to be statistically
significant, that is, not due to chance.3 These were the cash
surrender value of life insurance policies, equities in family
dwellings, total equities, current investments, Old-Age, Survi-
vors, and Disability Insurance (OASDI) coverage, and pension
plans other than OASDI. Other associated but independent
(well-fixed) factors were race, age, education, occupations, place
of residence and health of the head of the family; also annual
family income, family structure, and leisure activities of family
heads.
 It was hypothesized that if the magnitude of the dependent
variables were known, certain values, such as anticipated re-
tirement incomes and anticipated assets at age 65, could be
computed accurately. It was discovered, however, that this
assumption was not fully supported by the data. There were
elusive factors which could not be identified.

 Data Examined
 Both white and Negro families lived in most of the counties
studied. Approximately 86 per cent of all families were white
and 14 per cent Negro (Table 1). In analyzing the data secured,
various types of findings revealed that attributes affecting re-

 2All counties of western Texas and of the subtropical areas of Florida
were excluded because they were regarded as not representative of the
region under study.
 'Various descriptive statistics such as frequency distributions, means,
standard deviations, and variances, were applied to interpret the results
of this study.
 The chi-square criterion was used to determine the degree of relation-
ship between discrete variables, and product-moment correlations to
ascertain the degree of association between continuous variables. By use
of these analytical methods both dependent and independent variables
were selected for use in least-squares and multiple-regression models.