The results indicste that the photographic method is suf- 
ficiently accurate for obtaining housing data according to the 
previously described categories of prevalence of single-family homes 
and density of dwellings per block. Of particular interest was the 
finding that the photo interpretation errors were not distributed 
randomly. There were, for example, consistent overestimates of 
numbers of single-family structures, and consistent under-estimates 
of numbers of double-unit residences. This situation provided a com- 
pensating effect, producing a Mnett accuracy figure. In addition, 
the amount of absolute error (each individual photo interpretation 
discrepancy) was found to increase in areas having higher prevalence 
of multi-unit structures. These two findings concerning amount and 
direction of absolute errors indicate a basis for constructing 
systematic correction factors for further study in the continuing 
methodological development. 
For immediate purposes, the specific findings are summarized 
as follows: 1. Ninety-nine percent of the existing total of 3623 
residential structures studied were correctly identified. 2. The 
total number of individual dwelling-units in all types of structures 
was underestimated by seven percent. 3. The overall average density 
of dwelling-units per block was underestimated by 1.7; the range of 
error was from O to -7,1 units per block, ke The percentage of single 
dwelling-units, detached was overestimated by 5,3 percent; the range 
of error was from O to 15 percent, all discrepancies being under- 
enumerations, 
Although the absolute errors were present as indicated, the final 
evaluation of the photographic interpretation observations (net error) 
disclosed a comparatively accurate picture of the relative structural 
characteristics of the subareas under study. The consistency of the 
error direction resulted in each subarea being interpreted as having 
slightly greater proportions of single-family homes and slightly 
less dwelling-unit density than the ground contro) data revealed, 
Consequently, when the seventeen subareas were ranked according to 
the photo data on these two primary interest items, the correlations 
with their ranks according to the ground data were found to be ,98 and ,99 
respectively, The figures supporting these accuracy measures are 
shown in Tables 1 and 2.