Johnston] NAVAHO POPULATION 147 



scores below 24). Data of poorer quality, finally, would be indicated 

 by scores above 4, 6, and 24, respectively. In general, the higher the 

 scores, the poorer the data.^^ 



In order to properly interpret the scores shown in table 28, two im- 

 portant qualifications must be borne in mind. First, the adjustment 

 tliat is prescribed as an allowance for small populations may tend to 

 obscure real differences in the quality of data between populations 

 that differ markedly in size. As the authors of this joint score measure 

 indicate, its application to populations under about 10,000 is practically 

 meaningless. Similarly, the separate age and sex ratio scores are of 

 questionable validity with populations below 5,000. 



Secondly, it is essential to recognize that a high score is subject to 

 two possible interpretations : either the basic data are inaccurate, or the 

 actual age distribution of the population in question is marked by 

 sufficient irregularities (such as might be produced by the impact of 

 war or migration, for example) to produce high scores despite highly 

 accurate data. 



In view of these limitations, these scores must be regarded only as 

 approximate indicators of the relative accuracy of age distributions 

 obtained from populations that are roughly comparable in their gen- 

 eral characteristics. Wliere the populations in question differ mark- 

 edly either in size or in other characteristics, comparisons between 

 these scores must be regarded with extreme caution. 



The scores obtained from the 25 Navaho age distributions permit of 

 few generalizations, beyond the fact that the basic data, on the whole, 

 can be ranked as "poor" to "fair" in quality. The Navaho sex ratio 

 scores are particularly high as compared to those obtained from the 

 age distributions for the total United States Indian populations. The 

 Navaho age ratio scores are also generally higher than those of the 

 total U.S. Indian populations (table 31).^^ 



^ The cutting points set forth above to distinguish "excellent," "good," "fair," and 

 "poor," data are essentially arbitrary. They are in general agreement, however, with the 

 judgments expressed by the authors of this measure regarding the examples they cite 

 (United Nations, 1952, pp. 70-71),. 



»' The peculiarities of the age-sex distributions obtained for American Indians from the 

 returns of the censuses of 1950 and 1960 merit further comment. In 1950 the dis- 

 tribution displayed a marked excess of males 10 years and over, especially 10 to 14 

 years. Calvin Beale, who has had considerable experience In analyzing these data, sug- 

 gested the possibility of processing error in this regard. iThis notion was confirmed by 

 means of an ingenious investigation carried out hy Ansley J. Coale and Fredericls F. 

 Stephan (1962). Their chief finding was that in processing the 1950 census returns, the 

 IBM card punchers may occasionally have punched data one column to the right of the 

 proper position. Although most such errors would normally have been detected routinely, 

 It was suggested that enough such errors went undetected to produce a large excess of 

 male Indians in the 14 to 24 year group. 



The 1960 census returns, however, have produced a far more glaring discrepancy — a 

 very large excess of Indians of both sexes in the age group 55 to 59 years. A limited 

 check of the data processing instructions and related computer programs has not uncovered 

 evidence of processing error in this instance. It is possible that large numbers of older 

 Indians might have been uncovered in off-reservation areas for the first time in 1960, by 

 virtue of the self -enumeration procedure. But this would not explain the occurrence of the 



Footnote continued on following page. 



