1970 HUNGARIAN CENSUS OF POPULATION Information on the data collection and processing (1977)

I. THE COMPREHENSIVE REWIEV OF THE DATA COLLECTION AND PROCESSING - 11. The control of the reliability of the census

then the gross error is: G = b + c Its proportion to the census data of the criterion: r _ b + c gl a + b (It must be remarked that in the literature comparison is frequently made to the data of post-enumeration control (a + c). This method of comparison is justified by the fact that the post-enumeration control is regarded as the more reliable survey; however, comparison with the census returns expresses better its degree of reliability. For this reason the latter method was used in our calculations. ) Comparison with the total number of items: b + c Similarly, in case of net error: H = b - c b - c h, = 1 a + b b - c h2 n It is convenient to make a distinction between response dispersion and bias. The former occurs when the errors of declaration are dispersed at random, i. e. merely from inaccuracy, bad memory or are not the own declarations of the person questioned. Bias is a one-way error. It is usual to calculate the standard error of response dispersion, which is generally the square root of the gross error. Ő = \/ b + c, which is also usually expressed in rates. This can be compared with the extent of the net error. Also the one-way development of the sing of the net error hints at bias. A further index is the so-called index of inconsistency. Its formula is as follows: i . h. pq where p is the relative frequency of the criterion under survey, q = 1 - p. The index essentially expresses the share taken by the actual response dispersion as well as its square in the variance produced merely at random. In the literature the denominator frequently contains 2pq, expressing that half of the errors is attributed to the census, while pq is used as the mean of the rates of the two censuses. Following from what was said concerning the gross error, only census rates are used in our calculations, with the ommission of the multiplier 2. Namely in view of the fact that the post-enumeration control is considered as a more accurate survey, this method seems more proper, though we are aware of the fact that even this only approximates reality. The index of inconsistency is a further index-number of the reliability of census returns, its lower value means higher accuracy. 7. Népszámlálás 97