Comparability

(redirected from comparabilities)
Also found in: Dictionary, Thesaurus, Legal, Financial.

Comparability

 

in statistics, an essential prerequisite for the comparison and analysis of statistical indexes. For indexes to be comparable, several conditions must be met. For example, the same methods of computation and the same units of measurement must be used, observations of the indexed phenomenon must be carried out to the same extent, and constant territorial boundaries are required.

Statistical data that are not comparable because of different computation methods are encountered particularly often when statistical indexes from different countries are considered. An example is provided by the national incomes of the USSR and the USA. The values of these national incomes are not directly comparable not only because they are expressed in different monetary units but also because they are calculated by different methods. Such a lack of comparability can occur in the statistics of a single country if the methods of computing an index change. When the comparability of indexes is violated as a result of different computation methods, the indexes can be made comparable with respect to methodology by carrying out the appropriate recalculations.

A more complicated problem arises when indexes are not comparable because of the absence of a clear methodology and the gathering of data in an unscientific manner. In his article “On the Question of Our Factory Statistics” (1898), V. I. Lenin showed that Russian industrial statistics of the second half of the 19th century had year after year used incomparable data on the number of factories. Because a clear definition had not been arrived at as to which enterprises constituted factories, even small workshops were included in the number of factories. In later years, fewer such establishments were counted in the statistics, and the impression was created that the number of factories in Russia was shrinking. By presenting the data in a comparable form, Lenin demonstrated that in actuality the number had increased.

If observations of a phenomenon are made on the basis of a nonrepresentative sample, comparability is violated when the size of the sample is changed. For example, statistics on kolkhoz markets were based on observations in 71 large cities up to 1940 and on observations in 264 cities in the postwar period. Consequently, in order to maintain comparability, kolkhoz-market turnover and price indexes must be calculated on the basis of 71 cities when making comparisons with 1940 and on the basis of 264 cities when making comparisons with 1960.

Comparability is violated when territorial boundaries are changed. As a result, indexes based on the old boundaries must be adjusted in order to make comparisons with indexes based on the new boundaries.

Comparability is also violated when the units of measurement are changed. For example, because of the revaluation of the ruble, pre-1961 data expressed in rubles are not comparable with data for 1961 and subsequent years. Data for the pre-1961 period must thus be converted to comparable values. Because of a change in prices, such indexes as the value of output or of national income for different years may not be comparable. A conversion must therefore be made to comparable prices.

Some complex statistical indexes cannot be directly compared because the phenomenon under consideration exhibits a varying structure. Overall death rates, for example, are not always suitable for making comparative characterizations of mortality in different countries, since overall death rates are affected by the age makeup of the population, which may differ markedly in some cases. For the sake of comparability, death rates must thus be calculated on the basis of the same standard population structure.

The problem of comparability also arises in computing relative and average values in statistics. For example, in calculating the extent to which a plan is fulfilled, the actual fulfillment indexes and the plan indexes must refer to the same group of enterprises; in computing rates of change, the indexes must be given for equal time intervals; and in figuring average earnings, the wage fund must strictly correspond to the number of workers.

N. N. RIAUZOV