RATIO VARIABLES IN AGGREGATE DATA ANALYSIS - THEIR USES, PROBLEMS, AND ALTERNATIVES

• NCJ Number: 57691
• Journal: Sociological Methods and Research Volume: 7 Issue: 4 Dated: SPECIAL ISSUE (MAY 1979) Pages: 431-450
• Author(s): K A BOLLEN; S WARD
• Date Published: 1979
• Length: 20 pages
• Corporate Author: Sage Publications, Inc
• Publication Format: Article
• Language: English
• Country: United States of America

Annotation

AN OVERVIEW IS PROVIDED OF ISSUES THAT HAVE BEEN RAISED OVER THE USE OF RATIOS IN CORRELATION AND REGRESSION ANALYSES; SUGGESTED GUIDELINES FOR RESEARCHERS WORKING WITH RATIO DATA ARE INCLUDED.

Abstract

THERE ARE AT LEAST THREE SEPARATE CONFIGURATIONS OF THE RATIO PROBLEM IN SOCIAL SCIENCE RESEARCH. THE FIRST OF THESE IS THE THEORETICAL CONFIGURATION WHERE THE ISSUE IS THE NATURE OF THE RELATIONSHIP BETWEEN THE VARIABLES. IN THE USE OF RATIOS AS INDICES OF CONCEPTS, A PROBLEM CAN ARISE IF IT IS REGRESSED ON OTHER INDICES OR VARIABLES THAT CONTAIN A COMMON COMPONENT. FOR EXAMPLE, THE RELATIONSHIP BETWEEN TWO PER CAPITA MEASURES MAY BE CONFOUNDED WITH THE COMMON POPULATION COMPONENT IN EACH VARIABLE. THE SECOND ISSUE IS THE CONTROL CONFIGURATION THAT INVOLVES THE USE OF RATIOS TO STANDARDIZE FOR A THIRD VARIABLE, AND RESULTS SHOW THAT ONLY UNDER EXCEPTIONAL CONDITIONS WILL RATIO VARIABLES BE A SUITABLE MEANS OF CONTROLLING AN EXTRANEOUS FACTOR. THIRDLY, THERE IS THE USE OF RATIOS TO CORRECT FOR HETEROSCEDASTICITY. ALL THREE USES OF RATIOS MAY INVOLVE PROBLEMS WITH INFERENCE. IN PROVIDING GUIDELINES FOR RESEARCHERS WORKING WITH RATIO DATA A NUMBER OF ALTERNATIVES MAY BE SUGGESTED. THREE EXAMPLES OF ALTERNATIVES RELATED TO THE THREE RATIO CONFIGURATIONS ARE (1) FOR RATIOS AS VARIABLES OF THEORETICAL INTEREST, THE RESEARCH PROBLEM SHOULD, IF POSSIBLE BE REFORMULATED IN TERMS OF NONRATIO VARIABLES; (2) FOR RATIOS AS METHODS OF CONTROL, RESIDUALIZING TECHNIQUES MAY BE USED TO STATISTICALLY REMOVE THE EFFECTS OF THE 'DEFLATING' VARIABLE; AND (3) FOR RATIOS AS CORRECTION FOR HETEROSCEDASTICITY, IT IS SUGGESTED THAT THE ORIGINAL FORM OF THE MODEL SHOULD BE DETERMINED. THE MAJOR ARGUMENT HERE IS THAT INDISCRIMINATE USE OF RATIOS SHOULD BE AVOIDED. THE BEST RESEARCH STRATEGY MIGHT BE THE USE OF MULTIPLE TECHNIQUES AS A CHECK ON FINDINGS. ANY AMBIGUOUS OR CONTRADICTORY FINDINGS CAN THEN BE REPORTED AND AN ATTEMPT MADE TO EXPLAIN THEM. IT IS CONCLUDED THAT SUCH A COMPARATIVE APPROACH TO DATA ANALYSIS WILL INCREASE THE CERTAINTY THAT PARAMETER ESTIMATES REFLECT 'REAL' SOCIAL PROCESSES RATHER THAN STATISTICAL ARTIFACT. NOTES AND REFERENCES ARE PROVIDED. (MHP)