Journal of Data Science

Abstract: Loss of household income and purchasing power are shown to have broad and negative societal effects. The economic anxiety accompanying this loss has its strongest impact on consumer demand, which is the major factor in a nation’s gross domestic product (GDP). Negative effects of economic anxiety are also found on the propensity to vote, political trust, societal satisfaction, and the quality of life. These effects were verified in a cross national sample from the fifth round of the European Social Survey. Simple regression of the true value of consumer demand, etc. on the true value of economic anxiety is made possible by an estimate of the reliability of our economic-anxiety score (cf. Bechtel, 2010; 2011; 2012). This reliability estimate corrects the regression slope of each societal variable for measurement error in the anxiety score.

Abstract: Design-based regression regards the survey response as a constant waiting to be observed. Bechtel (2007) replaced this constant with the sum of a fixed true value and a random measurement error. The present paper relaxes the assumption that the expected error is zero within a survey respondent. It also allows measurement errors in predictor variables as well as in the response variable. Reasonable assumptions about these errors over respondents, along with coefficient alpha in psychological test theory, enable the regression of true responses on true predictors. This resolves two major issues in survey regression, i.e. errors in variables and item non-response. The usefulness of this resolution is demonstrated with three large datasets collected by the European Social Survey in 2002, 2004 and 2006. The paper concludes with implications of true-value regression for survey theory and practice and for surveying large world populations.

Abstract: The present paper addresses the propensity to vote with data from the third and fourth rounds of the European Social Survey. The regression of voting propensities on true predictor scores is made possible by estimates of predictor reliabilities (Bechtel, 2010; 2011). This resolves two major problems in binary regression, i.e. errors in variables and imputation errors. These resolutions are attained by a pure randomization theory that incorporates fixed measurement error in design-based regression. This type of weighted regression has long been preferred by statistical agencies and polling organizations for sampling large populations.