On the accuracy of economic observations

Abstract

Economic statistics are frequently presented with a degree of precision that disregards fundamental observational errors inherent in social data. Unlike the physical sciences, the social sciences lack a robust tradition of quantifying measurement uncertainty, resulting in “specious accuracy” where statistically insignificant digits inform critical policy decisions. Most economic data are administrative by-products rather than results of designed experiments, making them vulnerable to deliberate misrepresentation, shifting classifications, and temporal inconsistencies. Significant discrepancies between international trading partners and the frequent, large-scale revisions of national income accounts demonstrate that observational error often reaches margins of five to ten percent. These uncertainties undermine the mathematical stability of high-order econometric models, as small variations in input parameters can produce radically different solutions. Furthermore, the computation of national growth rates is exceptionally sensitive to base-year errors, often rendering year-to-year changes statistically indistinguishable from random noise. Effective economic analysis and decision-making require a fundamental adjustment to the reality of these limitations, specifically through the systematic reporting of error estimates. Theoretical frameworks must be reconciled with the actual resolution of available information to avoid drawing unwarranted conclusions from imprecise aggregates. – AI-generated abstract.