Tag
measurement_error
11 findings
Labour & employmentActual vs. Potential Experience BiasTraditional potential experience formulas understate experience payoffs, whereas tracking actual 5-day work experience demonstrates an average daily wage increase of $3.83 to $4.19 (an approximate 80% daily wage increase, p < 0.05), which equals nearly 10 years of potential experience in non-experimental Malawian estimates.Observational StudyEconomics (general)Administrative Social Security Data Linkage for Earnings MeasurementAdministrative linkage secured 100% verified annual earnings histories across 13 consecutive years per individual, confirming pre-lottery earnings parallel trends between nonwinners ($15,600 to $18,000) and winners ($12,000 to $14,500) while eliminating recall measurement bias.Observational StudyEducationEmpirical Bayes Shrinkage Estimation for Volatility AdjustmentUsing raw value-added scores without shrinkage generated actual achievement differences following random assignment that were only 43% as large as predicted (coefficient = 0.430, SE = 0.057). Single-year math value-added required multiplication by shrinkage factors of 0.396 in elementary and 0.512 in middle school to predict future performance.RCTMethods & evidenceHigh-Frequency Survey Recall ValidationDaily visits averaged 8.11 GH¢ cash in and 3.32 GH¢ cash out. Comparing monthly recall to aggregated daily data showed no statistically significant difference in revenue (p-value > 0.10, difference of -4.08 GH¢) or expenses (difference of 8.71 GH¢), proving that monthly recall surveys do not introduce strategic reporting bias in consulting evaluations, though daily data exhibited higher variance for record-keepers (squared difference 22,895 GH¢, p < 0.05).RCTEconomics (general)Inattention Mixture Model CorrectionInattention affects choice variance without biasing central tendencies; the inattention-corrected flow value parameter ($\tilde{z} = 0.58, s.e. = 0.04$) matches the standard logit baseline ($\tilde{z} = 0.58, s.e. = 0.04$) identically across all hour choices.RCTEconomics (general)Instrumental Variables Correction for Management Measurement ErrorThe IV point estimate for Total Factor Productivity (TFP) was 0.477 (p < 0.05, First-stage F-stat = 86.00) compared to an OLS estimate of 0.167, while the IV effect on quality defects was -1.694 compared to -0.558 under OLS, proving that uninstrumented estimates suffer from substantial downward attenuation bias.RCTLabour & employmentPotential Experience Measurement BiasNonexperimental wage returns relying on potential experience proxies require approximately 10 years of estimated potential experience in Malawi to match the wage gain produced by 5 days of actual measured work experience, demonstrating that potential experience drastically underestimates true per-day returns while overstating accumulated work history.Observational StudyLabour & employmentResume-Survey Experience Measurement DiscrepancyWhile only 15% of respondents reported no prior work experience in self-reported survey modules, 35% lacked work experience on their submitted résumés, revealing a 20 percentage point discrepancy caused by infrequent résumé updates and differing worker definitions of suitable CV entries.Observational StudyEconomics (general)Short-Term Contract Variance DilutionFixed-term and temporary contracts of 1 week or less comprised 53% of all contract flows, generating a standard deviation of hiring 9.9 times larger than permanent contracts and obscuring overall hiring treatment effects.RCTEducationSignal Variance Decomposition of Teacher EffectsThe estimated signal standard deviation of true teacher effects ranged from 0.16 to 0.18 student-level standard deviations in math and 0.10 to 0.12 in English Language Arts (ELA). ELA demonstrated lower signal variance and higher year-to-year volatility due to reliance on multiple-choice reading comprehension items without direct writing assessments.RCTEducationValue-Added Extreme Outlier Trimming and Anomaly DetectionIn English language arts, including an extreme outlier teacher with a 1.666 SD value-added swing dropped the predictive coefficient to an imprecise 0.375 (SE = 0.264); trimming the top 1% restored the ELA coefficient to 0.697 (p < 0.01, SE = 0.213) and overall composite predictive coefficient to 0.955 (SE = 0.123).RCT