|
Chapter 59
MEASUREMENT ERROR IN SURVEY DATA
JOHN BOUND'
University of Michigan and NBER
CHARLES BROWN
University of Michigan and NBER
NANCY MATHIOWETZ
University of Maryland
Contents
Abstract 3707
Keywords 3707
1 Introduction 3708
2 The impact of measurement error on parameter estimates 3710
2.1 Special cases 3712
2.2 General results linear model 3715
2.3 Differential measurement error an example 3716
2.4 Bounding parameter estimates 3721
2.5 Contaminated and corrupted data 3723
2.6 Measurement error in categorical variables 3724
2.7 Nonlinear models 3727
3 Correcting for measurement error 3728
3.1 Instrumental variables in the bivariate linear model 3729
3.2 Multivariate linear model 3733
3.3 Nonlinear models 3735
3.4 The contribution of validation data 3737
4 Approaches to the assessment of measurement error 3740
5 Measurement error and memory: findings from household-based surveys 3743
5.1 Cognitive processes 3743
5.2 Social desirability 3745
5.3 Essential survey conditions 3746
5.4 Applicability of findings to the measurement of economic phenomena 3747
6 Evidence on measurement error in survey reports of labor-related phenomena 3748
6.1 Earnings 3748
6.1 1 Annual earnings 3748
6.1 2 Monthly, weekly, and hourly earnings 3766
6.2 Transfer program income 3770
6.3 Assets 3779
6.4 Hours worked 3784
6.5 Unemployment 3791
6.5 1 Current labor force status, and transitions to and from unemployment 3792
6.5 2 Retrospective unemployment reports 3799
6.6 Industry and occupation 3802
6.7 Tenure, benefits, union coverage, size of establishment, and training 3805
6.8 Measurement error in household reports of health-related variables 3811
6.8 1 Health care utilization, health insurance, and expenditures 3811
6.8 2 Health conditions and health/functional status 3817
6.9 Education 3823
7 Conclusions 3830
References 3833
|