Improving GDP Measurement: A Measurement-Error Perspective

Improving the Accuracy of Economic Measurement with Multiple Data Sources: The Case of Payroll Employment Data

“…Our approach in the present paper is different from those above in that we explicitly inves-tigate the usefulness of ADP as a supplement to CES data for tracking the underlying state of the labor market. In this respect, our work is inspired by Aruoba et al (2016) who note difficulties in assessing the growth of aggregate output in real time given limitations on the comprehensiveness and timeliness of GDP measures. Two independent measures of GDP exist-the commonly reported expenditure-side approach and the income-based approach-and both are prone to measurement errors arising from various sources.…”

“…Two independent measures of GDP exist-the commonly reported expenditure-side approach and the income-based approach-and both are prone to measurement errors arising from various sources. Aruoba et al (2016) combine the two measures using a state-space framework, recovering an underlying state of output growth which they label "gross domestic output". We follow this general approach with a focus on employment rather than output.…”

“…As such, both CES and ADP-FRB are (approximately) truth plus mean-zero sampling error. This sampling error is captured by the 19 Aruoba et al (2016) use a similar approach to provide a better measure of output. 20 The approach is in contrast to Aruoba et al (2013), who assume that both the observation variables in their paper (GDP and GDI) have unit loadings on the unobserved state variable.…”

Normality tests for latent variables

“…Traditionally, the difference between the two, officially known as the “statistical discrepancy” (see Grimm ()), was regarded by many academic economists as a curiosity in the US National Input and Product Accounts (NIPA) elaborated by the Bureau of Economic Analysis (BEA) of the Department of Commerce. However, the Great Recession substantially renewed interest in the possibility of obtaining more reliable GDP growth figures by combining the two measures (see, e.g., Nalewaik (, ), Greenaway‐McGrevy (), and especially Aruoba, Diebold, Nalewaik, Schorfheide, and Song (), which provides the background for the Philadelphia Fed GDPplus measure). Some national statistical offices compute a simple equally weighted average of the different aggregate series, and in fact, BEA began providing this average in 2015.…”

“…We also assume that the two measurement errors are uncorrelated, which guarantees the nonparametric identification of the signal from the noise (see Almuzara, Fiorentini, and Sentana () for further details). The fact that the two measures of output are obtained from independent sources provides some plausibility to this assumption (but see Aruoba et al ()).…”