Abstract

Identified as a priority research concern was the improvement of the timeliness, accuracy and reliability of statistics on employment, wages and hours of work generated from the Employment, Hours and Earnings Survey (EHES) and Occupational Wages Survey (OWS) conducted by the Bureau of Labor and Employment Statistics (BLES). One major area of study was the effect of imputation of missing entry in an individual establishment data series. Other areas for consideration were estimation problems such as; (a) non-response or inadequate reporting in a cell; (b) changes in the industry/size stratification of sample establishments; (c) atypical data where data for an establishment showed a trend markedly different from the trends of the data for other reports in a cell; and (d) sample break which was caused by a change in the composition of the matched sample from that of the previous period (in the case of OWS which adopted the link-relative method of estimation). The absence of uniform or standardized guidelines in the treatment of these problems made comparability and consistency difficult among various establishment surveys. This research project was intended to investigate the problems attendant to the processing of major statistical survey series with particular emphasis on employment, hours of work and wages generated from establishment surveys. Research outputs were: (1) inventory of the current data imputations used in processing establishment surveys relative to the employment and wages in the Philippines and selected countries; (2) identification of imputation methods which may be useful in processing EHES and OWS data and applied to a subset of returns of these surveys; (3) development of a computer-based system to automatically validate and impute missing data and the system’s accompanying data processing and manual of operating procedures; and (4) identification of reporting and processing problems in the EHES and OWS relative to missing data.