Abstract

The project paper is titled as “An Assessment of the Quality of Statistics Generated by the Bureau of Labor and Employment Statistics (BLES)..” The objectives of this study are: (1) to review the current state of labor and employment statistics with emphasis on those being generated by the BLES, (2) to identify gaps and make recommendations towards the improvement of such statistics, and (3) to enhance the technical capability of the BLES personnel on survey sampling theory and methods trough an active participatory approach to the review of the process. An assessment was done on the quality of five surveys conducted by the BLES, namely: (1) Occupational Wages Survey (OWS), (2) Occupational Injuries Survey (OIS), (3) Survey of Specific Group of Workers (SSGW), (4) Industrial Relations in the Workplace Survey (IRWS) and (5) Employees’ Hours and Earnings Survey (EHES). Because of high non-response rates, there was difficulty in computing variance estimates using traditional formulas for stratified sampling (as the surveys were designed). Aside from coefficient of variation (CV), sensitivity analysis performed consisted of measures of relative bias, relative variance, and relative mean squared error. One of the most evident problems that may have affected the quality of data and statistics generated by the BLES surveys was the “large” incidence of missing data or nonresponse (71% for OWS, 1994; 58% for OIS, 1994; 49.7% for SSGW, 1995; and 61.6% for IRWS, 1996). There were various levels of sampling errors as measured by the CVs, ranging from 1% to 32% for selected parameters. Results of the sensitivity analyses showed that for all cases considered, the bias in the estimates cannot be ignored. Thus, there is need to pursue actions to the identified causes of non-response in order to control its incidence.