Development of Standard Imputation Procedure in Processing Establishment Surveys Using the Line Relative Method
The statistics produced by the current system of reporting establishment surveys are essential not only to the needs of national income accounts, but also to labor administration. In particular, these data are of significant input in the area of labor standard enforcement and employment facilitation. As such, the Department of Labor and Employment (DOLE) prioritized the improvement of timeliness, accuracy and reliability of these statistics on employment, wages and hours of work generated by the Employment, Hours and Earnings Survey (EHES) and Occupational Wages Survey (OWS). This study then aims to develop standard imputation and other techniques in processing establishment survey to minimize estimation errors and generate reliable and comparable data on employment and wages from major establishment surveys. The first part of this study employed the use of nine imputation procedures of the US Bureau of Labor Statistics to develop alternative imputation procedures for employment hours of work and wages of data: (1) ES – 202 Method of Imputation (ES); (2) Mean Imputation Method (MN); (3) Hot Deck Imputation Method – Random Selection (HD1); (4) Hot Deck Imputation Method – Nearest Neighbor (HD2); (5) Model 1: Yt,i = a+bt; (6) Model 2: Yt, I = a +bYt-1,I; (7) Model 3: Yt,i = a + bXt-1,I; (8) Model 4: ln(Yt,i) = a + bln(Yt-1,i) and; (9) Model 5: ln(Yt,i) = a + bn(Xt-1,i). For missing employment and hours of work data, Models 2 and 3 appear to be most acceptable as the imputed values are close to the reported past quarter data of the sample establishment. While for missing basic wage data, Models 2 and 5 obtained imputed values which are very closed to the reported data in the previous quarters.