210 likes | 378 Views
CES and LAUS: Advanced. Current Employment Statistics Local Area Unemployment Statistics. CES: CHANGES & DATA. CODE STRUCTURE, SIC TO NAICS DATA REVISIONS MSA DEFINITIONS SAMPLE AND ESTIMATION METHODS. CES DATA. INDUSTRY CODE CHANGES ARE SMOOTHED CES IS A TRUE TIME SERIES
E N D
CES and LAUS: Advanced Current Employment Statistics Local Area Unemployment Statistics
CES: CHANGES & DATA • CODE STRUCTURE, SIC TO NAICS • DATA REVISIONS • MSA DEFINITIONS • SAMPLE AND ESTIMATION METHODS
CES DATA • INDUSTRY CODE CHANGES ARE SMOOTHED • CES IS A TRUE TIME SERIES • THERE HAS BEEN A HISTORICAL RECONSTRUCTION OF NAICS DATA BACK TO 1990
CES DATA: REVISIONS • CES DATA IS REVISED MONTHLY (ONCE) AND ANNUALY (TWICE) • THE MONTHLY REVISION, BASED ON A MORE COMPLETE SAMPLE COLLECTION, IS PUBLISHED WITH THE FOLLOWING MONTHS DATA RELEASE
CES DATA: REVISIONS • IN THE ANNUAL BENCHMARK PROCESS THE PREVIOUS YEARS DATA IS REPLACED BY ADJUSTED QCEW (TAX RECORD) DATA THROUGH THE BENCHMARK MONTH AND THE BALANCE OF THE YEAR IS RE-ESTIMATED • AT THE SAME TIME, THE DATA FOR THE SECOND YEAR PREVIOUS IS COMPLETELY REPLACED BY ADJUSTED QCEW DATA
MSA CHANGES: COLORADO SPRINGS Present Additions
MSA CHANGES: DENVER PRESENT ADDITIONS
CES: SAMPLING • THE BASIC SAMPLE UNIT IS THE UEMPLOYMENT INSURANCE (UI) ACCOUNT • MULTIPLE WORKPLACE ENITIES ARE INCLUDED IN THE UI ACCOUNT • THERE ARE ADJUSTMENTS FOR BIRTH/DEATH
CES: SAMPLING • CES HAS CHANGED SAMPLING AND ESTIMATION METHODS. • OLD METHOD: QUOTA • NON-RESPONSE PROBLEMS WERE ADDRESSED BY FREE SUBSTITUTION, INTRODUCING BIAS. • BIRTH/DEATH ADJUSTMENT WAS MULTIPLICTIVE.
CES: SAMPLING • NEW METHOD: PROBABILITY • RANDOM SAMPLE SELECTION. • WEIGHTED ESTIMATORS CAN BE USED. • CONFIDENCE INTERVALS CAN BE ESTABLISHED. • IMPROVED MONITORING OF ACTUAL ECONOMIC CONDITIONS.
CES: MODEL • OLD METHOD: CURRENT LINK RELATIVE • L = aec/aep • Where: aec = all employment for current month's current matched sample • aep = total ae for previous month's current matched sample.
CES: MODEL EXAMPLE • A SAMPLE OF 3 MEMBERS WITH DIFFERENT EMPLOYMENT SIZES: • COMPANY ‘x’ HAS 10 WORKERS THIS MONTH, LAST MONTH 9. THERE ARE 10 COMPANIES THIS SIZE IN THE STATE • COMPANY ‘y’ HAS 100 NOW, 99 LAST MONTH. THERE ARE 5 THIS SIZE • COMPANY ‘z’ HAS 500 BOTH MONTHS. IT’S THE ONLY COMPANY THIS SIZE • THERE ARE 10,000 EMPLOYEES IN THE INDUSTRY STATEWIDE
CES: MODEL EXAMPLE OLD • THE CURRENT LINK WOULD BE: • L = ( x + y + z )now/(x + y + z)last • L = (10 + 100 + 500) / (9 + 99 + 500) • L = 1.0033 • LINK IS STABLIZED BY COMPANY z, THE ONLY COMPANY ITS SIZE
CES: MODEL • OLD MODEL - APPLY LINK TO GET THE EMPLOYMENT ESTIMATE: • AEc = (L)(Fc) (AEp) • Where: AEc = AE estimate for current month • AEp = AE estimate for previous month • L = current link relative • Fc = bias adjustment factor for current month
CES: MODEL EXAMPLE OLD • LET THE AEp = 10,000 • LET THE ADJUSTMENT FOR BIAS, Fc = 1.001 • THE ESTIMATE WOULD BE: • AEc = (L)(Fc)(AEp) • AEc = (1.0033)(1.001)(10,000) • AEc = 1043 • THE BIAS ADJUSTMENT INTERACTS WITH THE LINK
CES: MODEL • NEW MODEL; WEIGHTED LINK RELATIVE • AEc = AEp ((Σ(w * aec )/(Σ(w*aep)) + (BIRTH/DEATH FACTOR) • WHERE w IS THE WEIGHT OF A INDIVIDUAL SAMPLE MEMBER • aec IS THAT MEMBER’S CURRENT EMPLOYMENT • aep IS THE PREVIOUS MONTHS EMPLOYMENT • THE B/D FACTOR IS A SEPARATE BLS ESTIMATE
CES: MODEL EXAMPLE NEW • USING THE SAME COMPANIES, AND A B/D FACTOR OF 10 • THE WEIGHTS ARE DETERMINED BY THE NUMBER OF COMPANIES OF A GIVEN SIZE IN THE INDUSTRY BEING SAMPLED • COMPANY x HAS WEIGHT, w = 10 • COMPANY y HAS w = 5 • COMPANY z HAS w = 1
CES: MODEL EXAMPLE NEW • USING THE SAME DATA AS IN THE OLD EXAMPLE THE NEW ESTIMATE IS: • AEc = AEp ((Σ(w * aec )/(Σ(w*aep)) + (BIRTH/DEATH FACTOR) • AEc = 10,000 (((10 * 10) + (5 * 100) + (1 * 500))/((10 * 9) + (5 * 99) + (1 * 500))) + 10 • AEc = 10,000 * 1.0138 + 10 • AEc = 10138 + 10 = 10148
CES: MODEL EXAMPLE NEW vs OLD • THE OLD ESTIMATE WAS 10,043 • THE NEW MODEL ESTIMATE IS 10,148 • THE DIFFERENCE IS THE IMPACT OF SMALLER BUSINESS ON THE LINK • IF THE ESTIMATING CELL HAS LARGER EMPLOYMENT OR A GREATER LINK, THE OLD BIAS ADJUSTMENT FACTOR HAS A LARGER IMPACT
CES: MODEL • WHY THE CHANGES? • THE NEW SAMPLE IS MORE REPRESENATIVE AND RESPONSIVE • THE WEIGHTED SAMPLE GIVES MORE MODEL IMPACT TO SMALL BUSINESS • THE BIRTH/DEATH ADJUSTMENT IS ADDITIVE SO ITS NOT SCALED TO MONTHLY RESULT
CES DATA LIMITATIONS • Limited by area • Limited by industry detail • The results are designed to track direction as the primary item of interest • The absolute measure of level is the secondary item of interest, but this is addressed by Benchmarking the data • No demographic data is published