9: Multi-valued Variables in the Register system

Transcription

1 Course on Register-based Statistics INEGI Aguascalientes April 2011 Anders & Britt Wallgren Statistics Sweden and Örebro University 9: Multi-valued Variables in the Register system 1. Multi-valued variables 2. Estimation with combination objects 1

2 Chart 9.1 Number of employed and wage sums in different registers Register 1 Persons Register 2 Job activities Person Sex Wage sum 1 st Industry Job Person Local unit Wage sum Industry PIN1 M D J1 PIN1 LU A PIN2 F D J2 PIN3 LU A PIN3 M A J3 PIN1 LU D J4 PIN2 LU D Aggregation J5 PIN3 LU D Aggregation Register 3 Local units Local unit Industry Wage sum Nr empl LU1 A LU2 D This illustrates a general problem: When we integrate different registers and create new variables by aggregating qualitative variables, then inconsistencies can arise. 2

3 Four surveys with the same variable, the inconsistencies are clear! National Accounts get data from many surveys with different variables, inconsistencies are there, but where? Chart 4. Number of Employees November 2004 according to four different surveys Labour Force Survey Employees November 2004, thousands Business Register Employment Register LFS Economic activity Enterprises Establishments Persons main job's first NACE Error NACE 2002 (1) (2) (1)-(2) (3) (2)-(3) (4) margin (3)-(4) A Agriculture and forestry B Fishing C Mining and quarrying D Manufacturing E Electricity, gas and water F Construction G Wholesale and retail trade H Hotels and restaurants I Transport, communication J Financial intermediation K Real estate, business activities L Government M Education N Health and social work O Other service activities Unknown activity Total

4 One enterprise can have one or more Establishments (1) 1 th Industry for Enterprises (N= ) (2) 1 th Industry for Establishments (N= ) Chart 4. Number of Employees November 2004 according to four different surveys (1) 1 th Industry Enterprises Employees November 2004, thousands Business Register Employment Register LFS Economic activity Enterprises Establishments Persons main job's first NACE Error NACE 2002 (1) (2) (1)-(2) (3) (2)-(3) (4) margin (3)-(4) A Agriculture and forestry (2) 1 th Industry B Fishing Establishments C Mining and quarrying D Manufacturing E Electricity, gas and water F Construction (3) 1 th Industry G Wholesale and retail trade H Hotels for and restaurants the I Transport, Establish- communication J Financial ment intermediation of the K Real estate, business activities th Job of an L Government M Education employee N Health and social work O Other (4) service Same activities as (3) Unknown activity Total

5 This is a serious problem for economic statistics in all countries! At a NSI there e are three categories: 1. Subject matter specialists: Content, definitions 1 st NACE 2. Methodologists: Math Probability Inference Sampling, One survey at a time -paradigm 3. IT-specialists: Systems, multi-valued variables The problem is unnecessary! 5

6 The problem is unnecessary! The problem arise when (p148): Objects occur several times, generates multi-valued variables (Calendar year registers) Many to one relations generates multi-valued variables Multi-valued variables themselves (NACE, Occ., Education ), e.g. a gas station (45201, 45320, 47112, 47300, 56100, 77110) Combinations of these (NACE, Job -> Person) 6

7 Calendar year register for the population of persons during 2005 From date To date Weight = Time at the Person Address Municipality yyyymmdd yyyymmdd address, years PIN1 Address /365 = 0.37 PIN1 Address /365 = 0.63 PIN2 Address /365 = 1.00 PIN3 Address /365 = 0.20 PIN3 Address /365 = 0.53 PIN3 Address /365 = 0.27 PIN4 Address /365 = 0.43 PIN4 Address /365 = 0.57 Average population 2005 Municipality Absolute frequency Relative frequency = % = % Total %

8 Many to one relations generates multi-valued variables One person can have many jobs Number of employed and wage sums in different registers Person Local unit Wage sum Industry Weight PIN1 LU A 22/45 = 0.49 PIN1 LU D 23/45 = 0.51 PIN2 LU D 21/21 = 1.00 PIN3 LU A 18/27 = 0.67 PIN3 LU D 9/27 = 0.33 Employed by Industry Employed by Industry Industry Number of employed A = 1.16 D = 1.84 Total

9 Register on persons with occupational data Person Occupation Extent Weight alt 1 Weight alt 2 Weight alt 3 PIN1 Statistician PIN1 Farmer PIN1 Politician PIN2 Hospital orderly PIN2 Cleaner PIN3 Shop assistant PIN4 Shop assistant PIN5 Shop assistant PIN6 Shop assistant Total Persons and full-time employed by occupation, by occupation, three three alternatives alternatives ti Alternative 11 Alternative 2 2 Alternative 33 Occupation Nr. Per cent Nr. Per cent Nr. Per cent Politician Statistician Hospital orderly Shop assistant Farmer Cleaner Total Page 153 9

10 Estimation of turnover within one industry using two methods Page 164 Year Industry X Ltd X Ltd X Ltd s Weight X Ltd s contri- Industry R X Ltd s Industry R Quarter R excl. before after traditional X Ltd purchase purchase estimation share with bution to contribution within traditional Industry R, to R with Industry estimation traditional combination R estimation objects (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) with combination objects Blue Red 10

11 Turnover in an industry, two estimates SEK billions per quarter Traditional estimation, col (8) Estimation w ith combination objects, col (10) 0 Year 1 Year 2 Year 3 Year 4 Year 5 11

12 - These problems inconsistencies between surveys and level shifts in time series can be solved by using weights - In the Swedish Business Register there is information on weights for different industries for each establishment - In a similar way that weights are used to give estimates t in sample surveys - This is illustrated by the following example: 12

13 1. Local unit Register 2. Local unit * Industry Loacl unit 1 st NACE N employees Local unit NACE weight LU1 DJ 218 LU1 DJ 1 LU2 DH 293 LU2 DH 0.51 LU3 DJ 156 LU2 DJ 0.49 Sum: 667 LU3 DJ 0.4 LU3 DH 0.3 LU3 DK Match 1. and 2. by Local unit Id, gives a register with combination objects Loacl unit 1 st NACE N employees NACE weight =(N of empl)*weight LU1 DJ 218 DJ LU2 DH 293 DH LU2 DH 293 DJ LU3 DJ 156 DJ LU3 DJ 156 DH LU3 DJ 156 DK Estimates from 1: Estimates from 3: DH 293 DH 196 DJ 374 DJ 424 DK 0 DK 47 Sum: 667 Sum: