Statistics Econometrics Statistics Econometrics Statistics Econometrics Statistics for

INDEX STATISTICS 1 2 Analysis on Economically Active Population Designing Unemployment Rate Analysis in

STATISTICS 1. Analysis on Economically Active Population U/E Rate Population above Age 15 Non-Economically

STATISTICS 2. Designing Unemployment Rate Analysis in 1997 Sampling Framework (E. A. P =

STATISTICS 2. Designing Unemployment Rate Analysis in 1997 Sample Size Sample size of E.

3. Sampling Method STATISTICS Multi-Stage Cluster Sampling Deciding Sampling Framework Drawing out Sampling Cluster

4. Weighting on Samples STATISTICS Weighting Ø Why Weighting does matter? Adjust the revealed

4. Standard Error STATISTICS In E. A. P survey, Applying Multi-Stage Cluster Sampling method

4. Standard Error STATISTICS Half-Sample Method Ø How to know whether One Estimate Has

4. Standard Error STATISTICS Random Sample and Cluster Sample Ø Why We do not

6. Bias STATISTICS Bias Sampling Bias Non-response Bias As Sample survey does not cover

Slides: 16

Download presentation

Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics Statistics for Economist Ch. 17 Unemployment Rate Analysis 1. 2. 3. 4. 5. 6. Analysis on Economically Active Population Designing Unemployment Rate Analysis in 1997 Sampling Method Weighting on Samples Standard Error Bias Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics

INDEX STATISTICS 1 2 Analysis on Economically Active Population Designing Unemployment Rate Analysis in 1997 3 Sampling Method 4 Weighting on Samples 5 Standard Error 6 Bias Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 2

STATISTICS 1. Analysis on Economically Active Population U/E Rate Population above Age 15 Non-Economically Active Population Korea National Statistics Office Measuring Unemployment Rate using the Data from Census on Economically Active Population Labor force participation rate - Ratio of Economically Active Population to Population above age 15 Employed Unemployment Rate -Ratio of Unemployed Population to Economically Active Population Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 3

STATISTICS 2. Designing Unemployment Rate Analysis in 1997 Sampling Framework (E. A. P = Economically Active Population) ü Sampling Framework of E. A. P. Survey Ø 10% Sampling Cluster - Sampling Data Based on National Census (Sampling Cluster must contain at least 10% of All Administrative districts) - The Households included in Sampling Cluster represent All of Ordinary Households in The Country. A Population in use We use 10% Sampling Cluster-drawn out form all administrative districts- as Sampling Framework in E. A. P Survey Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 5

STATISTICS 2. Designing Unemployment Rate Analysis in 1997 Sample Size Sample size of E. A. P Survey -Designated number of sample in each cluster at survey in 1997 Ø Decide Sample Size Small Large Unreliable Budget Wasting In the level of satisfying Target Precision, Each Local office decide sample size considering workforce and budget Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 6

3. Sampling Method STATISTICS Multi-Stage Cluster Sampling Deciding Sampling Framework Drawing out Sampling Cluster After Dividing whole Country in accordance with administrative district, We Confirm the Sampling Framework consisting of 10% sampling cluster of 22, 029, and then make groups according to the guidelines Dividing Survey Area Diving Survey Area in The Household number Basis (Adjacent 8 Households) Drawing out Survey Area Cluster sampling 3 surveying areas-24 Households- from each Sampling Cluster. After Drawing out 1 Survey area in Random, Survey Northern and Clockwise also Drawing out Random Quota Samples as desire Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 8

4. Weighting on Samples STATISTICS Weighting Ø Why Weighting does matter? Adjust the revealed unbalance in the sampling process trough weighting - Because There’s no 1 on 1 relation between Samples in District and Households in District, A household in certain administrative district would have different probability to be included in the Sample - Other Factors like Gender and Age would have effect on the probability that Someone to be included the Sample Ø How do we give the weight on Samples? A Weighting process should make Each Sub-sample would represent the Component Ratio of the Population in Adjusted Sample Each Individual included in sample should have different weight according to residence, gender, age… etc. Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 10

4. Standard Error STATISTICS In E. A. P survey, Applying Multi-Stage Cluster Sampling method S. E. Cluster Sample If One survey area are included, The probability adjacent areas are chosen would get larger Information Random Sample The event one survey area get chosen and the event adjacent area get chosen are statistical independent As The Cluster sample has less information compared to The Random sample, The Formula used in calculating Standard Error of Random Sample does not hold. Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 12

4. Standard Error STATISTICS Half-Sample Method Ø How to know whether One Estimate Has real confidence or not? Doing identical survey independently again, then Comparing the results-this method would indicate the confidence of first survey, But there is a Budget concentration. Ø How to get same effect without doing repetition? Instead of repetition, We use half-sample. Dividing given sample to two part, then get the estimates from the each other (Half-Sample Method) Ø Ex. -Estimating the number of Unemployed in Korea Dividing Sample households in one survey area in half Estimate from the each half sample -1. 04 mil. , 1. 08 mil One Estimate=1. 06 mil(average) Difference to each estimate: 20, 000 (Standard Deviation) Unemployment in Korea = 1. 06 mil 20, 000 Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 13

4. Standard Error STATISTICS Random Sample and Cluster Sample Ø Why We do not use a Random Sample in real survey? 1. Incomplete information of residence. 2. Money. 3. Because We must do face to face interview, Using Cluster sample makes us spend less money. 4. (Money does matter in real world even in statistics. ) Ø Calculating Standard Error in Cluster Sample If you want to calculate the S. E. , You must know sampling method first (There needs more information related to sample). In Cluster sample, Characteristic of sample makes difference in S. E. The Half-Sample Method in cluster sample makes easier this complicated problem. Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 14

6. Bias STATISTICS Bias Sampling Bias Non-response Bias As Sample survey does not cover whole Population, There occurs Bias if a Chosen Sample not represent the Population. One Household in sample represent 430 households in population, so nonresponse of one household has great impact on survey result. In the Census-Surveying whole Population, the Omitted part is so small to consider-there’s no bias. In survey, A Omitted household is recorded to have identical contents that is from a adjacent household, This is irrational. Obscure in Classification Used Criteria like Having a Job, ability to work, Searching for job…etc. are arbitrary. There is difficulty to distinguish Employed/Unemployed. A Bias is a worse problem than a Standard Error. In Biased Sample, We can notice the existence of a bias after examining the data. Statistics & Econometrics Statistics & Econometrics Statistics & Econometrics 16