Asked 19th Jan, 2020

How to construct employment indicators using NSSO dataset (India)?

I wish to calculate the weighted average of the following employment related indicators in NSSO dataset at district level:
  1. Labour force participation rate (LFPR)
  2. Worker Population Ratio (WPR)
  3. Proportion unemployed (PU)
  4. Unemployment rate (UR) Can someone please guide me on the correct approach of constructing these indicators (I am not able to understand how to use weights)?
(1)Labour force participation rate (LFPR)
{(no of employed persons +no of unemployed persons)/total population} *100
(2)Worker Population Ratio (WPR)
(no of employed persons/total population) *100
(3)Proportion unemployed (PU)
(no of unemployed persons/total population) *100
(4)Unemployment rate (UR)
{no of unemployed persons/ (no of employed persons +no of unemployed persons)}*100

All Answers (3)

Goran Miladinov
Independent Researcher
You should first select a method that supports weight. Most commonly used method is weighted LS-Least Squares or TSLS. The point of the weighted Least Squares is to divide the regression model with a proportional factor (Z). In case that some of your indicators (variables) is a proportional factor than there is no need to add Z. The most popular measure for correction of heteroskedasticity is heteroskedasticity-consistent coefficient covariance. With using of the weighted LS (WLS), each of your error will be standardized according to the assumption of the OLS. Weighted LS (WLS) can be performed using inverse of the standard deviation of residuals as well. In the link below you can find an example how to calculate weighted average:
James R Knaub
Retired US Fed Govt/Home Research
It does not look like the question has anything to do with weighted least squares (WLS) regression - my favorite topic - but rather a desire to combine employment indicators to have one, say, 'super-indicator.' But the purpose is not clear. What should it measure?
I don't think this will work out. Consider that the first indicator is the sum of the next two. I see no meaningful way of taking a weighted average here, which was the question.
Subash S.P
Center for Development Research (ZEF)
Please see codes (written in stata) below. I have written the codes for rural region (so only considered visit1).
The variables name would change, but i hope it would give some direction
** Do file for LFPR, WPR, UR
svyset status_code_ps [pweight= mult]
gen employment_status_ps =0
replace employment_status_ps =1 if status_code_ps<=51
gen employment_status_ss =0
replace employment_status_ss =1 if status_code_ss<=51
gen employment_status = 0
replace employment_status =1 if employment_status_ss ==1 | employment_status_ps ==1
svy: proportion employment_status
///employment status is WPR
svy: proportion employment_status, over( sector)
gen unemployment_ps =0
replace unemployment_ps =1 if status_code_ps==81
gen unemployment_ss =0
replace unemployment_ss =1 if status_code_ss==81
gen unemployment =0
replace unemployment =1 if unemployment_ps ==1 | unemployment_ss ==1
*** LFPR calculation
gen lfpr = 0
replace lfpr =1 if employment_status ==1 | unemployment ==1
svy: proportion lfpr, over( sector)
*** unemployment rate (this is not matching) need to check it out
svy: ratio unemployment/lfpr, over( sector)

Similar questions and discussions

Related Publications

Full-text available
Background Many studies have investigated how unemployment influences health, less attention has been paid to the reverse causal direction; how health may influence the risk of becoming unemployed. We prospectively investigated a range of health measures and subsequent risk of unemployment during 14 years of follow-up. Methods Self-reported he...
The present paper aims to analyse the effects of employment protection on a heterogeneous labour force. To this end, a standard matching model with employment protection has been extended to allow for the existence of two types of workers differing from each other only in the probability of becoming less productive while holding a job. In working o...
Got a technical question?
Get high-quality answers from experts.