Participants

105

1. Kalyan Singhal: Can you define the poverty line in simple language?

Roy Van Der Weide: It measures a basic needs basket conceptually and evaluates the price one would have to pay in order to obtain such basic necessities. The .90 poverty line we use is the global poverty line the World Bank adopts for measuring extreme poverty.

2. Kalyan Singhal: Is CPHS data publicly available without payment?

Mahesh Vyas: The data is available publicly at a price.

3. Shinya Takamatsu: In Slide 7 to measure moments, NSS-2011 have multiple years. How this is possible if this was collected in 2011?

Roy Van Der Weide: This is the NSS-2011, we have just put it on the graph so that you can compare the different moments across years. We are not measuring NSS statistics over time, the dotted line represents estimates for NSS in 2011.

4. Shamim Mondal: Not sure I understand why the variance was calculated from the NSS for various years. What does it represent, and why isn’t this a horizontal line?

Roy Van Der Weide: The reason why NSS changes over time is because the number of states covered in the CPHS sample evolves over time and becomes larger over time. So in order to facilitate an apples to apples comparison what we do in any giver year is we only use the states that are available in the CPHS dataset.

5. Shamim Mondal: So it seems that the skewness of consumption is much higher in the urban NSS data compared to CPHS. That may result in underestimates of poverty, even after adjustment, especially for urban sample?

Roy Van Der Weide: This is precisely what motivated us to expand to adopt approach 2 in our paper.

6. Areendam Chanda: I appreciate the effort to compare it to NSSO 2011. However we are talking about 5-6 years where real consumption is growing at 5 or 6% per capita. Does it make sense to compare them.

Roy Van Der Weide: That is correct. This is why we are not using the first and second moment from the NSS-2011 so that’s a good point. We estimate the first and second moment by imputing NSS-type consumption into each of the CPHS years and those vary over time because of the imputed NSS-type consumption is driven by assumptions driven by changes in employment status, changes in asset ownership, housing characteristics and so forth to the extent that there is growth and that growth is picked up by changes in education and employment housing conditions asset ownership etc.

7. Shamim Mondal: Any specific reason for choosing MOM rather than MLE, given that you are assuming normality?

Roy Van Der Weide: MLE in this case would be equivalent to using OLS or general least squares if you assume that the errors would not be normal but we’re not able to implement OLS or MLE in this case because we dont have a sample of households for which both measures of consumption are observed in CPHS and NSS so we have to resort to something like the method of moments estimation.

8. Ilke Kardes: Can you define third moment of consumption?

Roy Van Der Weide: The third moment is literally the third moment of log expenditure. It is the expected value so you would subtract the mean of log expenditure then cube it and you take the expected value of it.

9. Anmol Somanchi:Thanks for the useful presentation. Is it possible to present some results which show how well the ’max entropy’ reweighting procedure worked to get representativeness?

Sutirtha Sinha Roy: We have slides on how the re-weighting performs with respect to benchmark indices in the appendix of this presentation, this presentation will be made available on the CPHS website. We are happy to continue a dialogue after this presentation as well.

Use of CMIE Product implies acceptance of the usage agreement & privacy policy   ♦   FAQs   ♦  Diagnosis