Classifying panel data by income each year of the period or just the last year of the period?

[Emmy Pham]

Hello everyone,

I have panel data from 52 countries and 19 periods (1996 to 2014). I want to classify the dataset by income using GNI per capita based on WB guidelines ($1,045 - $4,095 - $12,695 -$12,696).
I don't know how data have been categorized in most of the papers, that the category should be done with the last year 2014 or each year during the period?

From what I have seen, if we do this for each year of the period, we might have some problems, two of them are:
- As the thresholds are based on current US dollars in 2023, they can't be used to classify the income level of a country in 1996 or so.
- This way of classification will result in sub-datasets with some countries having less than 19 periods.
For example, country A is classified as low from 1996 to 2000 and high from 2001 to 2014, so when I run commands for the high-income group, country A has only 14 years while other countries have, let's say 18 or 19 years.

I would be grateful for every help.
Thank you!

[Carlo Lazzaro]

Emmy:
I am not clear with what you're after.
What gou you mean by the phrasing " I want to classify the dataset by..."?
Do you refer to -xtset- or else?

[Emmy Pham]

Emmy:
I am not clear with what you're after.
What gou you mean by the phrasing " I want to classify the dataset by..."?
Do you refer to -xtset- or else?

Oh sorry let me clarify my question.
I have already categorized my panel data by low, lower middle, higher middle and high income using the command:

xtset Country Year
egen category = cut(gni), at(0 1035 4045 12535 999999) icodes
label define Category 0 "Low-income" 1 "Lower middle-income" 2 "Upper middle-income" 3 "High-income"
label values category Category

My panel data (52 countries, 19 years) now have 4 income groups. When I run the regression for "Low-income" only, the number of years of some countries are less than 19.
For example, in the "Low-income" group, the data for country A after the category has 3 years spanning from 1996-1998 because from 1999, A is classified as "high income".
While other countries have 19 years.


Does Stata treat A as "missing data" from 1999 to 2014? Does this affect the result of the regression?

And if someone has experience analyzing data with different income groups, please share with me how you did it, just use the thresholds for the last year (2014) or for each year of the period?

Thank you!

[Carlo Lazzaro]

Emmy:
try:

list Country Year if Country==A

and see what Stata gives you back.

[Emmy Pham]

Thank you very much!