Decomposing productivity disparity between and within countries (ineqdeco)

[Oskar Solbraekke]

Hi,

I possess a measure of productivity of oil and gas field, that I calculated, called "revenue_return_lifecycle". This productivity measure is listed by "asset" and "country".

My goal is to decompose the productivity differences between and within countries. To my understanding, this decomposition can be done using General Entropy measures, such as GE(0) "Theil Index" and GE(1) "The Mean Log Deviation".

For this purpose, I have installed and usedprof. Jenkins "ineqdeco package" - which calculates several inequality measures (usually applied to income inequality, while I am looking at productivity inequality)

• Here is a link to the package: https://ideas.repec.org/c/boc/bocode/s366007.html
• How to install: ssc install ineqdeco
• Help ineqdeo: help ineqdeco
  
  
  My question are simply:

1. Do I need to take any precautions by using an ado.file for a different purpose (productivity inequality) than what it was initially intended (income inequality)?
2. I need help to interpret the resulting tables. I am not sure I understand what the "within-group" index would mean in my context. By looking at the pictures below, which one of the tables tell me the productivity heterogeneity within the countries? In other words, the difference in productivity of oil and gas assets within each country? Is it the "subgroup indices" or the "within-group" index?



They look like this:

Pic1:

Pic 2:

The function I have used is:

ineqdeco revenue_return_lifecycle, by(country_integer)


Dataset structure:

My dataset has almost a million lines and 32 string and number variables, but my question can be illustrated by showing only a sample (the data is generated/fake):



dataex country country_integer asset revenue_return_lifecycle

* Example generated by -dataex-. To install: ssc install dataex
clear
input str2 country long country_integer str8 asset float revenue_return_lifecycle
"US" 4 "Torrance"    .21875
"US" 4 "Wallace"   1.481203
"US" 4 "Birch"     1.256579
"US" 4 "Ding"     1.0169492
"US" 4 "Dong"           .04
"US" 4 "Desk"      .2847222
"US" 4 "Pulm"     3.5757575
"US" 4 "Past"      21.57143
"US" 4 "Torrance" .44444445
"US" 4 "Torrance" 1.3670886
"US" 4 "Birch"     .4916667
"BR" 2 "Asset 1"  2.2368422
"BR" 2 "Asset 2"   4.288889
"BR" 2 "Asset 3"   1.774648
"BR" 2 "Asset 4"   4.409091
"BR" 2 "Hag"       1.576271
"BR" 2 "Jikas"    2.1612904
"BR" 2 "Gul"      .51968503
"BR" 2 "Blo"         1.4375
"BR" 2 "Red"       2.448276
"BR" 2 "Hul"      1.1468531
"BR" 2 "Bets"      .7944444
"BR" 2 "Bongo"     .1440678
"AZ" 1 "Hat"       .9797297
"AZ" 1 "Haol"     1.5913043
"AZ" 1 "Hoal"      .6183206
"AZ" 1 "Pink"     15.083333
"AZ" 1 "Grey"     1.3636364
"DK" 3 "Sun"       6.111111
"DK" 3 "Boat"     2.2666667
"DK" 3 "Bies"      .5925926
"DK" 3 "Skies"     1.262295
"DK" 3 "Table"    .27642277
"DK" 3 "Book"      .8571429
"DK" 3 "Spoon"     .4919786
end
label values country_integer country_integer
label def country_integer 1 "AZ", modify
label def country_integer 2 "BR", modify
label def country_integer 3 "DK", modify
label def country_integer 4 "US", modify

[Stephen Jenkins]

Oskar: please read the help files in detail, and look at some of the literature on inequality decomposition by population subgroup. Your questions would be readily answered. Also, please report Stata output using CODE fragments, not pictures.

In short, with the MLD, total inequality can be written as the sum of Within-Group inequality and Between-Group inequality. Within-group inequality is the weighted sum of inequality within each of the groups, where the weights are 'population shares' (read the help for the precise definition) and Between-Group inequality is the inequality if each unit is attributed with the mean value for its group. You can apply the indices to any metric variable; not only "income". If you want to see the MLD estimates for each and every group, they are retained as saved results. Type -return list- after using the command to see the estimates. You can refer to these saved results in subsequent calculations.