Calculation of RCA (Revealed Comparative Advantage)

[Lee Eunice]

Dear Statalist,

I am trying to calculate RCA(Revealed Comparative Advantage) to use in my thesis. I am following the paper by Dalum (1999) and Laursen (2005), in order to explore more on RCA.
In the paper of Dalum (1999) and Laursen (2005), they not only calculated RCA and RSCA(Revealed Symmetric Comparative Advantage) but also the equation below, in order to examine country's specialization pattern.

I am having a difficult time in estimating the equation, which is as follows:


Which t1 is the initial year and t2 is the final year of RSCA(revealed symmetric comparative advantage) in sector i at country j.

For me, since I am having a dataset from 2005 to 2015, it would be RSCA2015 = 𝛼𝑖 + 𝛽𝑖𝑅𝑆𝐶𝐴 2005 + 𝜀𝑖𝑗. To explain a little bit more of my dataset, it is 10 ASEAN economies from 2005 to 2015 on 5 specific sectors. The variables I have are year, country, sector, exp, Xik, Xi, Xk, X, RCA, RSCA. (So, I have calculated RCA and RSCA.)
It is very difficult for me to understand the equation and how to regress the equation with data. I would appreciate anyone who knows would give me some help.

Could you give me any advice on how I should deal with this equation?
I would appreciate your advice and please excuse my posting if there are any missing points in the post.

[Tom Zylkin]

Hi Lee Eunice,
It looks like you meant to include an equation from a paper, but it did not show up. I am somewhat familiar with RCA methods but it is hard to follow your question without more details.

As an aside, you may want to check out the following paper: French, Scott. "Revealed comparative advantage: What is it good for?." Journal of International Economics 106 (2017): 83-103.

Regards,
Tom

[Lee Eunice]

Hello Professor Zylkin,

Thank you for the suggestion. I will check out the paper.

I am sorry that the equation didn't show up. This is the equation that I wanted to show:

RSCA_ij^{t2 =} a_i + b_i RSCA_ij^t1+e_ij

I tried to search everywhere to regress this equation, but I cannot figure it out.

I have already calculated the RCA and RSCA indices. However, I don't know how to approach the equation.
If there are any more details to add, please let me know and I will add it.

Regards,
Eunice

[Tom Zylkin]

Hi Eunice,

I assume the issue is not that you don't know you how to perform a regression of the form y=a+bx+e (that would just be reg y x with appropriate options for the standard errors.)

Thus I am under the impression that you are having a problem with how to regress a variable y at t=2 on that same variable at t=2. For that, I think you should reshape your data using reshape wide so that you have a separate variable for each year of RCA. But if that's not right, perhaps you could provide more details about what the issue is that you are having.

Regards,
Tom

[Lee Eunice]

Professor Zylkin,

Thank you for the reply.

Since my dataset is from 2005 to 2015, the equation should be RSCA_ij²⁰¹⁵⁼ a_i + b_i RSCA_ij²⁰⁰⁵+e_ij.
So, I tried 10-year leg, but it came out like this.

. xtset country year, delta(10)
time values with period less than delta() found
r(451);

Could you give me any advice on this?

I do not understand what you mean by

Thus I am under the impression that you are having a problem with how to regress a variable y at t=2 on that same variable at t=2.

Thank you.

Regards,
Eunice

[Tom Zylkin]

Hi Eunice,
I meant to say "I am under the impression that you are having a problem with how to regress a variable y at t=2 on that same variable at t=1." Sorry about that.

The error you are getting for delta has to do with the fact that you are telling xtset that the observations in your data are spaced out by 10 years, but actually there are observations that are less than 10 years apart. I don't think you actually need to put "delta(10)" in order to use a time series operator to refer to the 10 year lag. Assuming you have annual data, it should just be L10.

That said, my understanding is that you only have two years of data and you regressing data from the second year on data from the first. Using "reshape wide" so that you have one column for RCA at t=1 and another for RCA at t=2 makes more sense to me than using xtset based on this understanding. Others may disagree...

Regards,
Tom

[Lee Eunice]

Dear Prof. Zylkin,

I will show the dataset.

* Example generated by -dataex-. To install: ssc install dataex
clear
input int year str5 country float(exp1 RCA RSCA)
2005 "PHL"      168.4  .3120224   -.5243643
2005 "KHM"        8.7    .32541  -.50896704
2005 "IDN"     1308.9 1.1169997   .05526678
2005 "BRN"       30.9  .8391601  -.08745292
2005 "PHL"     1245.6  2.006744    .3348286
2005 "KHM"       25.1  1.584503   .22615688
2005 "IDN"     4659.3  .8761111  -.06603492
2005 "BRN"       42.6  1.005925  .002953779
2005 "IDN"      505.9  .4965255   -.3364289
2005 "THA"      105.2 .06821845   -.8722762
2005 "SGP"    54522.5  .9787697  -.01072906
2005 "MYS"        853  .6985787  -.17745504
2005 "VNM"      168.3  .4961784    -.336739
2005 "BRN"       30.8 1.4117076    .1707121
2005 "BRN"      232.5 1.2096813    .0948921
2005 "IDN"    19679.4 1.0281997  .013903795
2005 "ASEAN"  32062.4  .9712712 -.014573734
2005 "MYS"     1300.2 1.7971493   .28498632
2005 "IDN"     1098.7  1.819968   .29077205
2005 "PHL"     1053.1  3.293222    .5341494
2005 "ASEAN" 119609.8 1.0067886 .0033828234
2005 "MYS"    22423.7  .9776024 -.011325622
2005 "SGP"    16453.4 1.0630002   .03053815
2005 "KHM"      494.7  .9850138 -.007549679
2005 "PHL"     2853.9 1.0130807   .00649786
2005 "VNM"      311.8 1.0572072    .0278082
2005 "VNM"     5145.9  .9288254   -.0369005
2005 "MYS"       2561  1.823667   .29170123
2005 "ASEAN"   3544.3  .9458473 -.027829865
2005 "VNM"      144.4  .8263403   -.0950862
2005 "THA"        668  .7310891  -.15534206
2005 "SGP"     5360.4 1.8076453   .28765932
2005 "THA"      828.3  .4670286   -.3632999
2005 "THA"    30787.8 1.0628047  .030446265
2005 "PHL"     9114.6  .8990212  -.05317413
2005 "ASEAN"   6625.7 1.0476462   .02326875
2005 "THA"     8857.7  1.100446   .04782126
2005 "ASEAN"     7317 1.0059713 .0029767687
2005 "SGP"      412.7  .2348866   -.6195819
2005 "SGP"     2567.3  .7527701  -.14105093
2005 "VNM"     2118.1  1.375917    .1582197
2005 "BRN"      648.1  .9369514  -.03255043
2005 "MYS"     5521.8  .8663803  -.07159295
2005 "KHM"      158.9 1.1386682  .064838566
2005 "KHM"       27.7   .900868  -.05215091
2006 "SGP"    69907.3  .9634959 -.018591406
2006 "KHM"      704.4  .9067054  -.04892975
2006 "KHM"       17.4  .4518505   -.3775523
2006 "KHM"         47 1.1650611   .07623855
2006 "PHL"     1488.5 2.1241992   .35983595
2006 "MYS"        902 1.1983907   .09024362
2006 "ASEAN" 143824.3 1.0067208  .003349136
2006 "PHL"     1205.4   2.92013    .4898128
2006 "IDN"     3995.8  .6837119   -.1878517
2006 "VNM"       2734 1.3702326   .15620095
2006 "MYS"     1919.1 1.5019827   .20063396
2006 "SGP"     6236.6 1.7340993   .26849768
2006 "BRN"       33.5  1.262032   .11583927
2006 "THA"      201.9 .12144222   -.7834178
2006 "VNM"      180.7  .8536741   -.0789383
2006 "IDN"    21168.9 1.0444251   .02172989
2006 "ASEAN"   7420.4  1.000249 .0001244986
2006 "IDN"     1530.1  2.467899    .4232819
2006 "VNM"      207.6  .5777458  -.26763132
2006 "PHL"      345.2  .5160739   -.3191969
2006 "PHL"    13453.9  .9969895 -.001507525
2006 "IDN"      345.9  .3442931   -.4877708
2006 "BRN"       48.2 1.0696658  .033660404
2006 "BRN"      346.8 1.3860103   .16178064
2006 "THA"     9739.5 1.0070747  .003524888
2006 "VNM"      391.1 1.1402286    .0655204
2006 "THA"    35690.2 1.0641029   .03105605
2006 "SGP"      708.3  .3191341   -.5161461
2006 "ASEAN"  40058.9  .9724475 -.013968678
2006 "KHM"       43.2 1.8178546   .29024017
2006 "BRN"       30.2   .702106  -.17501496
2006 "KHM"      291.5  1.301295   .13092412
2006 "ASEAN"     7465  1.054157  .026364595
2006 "IDN"     1749.2 1.6619674     .248676
2006 "BRN"      773.9  .8918296  -.05717766
2006 "THA"      900.2  .8774093  -.06529778
2006 "VNM"     6315.7  .9126976  -.04564362
2006 "ASEAN"   4159.6  .9518247 -.024682203
2006 "SGP"    23794.3  1.137341   .06425789
2006 "THA"     1109.6  .6370964  -.22167516
2006 "PHL"       2675  .6874744  -.18520315
2006 "MYS"     7449.9 1.0500358    .0244073
2006 "MYS"      624.9  .5123569   -.3224392
2006 "SGP"     2413.9   .640694  -.21899638
2006 "MYS"    24054.8  .9776087  -.01132239
2007 "ASEAN" 178988.4 1.0049859  .002486765
2007 "VNM"     7498.5  .9419454  -.02989509
2007 "PHL"      381.7  .4298476   -.3987505
2007 "PHL"     4540.2    .88064  -.06346774
2007 "VNM"      218.8  .9804327  -.00988032
2007 "MYS"     2422.1   1.24979   .11102817
2007 "THA"      241.8 .11367525   -.7958556
2007 "IDN"      486.4  .4162372   -.4121928
2007 "VNM"      630.2 1.5149958   .20477004
2007 "KHM"      334.3 1.2502717   .11121843
end

With this dataset, how should I do it?
1. When regressing lagged, I encounter the error which I mentioned in previous post. If I use L10, this is how it shows.

 reg RSCA l10.RSCA
time variable not set
r(111);

2. For reshaping like you suggested, as I am not only having two years of data, then would it be appropriate to approach RSCA of 2005 at t=1 and RSCA of 2015 at t=2?


As I am a beginner in using Stata, I lack the knowledge of it. So, I might ask a lot of basic things but I try to learn by searching and reading.

Thank you for the advice.

Regards,
Eunice

[Tom Zylkin]

Hi Eunice,
It looks to me like your data is missing a sector code, since you have multiple observations per country-year and since you said "j" was the sector in your regression.

If you add the sector code, such that each observation is uniquely described by country-sector-year, then

reshape wide exp1 RCA RSCA, i(country sector) j(year)

should work.

Regards,
Tom

[Lee Eunice]

Professor Zylkin,

I am sorry, I missed showing the sector variable.

It worked! Thank you so much.
I have been struggling for five days, I sincerely appreciate your help.

Regards,
Eunice