not sorted error in spite of xtset

[Guest]

Dear Stata experts,

I have data on bilateral trade for different sectors. Here is the sample of my data:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(id year HS2) long(reporternr partnernr) double mean_tariff float ln_m_tariff
44685 1992 86 19 1     .          .
17480 2005  9  8 1  3.07  1.1216775
32398 2014 34 14 1  1.91   .6471033
12518 2015 82  6 1   1.9   .6418539
10069 1991 83  5 1 18.56   2.921009
 7476 1990 37  4 1     .          .
32410 1994 46 14 1     .          .
34819 1993  5 15 1  9.75  2.2772672
12507 2011 70  6 1  1.77  .57097954
42242 2016 89 18 1  1.18  .16551444
17480 2009  9  8 1  3.67  1.3001916
15067 1999 83  7 1   6.1  1.8082888
 5024 2007 35  3 1  3.45  1.2383742
37283 1990 23 16 1     .          .
10038 2017 51  5 1   .89 -.11653382
 7467 2014 28  4 1  1.91   .6471033
 7522 2014 84  4 1  1.91   .6471033
22443 2001 72 10 1   4.8   1.568616
17480 2013  9  8 1  1.84   .6097656
12486 2008 49  6 1  3.15  1.1474024
 7459 2017 20  4 1   .89 -.11653382
37313 2009 53 16 1  3.67  1.3001916
29823 2003  7 13 1  3.74  1.3190856
 5047 2012 58  3 1  1.81  .59332687
15061 1994 76  7 1     .          .
17562 1998 92  8 1  6.49  1.8702625
22379 2002  8 10 1  3.69  1.3056265
 5052 2003 63  3 1  3.74  1.3190856
22397 1991 26 10 1 18.56   2.921009
17481 1991 10  8 1 18.56   2.921009
57124 2010 30 24 1   2.8  1.0296195
32414 2013 50 14 1  1.84   .6097656
42228 2007 74 18 1  3.45  1.2383742
52241 2001 42 22 1   4.8   1.568616
19974 2005 53  9 1  3.07  1.1216775
54743 2015 95 23 1   1.9   .6418539
37316 2005 56 16 1  3.07  1.1216775
44649 2003 49 19 1  3.74  1.3190856
52227 2009 28 22 1  3.67  1.3001916
34863 2010 49 15 1   2.8  1.0296195
29866 2011 50 13 1  1.77  .57097954
49673 2000  8 21 1  5.83   1.763017
57109 2006 15 24 1  3.38  1.2178757
22387 1999 16 10 1   6.1  1.8082888
22413 1994 42 10 1     .          .
29822 2018  6 13 1   .86  -.1508229
17559 2016 89  8 1  1.18  .16551444
34903 2009 90 15 1  3.67  1.3001916
17481 1995 10  8 1     .          .
22390 2003 19 10 1  3.74  1.3190856
end
label values reporternr reporternr
label def reporternr 3 "BEL", modify
label def reporternr 4 "CAN", modify
label def reporternr 5 "CHE", modify
label def reporternr 6 "CHN", modify
label def reporternr 7 "CZE", modify
label def reporternr 8 "DEU", modify
label def reporternr 9 "DNK", modify
label def reporternr 10 "ESP", modify
label def reporternr 13 "GBR", modify
label def reporternr 14 "ITA", modify
label def reporternr 15 "JPN", modify
label def reporternr 16 "KOR", modify
label def reporternr 18 "NOR", modify
label def reporternr 19 "NZL", modify
label def reporternr 21 "RUS", modify
label def reporternr 22 "SWE", modify
label def reporternr 23 "TUR", modify
label def reporternr 24 "USA", modify
label values partnernr partnernr
label def partnernr 1 "AUS", modify

I wanted to have year-on-year logarithmic difference of importer's tariff. In the beginning I used the code:

xtset id year
gen D_ln_m_tariff=ln_m_tariff-L.ln_m_tariff

I noticed it's not correct though. It should be for every importer (here partnernr). In this case it's just taking the lagged values within the whole panel. I used the code:

xtset id year
bysort partnernr: gen D_ln_m_tariff=ln_m_tariff-L.ln_m_tariff

I saw an error: not sorted

I sorted partnernr and year. But wasn't successful.

I generated id by grouping exporter, importer and sector.

[Guest]

I've just changed the dataset with tariffs, year and partner only which I previously merged with my master data. However, I would be happy to know more elegant solution

[Carlo Lazzaro]

Guest:
as -AUS- is coded 1 in the whole dataset, my guess is that Stata finds nothing to sort by.
Conversely, the following works:

. bysort id (year): gen D_ln_m_tariff=ln_m_tariff-L.ln_m_tariff
(50 missing values generated)

[Nick Cox]

xtset id year

implies

sort id year

and is in general inconsistent with any sort based on other variables. So, if you change the sort order thereafter anything dependent on the xtset definition becomes impossible. That includes use of time series operators.

@Carlo Lazzaro's solution

bysort id (year): gen D_ln_m_tariff=ln_m_tariff-L.ln_m_tariff

is entirely consistent with

xtset id year

but could just be

gen D_ln_m_tariff=ln_m_tariff-L.ln_m_tariff

That may help to explain the first error in #1. I doubt it explains how to do what you want, which I don't understand.

[Guest]

Thank you Carlo and Nick for your answers. I mean I wanted to have log-on-year difference on tariff. Here is the dataset on tariffs only:

* Example generated by -dataex-. For more info, type help dataex
clear
input str3 partner int year double mean_tariff
"AUS" 1990     .
"AUS" 1991 18.56
"AUS" 1992     .
"AUS" 1993  9.75
"AUS" 1994     .
"AUS" 1995     .
"AUS" 1996  6.81
"AUS" 1997  6.82
"AUS" 1998  6.49
"AUS" 1999   6.1
"AUS" 2000  5.83
"AUS" 2001   4.8
"AUS" 2002  3.69
"AUS" 2003  3.74
"AUS" 2004  3.69
"AUS" 2005  3.07
"AUS" 2006  3.38
"AUS" 2007  3.45
"AUS" 2008  3.15
"AUS" 2009  3.67
"AUS" 2010   2.8
"AUS" 2011  1.77
"AUS" 2012  1.81
"AUS" 2013  1.84
"AUS" 2014  1.91
"AUS" 2015   1.9
"AUS" 2016  1.18
"AUS" 2017   .89
"AUS" 2018   .86
"AUT" 1990   4.8
"AUT" 1991   4.5
"AUT" 1992  4.18
"AUT" 1993  3.79
"AUT" 1994  4.16
"AUT" 1995  6.28
"AUT" 1996   4.4
"AUT" 1997  3.66
"AUT" 1998  3.47
"AUT" 1999  3.48
"AUT" 2000  2.41
"AUT" 2001  3.37
"AUT" 2002  2.45
"AUT" 2003  2.52
"AUT" 2004  2.24
"AUT" 2005  2.46
"AUT" 2006   2.3
"AUT" 2007  2.24
"AUT" 2008  1.72
"AUT" 2009  1.85
"AUT" 2010  2.13
"AUT" 2011  1.59
"AUT" 2012  1.46
"AUT" 2013  1.49
"AUT" 2014  2.03
"AUT" 2015  2.02
"AUT" 2016  2.09
"AUT" 2017  1.96
"AUT" 2018  1.83
"BEL" 1990   4.8
"BEL" 1991   4.5
"BEL" 1992  4.18
"BEL" 1993  3.79
"BEL" 1994  4.16
"BEL" 1995  6.28
"BEL" 1996   4.4
"BEL" 1997  3.66
"BEL" 1998  3.47
"BEL" 1999  3.48
"BEL" 2000  2.41
"BEL" 2001  3.37
"BEL" 2002  2.45
"BEL" 2003  2.52
"BEL" 2004  2.24
"BEL" 2005  2.46
"BEL" 2006   2.3
"BEL" 2007  2.24
"BEL" 2008  1.72
"BEL" 2009  1.85
"BEL" 2010  2.13
"BEL" 2011  1.59
"BEL" 2012  1.46
"BEL" 2013  1.49
"BEL" 2014  2.03
"BEL" 2015  2.02
"BEL" 2016  2.09
"BEL" 2017  1.96
"BEL" 2018  1.83
"CAN" 1990     .
"CAN" 1991     .
"CAN" 1992     .
"CAN" 1993  3.59
"CAN" 1994     .
"CAN" 1995  6.69
"CAN" 1996  5.43
"CAN" 1997  1.99
"CAN" 1998  1.44
"CAN" 1999  1.01
"CAN" 2000  1.31
"CAN" 2001   1.6
"CAN" 2002  1.49
end

after merging it with my master data I realized that I wanted log-on-year differences on tariffs. The problem is that it would take differences e.g. between Austria and Belgium. I wanted to have year differences only within the same countries. I decided to compute it in the using data set. I used a code:

gen ln_m_tariff=ln(mean_tariff)
encode partner, gen(partnernr)
xtset partnernr year
gen D_l_tar=ln_m_tariff-L.ln_m_tariff

and then merged with my master data. I was wondering if it was possible to do it when datasets have been merged. If this question sounds annoying- pardon me. I'm just doing my best to work efficiently in Stata

[George Ford]

once you xtset, it won't lag outside of the country.

egen cid = group(partner)
order cid, after(partner)
g ltariff = ln(mean_tariff)
xtset cid year
g dltariff = ltariff - l.ltariff

[Guest]

Thank you George! I wouldn't come up with the idea to group partner

[George Ford]

xtset no-like the strings