Cross-sectional regression

[Danilo Sultanazzi]

Dear all,
I'm working on the following sample.

* Example generated by -dataex-. To install: ssc install dataex
clear
input str6 company double year float earnings double decile
"001166" 2006      .1574603  6
"001166" 2007     .17622428  3
"001166" 2008      .0848003  7
"001166" 2009     .04289107 10
"001166" 2010      .3290127  6
"001166" 2011     .19211413  4
"001166" 2012     .05785522  8
"001166" 2013   -.008499536  1
"001166" 2014     .04031636  6
"001166" 2015     .05215391  5
"001166" 2016    .030050375  5
"001166" 2017     .04415479 10
"001166" 2018     .05634042 10
"001491" 2005     .04122434  4
"001491" 2006     .07826556  9
"001491" 2007     .06048017  8
"001491" 2008    .011324465  6
"001491" 2009   -.000723325  8
"001491" 2010    .013932136  6
"001491" 2011   -.013726063  8
"001491" 2012    .005304825  1
"001491" 2013    -.01955572  1
"001491" 2014   -.020802936  2
"001491" 2015    .005364127  1
"001491" 2016     .02345395  7
"001491" 2017      .0419029  7
"001491" 2018     .06463527  5
end

Within each decile, I would like to estimate annual cross-sectional regression of year-ahead earnings on current earnings. At this purpose, I would like to estimate the following regression separately for each decile each year:
earnings(t+1)=a+w*earnings(t)
How could I do that?
Thanks for the attention.

Daniel

[Andrew Musau]

Firms change decile groups from one year to the next and you want to perform regressions within these groups. That immediately rules out panel regressions. The question to you is should you ignore decile groups when taking leads of the earnings variable? If so, generate the leads first and then run the cross-sectional regressions.