Fixed effect and reverse causality

[Martin Jensen]

Hi Statalist,

i am analyzing a dataset of performance for Footballplayers and How a transfer affects Their performance.

To do This i have gathered an overall performance-target 83 players spread on 6 seasons.
Besides that i have a dummy variable which is 1 if the player is sold at the end of the season.

my data would look like This:

season - player - performance - sold
11/12 - Ronaldo - 7,2 - 0
12/13. - Ronaldo - 7,1 - 0
13/14. - Ronaldo - 8,1 - 1 (sold in summer 2014 AFTER the season)
14/15. - Ronaldo - 6,9 - 0
15/16. - Ronaldo - 7,0 - 0
16/17. - Ronaldo - 7,2 - 0

So i want to measure How performance develops leading up to a transfer and afterwards.

to do This i have made a fixed effect model with performance as My Y
and sold as My X. My model now shows me that when sold=1 the performance for that season is higher than normal which is also the effect i would expect.

but now i am Unsure if My coefficients would biased by reverse causality and How i should address This issue?

/Martin

[Martin Jensen]

Bump

[Carlo Lazzaro]

Martin:
despite being simphatetic with your striving to get a positive reply, bumping is discouraged on this list (please, see the FAQ) You have already received replies at https://www.statalist.org/forums/for...r-probit-model, with the advice of taking the reverse causation issue up to your teacher/professor/supervisor.
If you do not provide some update, I'm afraid your query has reached a standstill.

[Martin Jensen]

Hey Carlo,

i have apologize for bumping. The Thread you are linking to is about the probit model and there for another issue.

Here i am asking for ideas to imitigate the reverse causality problem in my fixed effect model with the binary variable AS the only regressor.

[Carlo Lazzaro]

Martin:
please post what you typed and what Stata gave you back (setting the endogeneity issue aside for a while). Thanks.

[Martin Jensen]

Dear Carlo,

Below I have typed dataex:

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte SeasonID str22 PlayerID str9 Year byte Sold double Performance int SoldID float clusterID
1 "Luis Suárez"      "2011/2012" 0               7.35  . 54
2 "Luis Suárez"      "2012/2013" 0   7.75048780487805  . 54
3 "Luis Suárez"      "2013/2014" 1   8.36142857142857  1 54
4 "Luis Suárez"      "2014/2015" 0               7.75  . 54
5 "Luis Suárez"      "2015/2016" 0  7.944772727272727  . 54
6 "Luis Suárez"      "2016/2017" 0  7.651395348837209  . 54
1 "James Rodríguez"  "2011/2012" 0                  7  . 41
2 "James Rodríguez"  "2012/2013" 1  7.169999999999999  2 41
3 "James Rodríguez"  "2013/2014" 1  7.556470588235294  3 41
4 "James Rodríguez"  "2014/2015" 0  7.667560975609755  . 41
5 "James Rodríguez"  "2015/2016" 0  7.285500000000001  . 41
6 "James Rodríguez"  "2016/2017" 0  7.161764705882352  . 41
1 "Ángel Di María"  "2011/2012" 0  7.366666666666666  . 83
2 "Ángel Di María"  "2012/2013" 0  6.978064516129032  . 83
3 "Ángel Di María"  "2013/2014" 1  7.543170731707317  4 83
4 "Ángel Di María"  "2014/2015" 1  7.207692307692308  5 83
5 "Ángel Di María"  "2015/2016" 0  7.855142857142857  . 83
6 "Ángel Di María"  "2016/2017" 0  7.319999999999999  . 83
1 "David Luiz"        "2011/2012" 0  7.216551724137932  . 21
2 "David Luiz"        "2012/2013" 0  7.166170212765956  . 21
3 "David Luiz"        "2013/2014" 1  7.127241379310345  6 21
4 "David Luiz"        "2014/2015" 0  7.187777777777778  . 21
5 "David Luiz"        "2015/2016" 0 7.2043333333333335  . 21
6 "David Luiz"        "2016/2017" 1 7.1416666666666675  7 21
1 "Alexis Sánchez"   "2011/2012" 0               7.31  .  6
2 "Alexis Sánchez"   "2012/2013" 0  6.953043478260869  .  6
3 "Alexis Sánchez"   "2013/2014" 1  7.505833333333334  8  6
4 "Alexis Sánchez"   "2014/2015" 0  7.746170212765957  .  6
5 "Alexis Sánchez"   "2015/2016" 0  7.807999999999999  .  6
6 "Alexis Sánchez"   "2016/2017" 1  7.756363636363637  9  6
1 "Diego Costa"       "2011/2012" 0  7.179999999999999  . 23
2 "Diego Costa"       "2012/2013" 0  6.943793103448276  . 23
3 "Diego Costa"       "2013/2014" 1  7.385333333333334 10 23
4 "Diego Costa"       "2014/2015" 0   7.31448275862069  . 23
5 "Diego Costa"       "2015/2016" 0  7.147941176470589  . 23
6 "Diego Costa"       "2016/2017" 1  7.410000000000001 11 23
2 "Luke Shaw"         "2012/2013" 0               6.96  . 55
3 "Luke Shaw"         "2013/2014" 1  7.157500000000001 12 55
4 "Luke Shaw"         "2014/2015" 0               6.79  . 55
5 "Luke Shaw"         "2015/2016" 0  7.218333333333334  . 55
6 "Luke Shaw"         "2016/2017" 0 6.9338461538461535  . 55
1 "Ander Herrera"     "2011/2012" 0               7.05  .  8
2 "Ander Herrera"     "2012/2013" 0 7.1406451612903235  .  8
3 "Ander Herrera"     "2013/2014" 1                7.4 13  8
4 "Ander Herrera"     "2014/2015" 0  7.330000000000001  .  8
5 "Ander Herrera"     "2015/2016" 0  6.772857142857141  .  8
6 "Ander Herrera"     "2016/2017" 0 7.3325000000000005  .  8
2 "Romelu Lukaku"     "2012/2013" 0               6.95  . 67
3 "Romelu Lukaku"     "2013/2014" 1   7.26735294117647 14 67
4 "Romelu Lukaku"     "2014/2015" 0  6.944102564102564  . 67
5 "Romelu Lukaku"     "2015/2016" 0               7.16  . 67
6 "Romelu Lukaku"     "2016/2017" 1                7.5 15 67
1 "Cesc Fàbgregas"   "2011/2012" 0  7.383636363636364  . 16
2 "Cesc Fàbgregas"   "2012/2013" 0  7.316315789473684  . 16
3 "Cesc Fàbgregas"   "2013/2014" 1              7.198 16 16
4 "Cesc Fàbgregas"   "2014/2015" 0  7.637560975609755  . 16
5 "Cesc Fàbgregas"   "2015/2016" 0  7.145249999999999  . 16
6 "Cesc Fàbgregas"   "2016/2017" 0  7.049999999999999  . 16
2 "Adam Lallana"      "2012/2013" 0               6.96  .  1
3 "Adam Lallana"      "2013/2014" 1  7.211578947368421 17  1
4 "Adam Lallana"      "2014/2015" 0  6.877307692307692  .  1
5 "Adam Lallana"      "2015/2016" 0  7.067058823529412  .  1
6 "Adam Lallana"      "2016/2017" 0                7.1  .  1
2 "Eliaquim Mangala"  "2012/2013" 0               6.81  . 26
3 "Eliaquim Mangala"  "2013/2014" 1               7.03 18 26
4 "Eliaquim Mangala"  "2014/2015" 0  7.056666666666667  . 26
5 "Eliaquim Mangala"  "2015/2016" 0 6.8482142857142865  . 26
6 "Eliaquim Mangala"  "2016/2017" 0               6.86  . 26
1 "Antoine Griezmann" "2011/2012" 0               7.21  . 10
2 "Antoine Griezmann" "2012/2013" 0               7.12  . 10
3 "Antoine Griezmann" "2013/2014" 1             7.1255 19 10
4 "Antoine Griezmann" "2014/2015" 0   7.11764705882353  . 10
5 "Antoine Griezmann" "2015/2016" 0  7.551272727272727  . 10
6 "Antoine Griezmann" "2016/2017" 0  7.364999999999999  . 10
1 "Medhi Benatia"     "2011/2012" 0               7.36  . 62
2 "Medhi Benatia"     "2012/2013" 1  7.409999999999999 20 62
3 "Medhi Benatia"     "2013/2014" 1               7.59 21 62
4 "Medhi Benatia"     "2014/2015" 0             7.1815  . 62
5 "Medhi Benatia"     "2015/2016" 0  7.013076923076922  . 62
6 "Medhi Benatia"     "2016/2017" 1            7.07625 22 62
1 "Dejan Lovren"      "2011/2012" 0  7.039999999999999  . 22
2 "Dejan Lovren"      "2012/2013" 1 6.8047826086956515 23 22
3 "Dejan Lovren"      "2013/2014" 1  7.261764705882353 24 22
4 "Dejan Lovren"      "2014/2015" 0  6.987037037037037  . 22
5 "Dejan Lovren"      "2015/2016" 0           6.998125  . 22
6 "Dejan Lovren"      "2016/2017" 0  7.130000000000001  . 22
1 "Toni Kroos"        "2011/2012" 0  7.604750000000001  . 78
2 "Toni Kroos"        "2012/2013" 0   7.60741935483871  . 78
3 "Toni Kroos"        "2013/2014" 1  7.432142857142858 25 78
4 "Toni Kroos"        "2014/2015" 0  7.396976744186047  . 78
5 "Toni Kroos"        "2015/2016" 0   7.33767441860465  . 78
6 "Toni Kroos"        "2016/2017" 0 7.5515384615384615  . 78
2 "Lazar Markovic"    "2012/2013" 1  6.650000000000001 26 50
3 "Lazar Markovic"    "2013/2014" 1  6.431428571428571 27 50
4 "Lazar Markovic"    "2014/2015" 0  6.594615384615385  . 50
5 "Lazar Markovic"    "2015/2016" 0  6.853333333333333  . 50
6 "Lazar Markovic"    "2016/2017" 0            6.66625  . 50
3 "Juan Iturbe"       "2013/2014" 1               7.31 28 46
4 "Juan Iturbe"       "2014/2015" 0  6.674761904761904  . 46
5 "Juan Iturbe"       "2015/2016" 0              6.362  . 46
end

And when I run the model I get the following.:

The goal is to see how Performance changes in the seasons where the players get sold.