Hi
I am looking to use the synth_runner package to construct a synthetic control sample for a data set which includes treated firms which are acquired in a given year (treated firms: acq=1), and untreated firms, which are not acquired (untreated firms: acq=0).
So the treatment period differs for each treated firm (ie the year of acquisition differs for each firm), but there are overlaps in the main data set.
Below is a dataex sample of the data:
Please note however, that there are only 3 treated firms in this small sample, and around 17 untreated firms. The actual data set has several hundred treated firms, and several thousand untreated firms.
Post is a dummy variable which takes the value 1 after a treated firm is acquired, and acqpost is an interaction between the acq dummy and the post dummy, so acqpost is equal to 1 for treated firms in the treatment period, and 0 otherwise.
Below is the code I have came up with on the basis of Galiani & Quirstoff (2017) (https://journals.sagepub.com/doi/pdf...7X1801700404):
However, when I run this, I am getting an 'unexpected end of file' error message. I am not sure as to why this is.
roa is the depvar, the predictor variables are: sic2, age, size, and debt. I tried to use the values of the predictor variables over the 3 years pre-intervention period. (a reminder, the intervention (ie the acquisition year) is different for different treated firms)
Moreover, I would like to ask whether it is possible to run a difference-in-differences estimation on the treated vs synthetic control samples once the synthetic sample has been identified? (I realise I am not at this stage yet though)
Any advice on how to progress with these issues would be greatly appreciated,
Paul
I am looking to use the synth_runner package to construct a synthetic control sample for a data set which includes treated firms which are acquired in a given year (treated firms: acq=1), and untreated firms, which are not acquired (untreated firms: acq=0).
So the treatment period differs for each treated firm (ie the year of acquisition differs for each firm), but there are overlaps in the main data set.
Below is a dataex sample of the data:
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input byte id int year byte acq int acqyear byte(post acqpost sic2) int age double(size roa debt) 1 2003 1 2006 0 0 23 17 75901 .1853731 .4104162 1 2004 1 2006 0 0 23 18 76838 .0156303 .5749499 1 2005 1 2006 0 0 23 19 80182 .1124442 .5577187 1 2006 1 2006 1 1 23 20 75771 .2295601 .2819284 1 2007 1 2006 1 1 23 21 62282 .1646222 .2521114 1 2008 1 2006 1 1 23 22 77092 .0258652 .6543221 1 2009 1 2006 1 1 23 23 51717 .1884487 .2377168 1 2010 1 2006 1 1 23 24 90815 .0996201 .6718273 2 2003 0 . . 0 24 22 127058 .0454438 .3140141 2 2004 0 . . 0 24 23 135544 .0202517 .4588326 2 2005 0 . . 0 24 24 99803 .0145787 .8432813 2 2006 0 . . 0 24 25 17924 .1045526 .3339656 2 2007 0 . . 0 24 26 19024 .108442 .2640349 2 2008 0 . . 0 24 27 96392 .0564155 .905459 2 2009 0 . . 0 24 28 140235 .0652405 .3132313 2 2010 0 . . 0 24 29 17925 .101 .35222 3 2003 0 . . 0 37 34 8314.93 .0331483 .4884822 3 2004 0 . . 0 37 35 8076.576 .0407139 .6114611 3 2005 0 . . 0 37 36 8565.934 .0684565 .5356479 3 2006 0 . . 0 37 37 6239.243 .0453068 .4800868 3 2007 0 . . 0 37 38 9442.293 .1006037 .6218933 3 2008 0 . . 0 37 39 10543.107 .0725632 .6204969 3 2009 0 . . 0 37 40 9087.624 .0923104 .6302685 3 2010 0 . . 0 37 41 14039.596 .1068429 .6269325 4 2003 1 2008 0 0 44 39 11980 .0269616 .7964942 4 2004 1 2008 0 0 44 40 13108 .0762893 .2413793 4 2005 1 2008 0 0 44 41 30827 .1728355 .4808772 4 2006 1 2008 0 0 44 42 11422 .0667134 .486167 4 2007 1 2008 0 0 44 43 13725 .1268488 .5987614 4 2008 1 2008 1 1 44 44 4657.897 .1532091 .421163 4 2009 1 2008 1 1 44 45 11058 .1110508 .3015012 4 2010 1 2008 1 1 44 46 5538.305 .1192598 .2606857 5 2003 0 . . 0 32 8 23171 .2823357 .2261879 5 2004 0 . . 0 32 9 10504 .1088157 . 5 2005 0 . . 0 32 10 50732 .2026532 . 5 2006 0 . . 0 32 11 15433 .3511955 .2621007 5 2007 0 . . 0 32 12 25202 .1146338 . 5 2008 0 . . 0 32 13 67735 .136414 .5211043 5 2009 0 . . 0 32 14 79909 .0524221 .6706128 5 2010 0 . . 0 32 15 98871 .0669256 .4794227 6 2003 1 2007 0 0 19 22 11716 .2779959 .3218675 6 2004 1 2007 0 0 19 23 46432 .1514473 . 6 2005 1 2007 0 0 19 24 86755 .0842833 .209498 6 2006 1 2007 0 0 19 25 13089 .1424097 .6418366 6 2007 1 2007 1 1 19 26 79649 .1240694 . 6 2008 1 2007 1 1 19 27 24922 .1705722 .1119092 6 2009 1 2007 1 1 19 28 11322 .1696697 .7583466 6 2010 1 2007 1 1 19 29 69842 .19053 . 7 2003 0 . . 0 12 17 57704 .2110772 .5504991 7 2004 0 . . 0 12 18 88898 .2035029 .6685865 7 2005 0 . . 0 12 19 95042 .2448076 .6113613 7 2006 0 . . 0 12 20 51367 .216131 .5614694 7 2007 0 . . 0 12 21 54134 .1617283 .696734 7 2008 0 . . 0 12 22 103750 .2724434 .6119133 7 2009 0 . . 0 12 23 51879 .1106228 .6339174 7 2010 0 . . 0 12 24 45079 .1120921 . 8 2003 0 . . 0 55 109 24090 .0913657 .3089249 8 2004 0 . . 0 55 110 53545 .0415165 .3730133 8 2005 0 . . 0 55 111 24700 .0281781 .2973279 8 2006 0 . . 0 55 112 65800 .0342249 .4561094 8 2007 0 . . 0 55 113 42668 .0413893 .2652808 8 2008 0 . . 0 55 114 45398 .1245429 .4090268 8 2009 0 . . 0 55 115 23286 .0005153 .2845487 8 2010 0 . . 0 55 116 64219 . .5183201 9 2003 0 . . 0 11 77 47774.543 .0375544 .4031637 9 2004 0 . . 0 11 78 42515.795 .0395456 .4626542 9 2005 0 . . 0 11 79 52995.306 .0309842 .4276892 9 2006 0 . . 0 11 80 34509.518 .0238699 .5587063 9 2007 0 . . 0 11 81 29541.557 .0451881 .4102418 9 2008 0 . . 0 11 82 30434 .0003943 .5799763 9 2009 0 . . 0 11 83 29854.4 .0162004 .412549 9 2010 0 . . 0 11 84 28404 .034643 .5154908 10 2003 0 . . 0 33 11 8469 .0852521 .3521077 10 2004 0 . . 0 33 12 11676 .0753683 .2419493 10 2005 0 . . 0 33 13 10861 .2940797 .2638799 10 2006 0 . . 0 33 14 10413 .042543 .2503601 10 2007 0 . . 0 33 15 10946 .0414763 .265211 10 2008 0 . . 0 33 16 11944 .0620395 .3029136 10 2009 0 . . 0 33 17 11137 .227889 .2555446 10 2010 0 . . 0 33 18 10750 .231814 .2872558 11 2003 0 . . 0 26 59 23597.765 .2280142 .36312 11 2004 0 . . 0 26 60 14919.5 .0415092 .4155757 11 2005 0 . . 0 26 61 36228 .0109308 .7736833 11 2006 0 . . 0 26 62 15040.704 .1399153 .4204167 11 2007 0 . . 0 26 63 35803.97 .1712919 .19542 11 2008 0 . . 0 26 64 35015.4 .0017664 .7570628 11 2009 0 . . 0 26 65 15428.225 .1544364 .3453543 11 2010 0 . . 0 26 66 14681.708 .0133162 .5495813 12 2003 0 . . 0 14 55 19218 .0493808 .1590176 12 2004 0 . . 0 14 56 19382 .0001032 .6363636 12 2005 0 . . 0 14 57 24843 .0049108 .4855694 12 2006 0 . . 0 14 58 22438 .0362777 .485783 12 2007 0 . . 0 14 59 20163 .0614988 .136934 12 2008 0 . . 0 14 60 16359 .0355767 .5700226 12 2009 0 . . 0 14 61 22202 .0966129 .6657959 12 2010 0 . . 0 14 62 17653 . .8522064 13 2003 0 . . 0 31 44 3527.23 .0373352 .8529379 13 2004 0 . . 0 31 45 2255.582 .0567157 .6030652 13 2005 0 . . 0 31 46 3884.253 .0296165 .8362076 13 2006 0 . . 0 31 47 6509.631 .1099389 .6345384 end
Please note however, that there are only 3 treated firms in this small sample, and around 17 untreated firms. The actual data set has several hundred treated firms, and several thousand untreated firms.
Post is a dummy variable which takes the value 1 after a treated firm is acquired, and acqpost is an interaction between the acq dummy and the post dummy, so acqpost is equal to 1 for treated firms in the treatment period, and 0 otherwise.
Below is the code I have came up with on the basis of Galiani & Quirstoff (2017) (https://journals.sagepub.com/doi/pdf...7X1801700404):
Code:
tsset id year program my_pred, rclass 1. args tyear 2. local sic2v "sic2(`=`tyear ́-3 ́(1)`=`tyear ́-1 ́)" 3. return local predictors "`sic2 ́ age ́ size ́ debt(`=`tyear ́-3 ́(1)`=`tyear ́-1 ́)" 4. end program my_xperiod, rclass 1. args tyear 2. return local xperiod "`=`tyear ́-3 ́(1)`=`tyear ́-1 ́" 3. end . program my_mspeperiod, rclass 1. args tyear 2. return local mspeperiod "`=`tyear ́-3 ́(1)`=`tyear ́-1 ́" 3. end synth_runner roa sic2 size age debt, d(acqpost) pred_prog(my_pred) trends xperiod_prog(my_xperiod) mspeperiod_prog(my_mspeperiod) effect_graphs
However, when I run this, I am getting an 'unexpected end of file' error message. I am not sure as to why this is.
roa is the depvar, the predictor variables are: sic2, age, size, and debt. I tried to use the values of the predictor variables over the 3 years pre-intervention period. (a reminder, the intervention (ie the acquisition year) is different for different treated firms)
Moreover, I would like to ask whether it is possible to run a difference-in-differences estimation on the treated vs synthetic control samples once the synthetic sample has been identified? (I realise I am not at this stage yet though)
Any advice on how to progress with these issues would be greatly appreciated,
Paul
