Problem with Propensity Score Matching

[Sebastian Grein]

Hi,

I want to conduct Propensity Score Matching to create a control group. I want to use the propensity score matching for creating a sample (observations and control group) for further calculations as done in Bena/Li (2014).
Therefore, I'd like to ask you for advise on which of the different PSM commands (psmatch2, teffects psmatch, kmatch) to use.
Additional information are:
- I want to create 2 control groups: 1. Size (SIZE) and industry (SIC) 2. On Size (SIZE), industry (SIC) and book-to-market (BTM)
- My treatment variable is deal (deal) which is 1 if the observation describes a deal and 0 otherwise
- Year (Year) and industry (SIC (2-digit SIC code)) should be exact matches
- I need 5 control firms for each firms in the initial sample (initial = only the deals (deal =1))
- Preferably, I'd want to use the no replace option

This is an example of my data:

* Example generated by -dataex-. For more info, type help dataex
clear
input float deal str6(a_cusip_six Target6digitCUSIP) byte SIC_two float(Year BTM_lag1 BV_ln_lag1 ind_year total_acc_freq_5)
1 "000307" "7C7919" 80 2015 .22084355  4.983278  988 2274.3284
1 "000307" "5C7040" 80 2015 .22084355  4.983278  988 2274.3284
1 "000307" "7C1485" 80 2015 .22084355  4.983278  988 2274.3284
0 "00032Q" ""       28 2022  .3129321  5.062025 1358 281.83923
0 "000360" ""       35 2000  .6573898   4.07169   82  1447.444
0 "000360" ""       35 2001  .7536119  4.341439  142 1442.1274
0 "000360" ""       35 2002  .3597853 4.3346076  202 1433.0425
0 "000360" ""       35 2003  .3818808 4.5186644  261  1711.893
0 "000360" ""       35 2004  .4200801  4.625806  319 1985.6843
0 "000360" ""       35 2005 .53020567   4.65612  377  1690.809
0 "000360" ""       35 2006  .5193562  4.732736  435 1391.1727
0 "000360" ""       35 2007  .4010745  4.867965  494  1926.067
0 "000360" ""       35 2008  .3832543  4.921002  552 1907.8893
0 "000360" ""       35 2009  .3916884  4.946936  610  3227.913
0 "000360" ""       35 2010  .4655783  5.051208  668  4255.303
0 "000360" ""       35 2011  .3442122  5.076903  726 4750.8145
0 "000360" ""       35 2012  .3548234   5.18728  784  6022.014
0 "000360" ""       35 2013  .3781444  5.265241  842  6746.429
0 "000360" ""       35 2014  .1836823  5.372701  900  5929.078
0 "000360" ""       35 2015  .1926586   5.45154  958 4858.6914
0 "000360" ""       35 2016 .18916784  5.450412 1016  5600.611
0 "000360" ""       35 2017 .14742124  5.547246 1074  4321.739
0 "000360" ""       35 2018 .15425764  5.692991 1132  5071.652
0 "000360" ""       35 2022 .15583254  6.477249 1365  6219.215
0 "000361" ""       50 2001  1.987912  6.607998  156  320.2961
0 "000361" ""       50 2002 1.8610992  6.553725  216  399.1839
0 "000361" ""       50 2003 1.9479238  6.565545  275 557.50214
0 "000361" ""       50 2004  4.790507  6.531783  333   873.725
0 "000361" ""       50 2005 2.2961338  6.564267  391  713.7247
0 "000361" ""       50 2006  1.400916  6.596095  449  790.8918
0 "000361" ""       50 2007 1.1089821  6.886347  508 1023.9554
1 "000361" "00320W" 50 2008  .8706896  6.973199  566  941.3951
0 "000361" ""       50 2012 1.6228745  7.440574  798 1474.8085
0 "000361" ""       50 2013 4.5244174  7.694235  856   1628.44
0 "000361" ""       50 2014  2.704927  7.667111  914 1629.7904
1 "000361" "9E4796" 50 2015 2.2880285  7.695985  972 1944.0752
0 "000361" ""       50 2019   .982753  7.329553 1204  1736.794
0 "000361" ""       50 2020 1.4494097  7.324622 1262  1642.347
0 "000361" ""       50 2021 2.9368286  7.639642 1320 1812.9552
0 "000361" ""       50 2022 1.0425171  7.339343 1379  2066.439
0 "000361" ""       50 2023  .9222679  7.361312 1428         .
0 "000752" ""       34 2000 1.2341483  6.117703   81  482.5416
0 "00081T" ""       27 2006  1.490755  7.565016  427  790.8918
0 "00081T" ""       27 2007 1.2994742  7.522725  486 1023.9554
0 "00081T" ""       27 2008 2.1877666   7.54882  544  941.3951
0 "00081T" ""       27 2009  6.839888  7.156332  602   937.736
0 "00081T" ""       27 2010  2.785915  7.009228  660  934.5398
0 "00081T" ""       27 2011  2.456704  7.047169  718 1087.9885
0 "00081T" ""       27 2012 2.0859506  7.018133  776 1474.8085
0 "00081T" ""       27 2013  3.019617  7.827121  834   1628.44
0 "00081T" ""       27 2014 3.1197054  7.776073  892 1629.7904
0 "00081T" ""       27 2015 2.2080333  7.708141  950 1944.0752
0 "00081T" ""       27 2016 2.5934224  7.577327 1008 2027.8108
0 "00081T" ""       27 2017   1.46607  7.632643 1066 1643.2063
1 "00081T" "4H9024" 27 2018 2.1505983  7.937053 1124 1886.0978
0 "00081T" ""       27 2022  3.905836  8.036347 1357  2066.439
0 "00086T" ""       59 2000 2.1053352  4.506642  105  1067.537
0 "00086T" ""       59 2001 1.7689255 4.6764855  165  941.9887
0 "00086T" ""       59 2002  .5621461  4.818756  225  935.3834
0 "00086T" ""       59 2003  .8227532  5.281466  284  930.0809
0 "00086T" ""       59 2004  .6217065  5.445849  342  810.6956
0 "00086T" ""       59 2005  .5370531  5.717396  400  577.9028
0 "00086T" ""       59 2006  1.084691  5.745427  458  577.2553
0 "00086T" ""       59 2007  .7547431  5.798599  517 576.17725
0 "00086T" ""       59 2008 1.1532676   5.77421  575  688.7772
0 "00086T" ""       59 2009 10.251554  5.682715  633  799.4882
0 "00086T" ""       59 2010  3.633331  5.581498  691  568.2173
0 "00087B" ""       50 2001 2.8610985  3.803145  156  515.9566
0 "00087B" ""       50 2002  9.084736  3.999704  216 507.13855
0 "00087B" ""       50 2003  11.52586  4.036539  275  544.6887
0 "00087B" ""       50 2004  13.02537 3.9651465  333  401.8929
0 "00087B" ""       50 2005  3.791897    4.0729  391  307.3124
0 "00087B" ""       50 2006 1.5377173  4.215145  449 299.66293
0 "00087B" ""       50 2007  1.810381  4.331207  508 335.68335
1 "000886" "11138X" 36 2000  .2335486  7.422092   83  810.7519
1 "000886" "45052K" 36 2000  .2335486  7.422092   83  810.7519
1 "000886" "02152K" 36 2000  .2335486  7.422092   83  810.7519
1 "000886" "695934" 36 2000  .2335486  7.422092   83  810.7519
1 "000886" "152317" 36 2000  .2335486  7.422092   83  810.7519
1 "000886" "19974H" 36 2001  .2411455  8.286647  143  802.7621
0 "000886" ""       36 2005  .7976782    7.2641  378  976.6392
0 "000886" ""       36 2006  .7550697  7.336286  436  968.3251
1 "000886" "50175R" 36 2007  .9608069  7.384859  495 1152.2501
1 "000886" "15618J" 36 2007  .9608069  7.384859  495 1152.2501
0 "00088E" ""       73 2000  .0818547  3.337618  108  421.0132
1 "00088E" "45723V" 73 2001 1.0846641  5.464476  168 419.22485
0 "00088U" ""       73 2000 .12377501 2.0458848  108  972.3721
1 "00088U" "46010P" 73 2001   .689868 1.9194195  168  954.9915
1 "00088U" "65423N" 73 2002  .4505208 2.2885876  228  940.2632
1 "00088U" "68286F" 73 2006  .3305296  3.085573  461 1310.6427
0 "000899" ""       28 2015 .26232374  3.304209  951 1127.3561
0 "000899" ""       28 2016 .27430806  3.166108 1009 1239.6344
0 "000899" ""       28 2017  .3589648  3.164842 1067 1125.0939
0 "000899" ""       28 2018   .742564 4.6823072 1125  787.2484
0 "000899" ""       28 2019   .802248  4.487242 1183 1237.1703
0 "000899" ""       28 2020  .5356342  4.844903 1241 1186.4073
0 "000899" ""       28 2021 1.0152136  5.335965 1299 1294.7832
0 "000899" ""       28 2022 1.0005624  5.621317 1358 1511.6017
0 "00089C" ""       38 2001 .55929387  4.889371  145 1250.4623
0 "00089C" ""       38 2002  .6786637  4.988437  205  934.9863
end

I tried the following command, but it leads to significant differences between the control group and treated group in the matched sample:

ssc install psmatch2, replace
logit deal i.SIC_two i.Year BTM_lag1 BV_ln_lag1
predict double pscore if e(sample)
gen double pscore2 = ind_year*100+pscore
psmatch2 deal (pscore2), out (total_acc_freq_5) neighbor (5) caliper(0.05) common //noreplace

psmatch2 deal (pscore2), out (total_acc_freq_5) neighbor (5) common

pstest BTM_lag1 BV_ln_lag1, both

Additionally, I want to run a regression on the matched sample with deal as the dependent variable (y) and total_acc_freq_5 as the independent variable (x1) + control variables (x2, x3...). But using the following command does not work. I know that the command should be structured like this: reg y x1 x2 t [fweight=_weight], but my treatment variable is also my dependent variable.

regress deal total_acc_freq_5 BV_ln BTM LEV ROA REV R_D [fweight=_weight]

So, questions are:
1) Does using psmatch2 make sense?
2) What are steps to get a matched sample where treated and untreated groups do not differ significantly?
3) How do I process a regression with the results of the PSM including the weights?

Best,
Sebastian