Asreg, Difference between Xtset by monthly date versus fiscal year.

[Julien Maas]

Dear Stata Community,

Consider the following data example,

* Example generated by -dataex-. For more info, type help dataex
clear
input long permno int fyear float(date yearly_return pastreturn MVlog logBM)
10001 2003 526   -.04000033            .  2.7470114     .9929781
10001 2004 538      .753453   -.04000033   2.851246     .9102439
10001 2006 562     .2048728            .   3.275811     .9014823
10001 2012 640    .05517977            .  4.3578887     .9948049
10001 2013 652   -.07614732    .05517977   4.429978    1.0337856
10001 2014 664    -.2750734   -.07614732  4.7499437     .9616078
10001 2015 676     .7123322    -.2750734  4.3600655     1.045629
10005 1988 352   -.40000015            .  -.6552504   -.37912425
10005 1989 364     .6666673   -.40000015 -1.3483976   -.01468603
10005 1990 376           .6     .6666673 -1.3483976    .24672206
10014 1972 163    -.2777788            .   2.643654    1.1760638
10018 1994 422    -.8749999            .  1.2255386     .9414079
10019 1992 394     .4615384            .  3.3356984    1.0492061
10021 1990 376    -.6538458            . -.37469345   .005342972
10021 1991 388   .066666216    -.6538458 -.37469345   -.25921646
10021 1992 400    -.4479162   .066666216  -.8498009    -.3583653
10021 1993 412     4.433959    -.4479162  -.2987674   -1.2088495
10025 2003 530      .847825            .  4.1162286     .9519883
10026 1995 433     .0833334            .  4.6749806     .9765223
10026 1996 445    .09615196     .0833334   4.543845    1.0061294
10028 1990 376     8.733331            . -.14050418   -2.2353818
10028 2015 676    1.9943388            .  1.4006114     .9663729
10029 1989 361     .4848481            .   2.577182    1.0336713
10032 2004 541     1.457466            .   6.166995     .9505379
10035 1995 433    -.2394371            .    5.04009     .9747478
10035 1996 445    .03703811    -.2394371  4.6861973    1.0374911
10035 1997 457    -.4107135    .03703811   4.623489     1.032027
10035 1998 469     .8181817    -.4107135   4.427463    1.0657483
10036 1988 346    -.4285715            .   .1982359     2.102216
10036 1989 358         -.25    -.4285715 -1.1960907 -.0008356783
10036 1990 370      .666667         -.25 -1.1960907      .383638
10039 1989 364    -.2499995            .  .10989607     1.742068
10042 1997 460    -.7941174            .    3.89323    1.0294901
10042 1998 472     .4999995    -.7941174   2.445321     1.559421
10042 1999 484   -.12381048     .4999995  2.6762586     1.427618
10042 2000 496    .26087007   -.12381048  2.5256486    1.3998827
10042 2001 508   -.19999857    .26087007   2.821153    1.2433417
10043 1989 367    .08749946            .   3.172727    1.0132569
10043 1991 391     .3255827            .   3.116843    1.0842364
10043 1992 403  -.008772658     .3255827  3.1422324    1.0917183
10043 1993 415    -.0442472  -.008772658   3.564421     .9929457
10043 1994 427    .05555877    -.0442472  3.3890405    1.0668535
10043 1995 439     .3684202    .05555877   3.432576     1.069369
10043 1996 451    .53846186     .3684202   3.487925    1.0575354
10046 1988 352    -.4714283            .  1.9505422    1.1698655
10046 1989 364   -.16216275    -.4714283  1.9505422     1.145754
10046 1990 376   -.22580606   -.16216275  .56424785    3.7372894
10046 1991 388           .5   -.22580606   .7465695     2.736219
10046 1992 400   -.05555509           .5  1.0944054     1.726776
10046 1993 412    -.6470591   -.05555509  1.1520345      1.60054
10047 1992 398   -.59183615            .  2.0512831    1.8745857
10047 1993 410    -.7000002   -.59183615  3.3681834    1.0656427
10051 1994 424     .6190451            .   3.213783     1.049671
10051 1995 436    .58823633     .6190451  3.1267715    1.1012415
10051 2000 496    11.723228            .  3.2116246    1.5691178
10051 2001 508    -.2764922    11.723228  4.7392464    1.0510893
10051 2004 544    .14897396            .   5.165761     .9725549
10051 2005 556     .7764512    .14897396   4.837208    1.0558592
10051 2006 568   -.01006608     .7764512   5.120608    1.0002797
10055 1992 404    -.8846153            .   2.694196    1.1413442
10055 1993 416    -.8333333    -.8846153   3.225752    1.1267066
10057 1975 193    .27027106            .   3.547568    1.2364576
10057 1976 205    .08510655    .27027106  3.6385396    1.2077203
10057 1982 277      .326668            .   4.409622    1.0874184
10057 1984 301   -.26845762            .  4.5096087    1.0498995
10057 1985 313    -.1651371   -.26845762    4.40216     1.046579
10057 1986 325   .032965865    -.1651371   4.127217    1.0790898
10057 1988 349   .024997944            .  4.0345945    1.0201061
10057 1990 373     .7906957            .   3.543868    1.1739488
10057 1991 385   -.09090906     .7906957  3.6717014    1.1141077
10057 1992 397    .19999987   -.09090906  3.7486625    1.0962434
10057 1993 409      .488097    .19999987  4.0904117    1.0057707
10058 1981 265   -.18749987            . -.15840973    -4.021886
10058 1982 277 -5.32967e-07   -.18749987  -.5709295   -1.1205348
10058 1983 289   -.15384577 -5.32967e-07 -.57892597     .3156217
10064 1990 376    1.4236153            .  4.2473445    1.0768205
10064 1993 412    -.3571429            .   5.671184     .9919688
10064 1994 424 9.187055e-07    -.3571429   5.274479     1.078993
10064 1995 436     1.000002 9.187055e-07   5.054471    1.1041697
10077 1988 353    -.4067796            .  3.4168246    1.1909196
10077 1989 365    -.4285713    -.4067796  3.0346165    1.3315755
10077 1990 377     -.533333    -.4285713   2.845374    1.3919848
10077 1991 389   -.14285642     -.533333   1.746762    1.8406335
10077 1992 401    -.9583333   -.14285642  1.8073868     1.691531
10079 1988 357    .09677254            .  1.8399146    1.0226563
10080 1988 352     .4000008            .   2.246491    1.1656886
10080 1989 364   -.04761967     .4000008  2.3365517    1.0980461
10080 1992 400   -.05555581            .  2.4722645     1.059117
10080 1993 412      .235294   -.05555581  2.3092132    1.1556388
10080 1994 424   -.26190433      .235294  2.3511372    1.1723671
10080 1995 436   -.22580636   -.26190433  2.0602746     1.359023
10080 1996 448   -.17708294   -.22580636  1.9464457    1.4038653
10080 1997 460   -.13923998   -.17708294   1.482882    1.7383646
10080 1998 472   -.20588204   -.13923998  1.4874398    1.7412082
10080 1999 484    -.4074079   -.20588204  1.1105407    2.3601463
10080 2002 520    .13513619            .  1.5631007    1.6491514
10082 1987 340    1.1999996            .    .368083    2.7044754
10083 1988 346    .06250124            .   3.478737    1.0332339
10084 1988 349      .260869            .  1.9139034    1.1617838
10084 1990 373   -.04761912            .   .4965999    3.3872955
end
format %tm date

Where yearly return refers to returns calculated from month t until t+11
And where past return refers to returns calculated from month t-12 until t-1

and MVlog and logBM are computed at t-5.

What is the difference between;

xtset permno fyear

xtset permno date

when running;

asreg yearly_return pastreturn MVlog logBM, newey(16) fmb window(date -120 -1)


Thank you in advance for your time and consideration,

[George Ford]

using date in asreg would suggest xtset permno date

otherwise you're window is going to be in years, which is not the objective.

[Julien Maas]

Dear Mr Ford,

Thank you for your response! The reason I was asking was because xtset permno date, for some reason gives me a pretty high R-Square while the coefficients are mostly insignificant. So I was wondering where I made a mistake.

In Addition;

asreg yearly_return pastreturn MVlog logBM, newey(16) fmb window(date -120 -1)

Gives me the same R square and coefficients for the summary slopes independent of what window I use,

I am sure I am just overlooking something or misunderstanding the procedure,

If you have any suggestions or ideas what may be the case, I'll be glad to recieve them.

Thank you for your time and consideration,

Kind regards, Julien.

[George Ford]

is all the data annual?

g

[Julien Maas]

Dear mr Ford,

Thank you for your swift response,

Yes, it should all be annual data, I calculated the ratio's by each fiscal year. And the returns as twelve month periods starting from the fifth month after the fiscal year end. I only kept the fifth month after each fiscal year before collapsing the data.

So that for example for a fiscal year of January to December 2010. I would have ratio's computed over this period. And then a return period starting from the begining of May 2011 until and including April 2012. And a previous return period of May 2010 until April 2011.

The only observation that would be kept before collapsing would be that of May 2011 yet with the twelve month return and ratio's.

[Julien Maas]

Also I just recieved a message from the creator of asreg in this forum, https://www.statalist.org/forums/for...ressions/page3

that the window option is not meant to be used with the fmb option, which explains why using different windows does not influence the results displayed when using the fmb option.

[George Ford]

If your data is annual, then you need to xtset by year. The date does not appear to matter if its all may dates. I doubt you have 120 annual observations though.

[Julien Maas]

I may have been somewhat ambigious and confused,

To be safe, since different permno's have different fiscal year end dates I have multiple months for a given year, yet I have only kept one month for each permno fyear. And the numbers are all computed on a twelve month (or fiscal year) basis.

Also I read that;

On page 30, Lewellen states, "The regressions are estimated monthly and the t-statistics incorporate a Newey-West correction with ten lags for 6 month returns and sixteen lags for 12 month returns to account for the overlap in successive monthly regressions"

So I believe here he does do the regression for each month, while some of his signals like Return on assets, Asset growth and Accruals are calculated in the prior fiscal year and are annual numbers.

He also states "accounting data are assumed to be known four months after the end of the fiscal year (thus, sales, earnings, etc. are assumed to be observable by the end of April for a firm whose fiscal year ends in the prior December").

So I am still trying to understand in this case, what is the best approach.

[George Ford]

Are you trying to replicate his method, or are you using his results?

[Julien Maas]

I am trying to replicate his method, yet apply it to a different sample and with a different set of signals (based on other papers).

[George Ford]

then you need monthly data?

[Julien Maas]

Dear professor Ford, Thank you for response! I am now using monthly data again, without collapsing and it works fine. I also found out the issue was due to a mistake I made that caused data to overlap, I have now fixed this issue.