7,627
edits
Changes
Jump to navigation
Jump to search
Note that the new CS data goes up The matching scripts, LBOmatchingscript.jl and LBOmatching.jl, were rewritten by Ed to July 2019run with Julia 1.1.1.
The Effect of Leverage Buyouts on Innovation (view source)
Revision as of 13:09, 25 July 2019
, 13:09, 25 July 2019→Code
The COMPUSTAT query is saved on the mcnair (AmoDate) account, and the last result is saved as e28f8123921e8343.txt
The SQL code assembles data for all gvkey year combinations, so many observations have blank data. LBOValid restricts the LBOs to:
*dealnum NOT IN (281,47,191,561,476)
*CS data available one or two years before the LBO, and US HQ in CS. Note that the new CS data goes up to July 2019.
SELECT (gopvtyear-csfyear) as diff, count(*) FROM lbovalid GROUP BY (gopvtyear-csfyear);
diff count
1 422
2 56
The final data contains 478 LBOs, 180 of which have patents at some point in their lives.
The output of this SQL code is:
Masterv2-12.txt
We then run '''DataSetup.do''', which (gvkeys, LBOs, LBOs w/patents ever):
*Loads Masterv2-12.txt (84 83 vars, 1,315294,542 171 obs): 30594, 481478, 181180*Restricts to declared assets >= 40m at some point: 17788, 475472, 180179*Requires non-LBO firms to be US non-OTC exchanges: 12131 (was 12106 with NNA exchanges), 472, 179 We find that Tobin's Q, lvg and revgrowth are heavily skewed so we log transform them or use a polynomial gen tobinql=ln(tobinq) gen revgrowthl=ln(1+revgrowth) gen lvgsq=lvg*lvg We also create markers:*LBOmarker, 1/0 for when the LBO takes place*LBOregmarker, 1/0 for the last year of CS data (-1 or -2)*Somepatenting, 1/0 for patent(s) at some point in history*matchfilter for whether the obs is in the match set*regfilter for whether the obs is in the reg*savefilter for whether the obs will be saved for the analysis The propensity score regression is: . logit lboregmarker lvg lvgsq tobinql revgrowthl taxtoassets ebitdatoassets revtl if regfilter==1, robust Iteration 0: log pseudolikelihood = -1146.5017 Iteration 1: log pseudolikelihood = -1103.9281 Iteration 2: log pseudolikelihood = -1091.3342 Iteration 3: log pseudolikelihood = -1088.5307 Iteration 4: log pseudolikelihood = -1088.0634 Iteration 5: log pseudolikelihood = -1088.0171 Iteration 6: log pseudolikelihood = -1087.0487 Iteration 7: log pseudolikelihood = -1083.1541 Iteration 8: log pseudolikelihood = -1082.5635 Iteration 9: log pseudolikelihood = -1082.549 Iteration 10: log pseudolikelihood = -1082.549 Logistic regression Number of obs = 51,432 Wald chi2(7) = 166.01 Prob > chi2 = 0.0000 Log pseudolikelihood = -1082.549 Pseudo R2 = 0.0558 -------------------------------------------------------------------------------- | Robust lboregmarker | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---------------+---------------------------------------------------------------- lvg | 3.78305 1.430146 2.65 0.008 .9800142 6.586085 lvgsq | -3.08942 1.11948 -2.76 0.006 -5.28356 -.8952793 tobinql | -.5689659 .0548885 -10.37 0.000 -.6765454 -.4613865 revgrowthl | -.9232836 .848426 -1.09 0.276 -2.586168 .7396008 taxtoassets | 4.224245 1.03806 4.07 0.000 2.189684 6.258806 ebitdatoassets | 2.020758 .4627235 4.37 0.000 1.113836 2.927679 revtl | .5396504 .0627495 8.60 0.000 .4166636 .6626372 _cons | -6.332186 .7769002 -8.15 0.000 -7.854882 -4.809489 -------------------------------------------------------------------------------- Note: 1210683 failures and 0 successes completely determined. With the increased sample, we now find that LBOs are associated with reduced likelihood of patenting overall. However, this results is driven by the LBOs in period 1 from 1980 to 1995. In period 2, 475from 1995 to 2015, 180there is no statistically significant association. We also tried using decades but this wasn't instructive.
=Ed's version from 2018=
