Difference between revisions of "CEO Compensation and Returns to Public and Private Acquisitions: Testing the Shareholder Value Hypothesis"
(→To do) |
|||
| Line 25: | Line 25: | ||
==To do== | ==To do== | ||
| + | |||
| + | We want to add: | ||
| + | *A unique CEO-Firm key | ||
| + | *IT target - and sum through | ||
| + | *VC backed target - and sum through | ||
| + | |||
| + | ===Same Day Acquisitions=== | ||
We need to solve the multiple acquisitions on a single day problem. The result needs the variables listed for acquisitions below (crucially including pub/priv indicator, CAR5, sameday indicator, sharesoutstanding, with a cusip-year key). | We need to solve the multiple acquisitions on a single day problem. The result needs the variables listed for acquisitions below (crucially including pub/priv indicator, CAR5, sameday indicator, sharesoutstanding, with a cusip-year key). | ||
| − | + | An outline of a process to do this: | |
| − | + | *SDCv2 is the import with some simple exclusions and basic processing 49031 - but it was likely overwritten and became 40216. A rebuild from SDCBase appears the replacement but this seems to have thrown out same day acqs. | |
| − | + | *We therefore need to go back to the raw import of SDC_MA_Oct10v4-fixed-normal.txt into table "SDC". | |
| − | + | *Load table prc from backup: SHOUT appears to have come from prc, which in turn came from CRSPRelNOInc. *Load table CRSPRelNOInc: this a base table for both estimation and event window builds. | |
| − | + | *Recode some queries from the previous work: | |
| − | *SDCv2 is the import with some simple exclusions and basic processing 49031 - but it was likely overwritten and became 40216 | ||
| − | |||
| − | *SHOUT appears to have come from prc, which in turn came from CRSPRelNOInc. CRSPRelNOInc | ||
**We would need to calc day number again for SDC, which would mean bringing in the dayno table. | **We would need to calc day number again for SDC, which would mean bringing in the dayno table. | ||
**Then we'd need to create new estimation data for the missing records: acq_cusip, date_announced, acqkey, dayno, avg(ret) as ret, primex, vret, eqret, reldayno, retdayno (see EstimationBase) - but maybe this time take SHOUT with you! | **Then we'd need to create new estimation data for the missing records: acq_cusip, date_announced, acqkey, dayno, avg(ret) as ret, primex, vret, eqret, reldayno, retdayno (see EstimationBase) - but maybe this time take SHOUT with you! | ||
| − | + | *Run the estimation (using script ???.do) and build a new estimationtotal file and import it in estimationoutput | |
| + | *Recode some queries from the previous work: | ||
**Build a new EventBase, and everything in CRSP_TABLE.sql through to Flatter | **Build a new EventBase, and everything in CRSP_TABLE.sql through to Flatter | ||
**Join this to EstimationOutput, calculate the abnormal returns and then CARs. | **Join this to EstimationOutput, calculate the abnormal returns and then CARs. | ||
| − | + | *Finally join all of this mess back to the data that we already have. We may need to recompute targetit and targetvc. | |
| − | |||
| − | |||
| − | |||
| − | |||
==Variables== | ==Variables== | ||
Revision as of 16:49, 12 August 2017
| Academic Paper | |
|---|---|
| Title | Executive Compensation and Acquisitions |
| Author | Ed Egan, Jim Brander |
| Status | In development |
| © edegan.com, 2016 | |
Contents
Draft
There's an early draft, without tables called "CEO CompensationJB2.docx". It's in the dropbox (Dropbox\coauthoredprojects\ResearchWithJim\ExecCompAcqs) and in the main folder (E:\McNair\Projects\CompAcqs\). There's also a CEO.log file that details results run on the old dataset, for information only.
Database, dataset and scripts
The database is compacqs
The SQL script and input data is in:
Z:\compacqs\
The SQL script is:
BuildCompAcqs.sql
The STATA do file is Master.do and is in:
E:\McNair\Projects\CompAcqs\
The current version of the dataset is:
MasterV1-1.txt
To do
We want to add:
- A unique CEO-Firm key
- IT target - and sum through
- VC backed target - and sum through
Same Day Acquisitions
We need to solve the multiple acquisitions on a single day problem. The result needs the variables listed for acquisitions below (crucially including pub/priv indicator, CAR5, sameday indicator, sharesoutstanding, with a cusip-year key).
An outline of a process to do this:
- SDCv2 is the import with some simple exclusions and basic processing 49031 - but it was likely overwritten and became 40216. A rebuild from SDCBase appears the replacement but this seems to have thrown out same day acqs.
- We therefore need to go back to the raw import of SDC_MA_Oct10v4-fixed-normal.txt into table "SDC".
- Load table prc from backup: SHOUT appears to have come from prc, which in turn came from CRSPRelNOInc. *Load table CRSPRelNOInc: this a base table for both estimation and event window builds.
- Recode some queries from the previous work:
- We would need to calc day number again for SDC, which would mean bringing in the dayno table.
- Then we'd need to create new estimation data for the missing records: acq_cusip, date_announced, acqkey, dayno, avg(ret) as ret, primex, vret, eqret, reldayno, retdayno (see EstimationBase) - but maybe this time take SHOUT with you!
- Run the estimation (using script ???.do) and build a new estimationtotal file and import it in estimationoutput
- Recode some queries from the previous work:
- Build a new EventBase, and everything in CRSP_TABLE.sql through to Flatter
- Join this to EstimationOutput, calculate the abnormal returns and then CARs.
- Finally join all of this mess back to the data that we already have. We may need to recompute targetit and targetvc.
Variables
The unit of observation is an executive's compensation year, or firm and (year-1). Each observation must have a CEO identifier, a year, and CEO compensation info. Each observation also must have lagged acquisition information, though many of these variables will be zero, and should have lagged firm level control variables.
CEO variables: Year - the year of the CEO's compensation package Execid - belonging to CEO in that year age - of ceo in year tenurefirm - tenure with firm in years (int) calc'd from joined firm date tenureceofromdate - tenure as CEO in years (int) cal'd from becameceo date tenureceo - tenure since first appeared as ceo (for firm) tenurecomp - tenure since first appeared in execcomp (for firm) to first appeared as ceo tenurecfo - tenure since first appeared in comp as cfo (for firm) to first appeared as ceo tenurecfocomp - tenure since first of cfo/comp to ceo (for firm) prevcfo - 0/1 (with firm) ceobefore -0/1/ (any firm in execcomp) gender - 1 for male, 0 for female salary - from execcomp bonus - from execcomp optionsfv - options fairvalue optionsrpt - options reported value optionsblk - options black-scholes values tdc1 - total compensation from execcomp Acquistions (joined on cusip6, ayear s.t. ayear+1=year): numacqs - number of acquisitions within year for the firm numacqspriv - number of private acquisitions within year for the firm numacqspub - number of private acquisitions within year for the firm car5tot - the total 5-day Cummulative Abnormal Return (0 when no acqs) car5av - the average 5-day CAR(0 when no acqs) valctot - sharesoutstanding (in millions?) x car5tot valcav - sharesoutstanding (in millions?) x car5tot sharesoutstanding - only available when their was >=1 acquisition hasacqs - 0/1 of there were any acqs within year transactionvalue - the total reported transaction value of acqs targetvalue - the total reported value (transaction, enterprise, or equity in that order of appearance) of the acqs Firm (acquirer, joined on GVKEY, fyear s.t. fyear+1=year): fyear - fiscal year from CSTAT naics - as a string (for redundancy) naics2 - 2dg NAICS naics3 - 3dg NAICS state - name of state statenum - number assigned to states [0-61] as some non-states at - total assets ppent - plant, property equipment net total ppegt - plant, property equipment gross total revt - sames as sales ni - net income acqi - current income from acquisitions in CSTAT acq - cost of acquisitions from CSTAT mktvalue - market value CPI: deflator - for exec comp with 2010 as base year cpiavg - the average CPI for the compensation year deflatorlag - for fiscal/acq year with 2010 as base year cpiavglag - the average CPI for the fiscal/acq year
Notes from the draft
Execucomp data
Execuomp provides data on various characteristics of each CEO that can be used as explanatory variables for compensation, including age, sex, and experience.
Various compensation measures are provided, including salary, bonuses, and the value of option awards. We focus primarily on total compensation (variable TDC1 in Execucomp), which includes salary, bonuses, the value of option awards and other types of compensation.
Compustat data
We use Compustat data (also obtained through WRDS) for data on various firm-level explanatory variables, particularly firm size measures (employment and real plant and equipment) and industry identifiers (NAICS codes at the 2 digit and 3 digit level).
CPI data for deflating
We use annual data, converting compensation over time to real values using the CPI index as a deflator.
Old distribution from paper
After a small amount of data cleaning (consisting mainly of dropping observations for which data on important variables is missing) we are left with 6447 panels and 34,956 observations, implying an average tenure of about 5.4 years per CEO in our data.
For most observations (29,775), the number of acquisitions is zero. Exactly one acquisition occurs in 4,228 observations, exactly two occur in 1,122 observations, three occur in 391 observations, and in 310 observations there are four or more acquisitions. The data set contains 9,452 acquisitions overall, of which 8,211 are acquisitions of privately held companies and 1,241 are acquisitions of publicly traded companies.
In our primary analysis we therefore include observations (firm-CEO-year combinations) with more than acquisition. In such cases we consider two alternative measures of acquisition performance – the average abnormal return over all acquisitions in the year and the total abnormal return over all acquisitions. To see if using multiple acquisitions has an effect we can of course consider a subsample consisting only of observations with no more than a single acquisition.
Data Source Draws
COMPUSTAT
Query URL: https://wrds-web.wharton.upenn.edu/wrds/ds/comp/funda/index.cfm?navId=80
Execucomp
Query URL https://wrds-web.wharton.upenn.edu/wrds/ds/comp/execcomp/anncomp/index.cfm?navId=72
Saved Query: ExecCompForAcqsPaperV
Query spec:
Data Request ID e0a8b1599f177e67 Libraries/Data Sets comp/anncomp / Frequency/Date Range ann / 1992 - 2017 Search Variable TICKER Input Codes all item(s) -all- Conditional Statements n/a Output format/Compression tab / Variables Selected CONAME CUSIP GVKEY NAICS STATE EXECID EXEC_FULLNAME GENDER PAGE BECAMECEO CO_PER_ROL GVKEY JOINED_CO AGE BONUS CEOANN CFOANN OPTION_AWARDS_BLK_VALUE OPTION_AWARDS_FV OPTION_AWARDS_RPT_VALUE SALARY TDC1 TDC2 TITLEANN TOTAL_CURR YEAR
Note: We need GVKEY, CUSIP, and year to do the joins. NAICS and STATE can be supplemented by the COMPUSTAT and EXECUCOMP data.
Acquisition Data
This is currently taken directly from the last version of the Winner's Curse Dataset (see Z:\compacqs).
We need to rebuild this to include all of the times when we had multiple acquisitions on the same day. As we use various SDC dataset measure (like a public indicator, state, and naics) it isn't sufficient to just take the CAR5 and sharesoutstanding from the event regressions - we need to join it back to the SDC data first. We will then calculate car5tot and car5av and count private and public acquisitions by blowing out the data (so that the unit of obs is an acquisition, not an announcement day) and recalculating the aggregate variables.
