{{McNair ProjectsAcademicPaper|Project TitleHas title=Winner's Curse in Acquistions Acquisitions (Academic Paper),|Topic AreaHas author=Economic and Business TrendsEd Egan,Jim Brander|OwnerHas RAs=Ed EganAmir Kazempour, Will Cleland, Amir K, Jake Silberman,|Start Term=Fall 2016,|Status=Active,|Deliverable=Academic Paper,|AudienceHas paper status=Academics,|Primary Billing=AccMcNair01,Published
}}
==Paper DiscussionSummary== This paper finds evidence consistent with a Winner's Curse in the acquisitions of privately-held firms. This Winner's Curse is aggravated when there are more potential bidders and/or greater information asymmetries. The paper was accepted for publication in the Quarterly Review of Economics and Finance in February of 2017. ==Dbase Notes== --SDC_TABLE.sql --START AT NEW CODE SDC --Raw data in SDCBase --No dups SDCV2 --Code and restrict 40216 SDCNo --Build AcqNo SDCA1 --Add previous acqs SDCA2 --Add bin for IT (from SDCwIT) SDCA3 --Add bin for pub/priv (from SDC_test) SDCA4 --Add indu FE (40216) --CRSP_TABLE.sql CRSPFinal --Processed Raw CRSPMain --w/ cusip6 CRSPRelNoInc --Add in the DayNo and lose irrelevant CUSIP EstimationBaseNoDups Estimation1-5 5 --output tables EstimationTotal --THIS IS LIKELY DEPRECATED Competition EventBase --Only SDC relevant CUSIP -3 to +3 ... Flatten1 -- Rn3 to Rp3 and VRn3 to VRp3 Flatter -- Collapse 32109 EstimationOutput --Load it back up 30950 AbnormalReturn1 --Join back to Flatter ABR2 --Abnormal returns (plus momentum) AbRet --CARs 30950 ------------- Turnover --NEEDS REWORKING tobinq -- CPI Below that is supersceeded --GOV_Table and GOVERNANCE_NOTES.txt --Malmendier_tate.sql
| Governance Friendly vs. Hostile || CompustatSDC || D.N.E.
|-
| Friendly vs. Hostile CEO Overconfidence || SDCExecucomp || malmendier_Tate
|}
Boom - Created. Code in E:\McNair\Projects\Winner's Curse\Data\SQL\SDC_TABLE. Final variable in table SDCno_boom.
Competition - Created. Code in E:\McNair\Projects\Winner's Curse\Data\SQL\CRSP_TABLE. Final variable in table competition final.
IT Fixed Effect - Create an indicator variable taking the value one if the target has a NAICS code listed on page 31 of the article.
Idiosyncratic Volatility - RMSE, it should already be up on the database.
Industry Fixed Effect - Created. Jake, please update this.
Momentum - To control for systematic variations in return over time. The summation of the NYSE-Nasdaq-AMEX composite market return (as
calculated by CRSP) over the event window
Private Financial Data - Look at [https://wrds-web.wharton.upenn.edu/wrds/query_forms/variable_documentation.cfm?vendorCode=CRSP&libraryCode=crspa&fileCode=ccmfunda&id=bspr]. The variables takes value 1 when the balance sheet is not publicly available.
Percent Cash - Possibly this is going to be the PCT_CASH variable in SDC. This should be the PC_CASH field in the SDC table.
Previous Acquisitions - The number of previous acquisitions (from 1985 on) of privately-held firms completed by the acquirer prior to its announcement of the acquisition.
Tobin's Q: The ratio of market-to-book value of assets. This would be MKVALT/ AT in the SDC table.
Q) Is CUSIP in RiskMetrics current or historical? How can I use it to merge with CRSP?
A) The Legacy files have the current CUSIP. With each update, WRDS updated the historical data with the latest value. Starting with the new Director and Governance files, we no longer modify the historical data.
To merge any of these datasets with CRSP, you can join on the NCUSIP field. This will match a single Permco (company identifier), but you may still need to select the primary Permno because some companies issue more than 1 security.
Merging EXECUCOMP AND IRRC:
Q) What is the best way to merge Execucomp and IRRC (director and governance) databases? I understand they both have CUSIP, but that matching with CUSIPs isn't always ideal since the CUSIP can change over time for the same firm.
A) All Execucomp (CUSIP), IRRC Governance (CN6), and IRRC Directors (CUSIP) CUSIPs are header cusips, in that they reflect the most recent CUSIP that the company has had. There are however 2 main differences between Execucomp's CUSIP and IRRC's CUSIP and CN6:
1. While Execucomp uses 8-digit CUSIP, both IRRC datasets have 6-digit CUSIPs. Therefore, you need to use a SAS function like "substr" to take the first 6-digit of Execucomp CUSIP before merging it to IRRC data. ex: cusip6 = substr(CUSIP,1,6);
2. While header convention reflect "most recent" value of an identifier variables, it is dependent on the date of update of each dataset. Therefore, for CUSIPs that change in between the time that both databases are updated, you might find some discrepancy. To overcome this issue, we recommend that you make use of the CRSP name files (ex: stocknames), that contains all historical CUSIPs that a company has had.
Also, good to know:
Best Identifier to Use in RiskMetrics Directors- Changes to Identifier between Legacy and Current Files
1. The primary identifier in the current dataset is year-ticker-director_detail_id. Ticker is better populated than cusip in earlier years, but should be about the same in the latest years. For the legacy directors, director_detail_id was in use for some years and legacy_director_id for others.
2. Cusips in the legacy file are "header". With each update, WRDS modified all previous cusips for a company to match the latest. Although it may help in some cases, it may hurt in others and it is not quite the way the data was delivered. Starting with the new directors file, Cusips appear in the file as they were delivered (effectively making them historical going forward). Cusips in the legacy data were also uniformly truncated to 6 digits for consistency. The new file reports them as they are delivered, typically (but not always) 9 digits.
3. Tickers were not changed over time and appear as they were delivered. One example is legacy company id (legacy_pps_id) 25 has tickers ASN and IKN.
4. Each time IRRC - ISS - Riskmetrics was acquired, new identification methods were introduced. This is why you see legacy company and director IDs. Neither is more correct, we recommend choosing the most populated one for the years in which you are working.
5. Constructing a complete time series in the directors data is not easy. There is no single variable that is populated for all companies and all years.
6. Be very careful when combining the legacy data (through 2006) and the current data (from 2007). They were separated because the collection methods and meaning of some variables changed.
==Paper Discussion==
==Importing the CRSP data into the `winner' database (Amir) ==
Two important files:
Project --> Winner's Curse --> Data --> SQL --> CRSP_NOTES
Project --> Winner's Curse --> Data --> SQL --> CRSP_CODES
Data file is located in: E:\McNair\Projects\Winner's Curse\Data\CRSP-Oct10-c30cc89e52b70545-1.txt
Issue 1: The text file is 8 GB large and there's no way to open it in Textpad. I'll try to read the file in LogExpert which seems to do a neat job in in opening large files. Otherwise, I'll try to read the first few lines in the terminal.
Update: I managed to open the file in LogExpert. However, extremely inefficient to scroll over the file.
I copied the file to Z:\WinnersCurse\CRSP-oct10
Entered the following command in the terminal to read the header:
head -1 CRSP-oct10.txt
CREATE TABLE CRSP (
PermNo int,
date date,
EXcode int,
SIC int,
NCUSIP varchar(10),
Ticker varchar(10),
Name varchar(100),
NAICS int,
PrimEx varchar(1),
PERMCO int,
CUSIP varchar(12),
Price real,
VOL int,
RET real,
ShOut int,
trade int,
EQRET real,
VRET real
);
We have unexpected data types in few columns. We'll first start by copying data into a table named CRSPtmp with varchar type for most of the fields.
CREATE TABLE CRSPtmp (
PermNo varchar(10),
date date,
EXcode varchar(10),
SIC varchar(10),
NCUSIP varchar(10),
Ticker varchar(10),
Name varchar(100),
NAICS varchar(10),
PrimEx varchar(1),
PERMCO varchar(10),
CUSIP varchar(12),
Price varchar(15),
VOL varchar(10),
RET varchar(10),
ShOut varchar(10),
trade varchar(10),
EQRET varchar(10),
VRET varchar(10)
);
\COPY crsptmp FROM '/bulk/WinnersCurse/CRSP-oct10.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV
-- COPY 64614281
Next, to create an index on CUSIP:
CREATE INDEX CUSIPIndex
ON crspfinal(CUSIP);
===CRSP2 Variable Issues (Jake)===
permno-
date-
excode -
sic - letter "Z", value 711 (most are 4 digits),
ncusip -
ticker -
name -
naics -
primex -
permco -
cusip -
Price - Null values, negative numbers,
vol - null values
ret - null values, letter "B" as value multiple times, letter "C",
11332 rows were removed as a result of our regular expression use.
=== Normalizing SDC data ===
Make sure that the first line of the data file starts with the column names and column name are not preceded by empty lines. Also, remove the summary/report lines at the bottom of the text file.