DIVERSIFICATION DISCOUNT IN FIRM VALUATION:
PROJECT INTRODUCTION Dr.
Diversification discount in a nutshell
Copyright By PowCoder代写 加微信 powcoder
Example segment-level and firm-level data
Elements of competent research
Data options
Typical sections of project report How to make a contribution Expectations
What R can do for you
DIVERSIFICATION DISCOUNT IN A NUTSHELL
THIS THEME GROUP CONCERNS
Discount on diversified firms¡¯ valuations,
when benchmarked against focused firms¡¯ valuations (Berger and Ofek 1995)
The literature often assumes
diversified firms = multiple-segment (MS) firms focused firms = single-segment (SS) firms
VALUATION METRIC FOR COMPARISON
Berger and Ofek (1995) use various asset/sales/EBIT multiples, incl. the q ratio used by Custodio (2014)
q = Market Value to Book Value of Total Assets BV of TA is measured with Compustat¡¯s variable at
MV of TA is estimated with
at – ceq + prcc_f*csho,
ceq = BV of common equity
prcc_f = closing share price at fiscal year end csho = number of common shares outstanding
THE COMPARISION BENCHMARK
For any real MS firm,
find an equivalent ¡®synthetic¡¯ MS firm as the comparison benchmark
Then compare the real MS firm¡¯s actually observed q (q_Obs) to the q imputed for the synthetic MS firm (q_Imp)
A valuation discount exists if q_Obs/q_Imp < 1, which means Excess value xv = log(q_Obs/q_Imp) < 0
IMPUTED Q RATIO
Suppose the real MS firm in concern has two segments. Segment A in the SIC = 2834 industry
Segment B in the SIC = 2023 industry
The 4-digit code is the Standard Industry Classification (SIC) code To construct the synthetic MS firm as the comparison benchmark,
select a typical SS firm from the SIC = 2834 industry;
select a typical SS firm from the SIC = 2023 industry.
The synthetic MS firm is a firm formed with these SS firms, as though they were the synthetic MS firm¡¯s segments
By typical, it means the median (or the average) in Custodio (2014)
DIFFERENT OPERATION SCALE?
The operation scales of the SS firms are different from the segments¡¯ of the MS firm? No worries
The q ratio is scale-free!
simply use the actually observed q ratios of the SS firms as the imputed q ratios for the segments of the MS firm
weight these q ratios by the segment assets (or segment sales) of the MS firm to obtain
a weighted average q ratio as the imputed q ratio for the MS firm (q_Imp)
accordingly, the excess values obtained are the asset- and sales-weighted excess value variables (denoted by xv_aw and xv_sw in the firm_xvExample.csv file generated by the R demo code)
EXAMPLE SEGMENT-LEVEL AND FIRM-LEVEL DATA
Segment-level data
Firm-level data
ELEMENTS OF COMPETENT RESEARCH
Research questions / Objectives Literature search
identify key prior studies to build upon
e.g., Berger and Ofek (1995) and Campa and Kedia (2002)
Analysis techniques (e.g., statistical models and tests)
borrow techniques from prior studies; mix and match and modify them
Implementation tool (e.g., R, Stata) Writing the project report
DATA OPTIONS
Commercial databases subscribed by Bayes
a range of choices on WRDS (but only those with subscriptions)
currently include Compustat, CRSP, Audit Analytics, and several more
important to understand the meaning of the data variables and the structure of the databases
e.g., sales (segment-level sales) vs. sale (consolidated firm-level sales)
For full lists of Compustat variables, click the ¡®Variable Descriptions¡¯ tab on the pages below:
Historical Segments (segment-level) Fundamentals Annual (consolidated firm-level )
TYPICAL SECTIONS OF PROJECT REPORT
Introduction
Literature Review Research Methodology Analysis and Results Conclusions
A very brief summary of your research
highlighting the most essential elements, findings, and/or takeaways often around 250 words
up to about half of the first page of the report
INTRODUCTION
describe the objectives of your research, e.g.,
to answer a research question that has not yet been fully addressed
to compare the findings of different analysis approaches and assess their relative usefulness
motivate the objectives, i.e., explain why they are interesting and important
not just to you but to most other people (or certain important parties, e.g., investors, regulators, etc)
give a quick ¡®executive summary¡¯ of the research, including key results
your contributions
(i.e., what you incrementally add to the world and why it is worth knowing your addition)
LITERATURE REVIEW
give an overview of the related literature, especially those most directly related to your project
(e.g., where you borrowed the analysis techniques)
explain how the reviewed studies are related to your project, or how they are related to each other in the literature
RESEARCH METHODOLOGY
explain the sample construction, the analysis methods, and how they fit with each other
explain why the methodology can serve the project objectives (e.g., answer your research questions)
FAQ: How large should the sample be?
The larger it is, the stronger your evidence is, the more general your conclusions are. Thus, as large as it could feasibly be (without freezing your pc, making it too slow, etc)
Note: Only 1GB memory for RStudio Cloud¡¯s free-tier users
ANALYSIS AND RESULTS
report the main analysis in professionally-looking tables (and figures)
see published research articles for what professionally-looking means
make sense of the analysis results
e.g., explain why certain results are not as expected, highlight certain expected results
report any additional analysis (e.g., subsample analysis, robustness checks)
CONCLUSIONS
summarize the key takeaways most consistent with the evidence fom your analysis
HOW TO MAKE A CONTRIBUTION
baseline: replicate the diversification discount finding using more recent data additionally, in any of the following directions
deepen the understanding for certain interesting scenarios how about for a particular industry?
how about for a particular sample period/year?
what else?
consider new explanatory variables
use alternative methodologies
in any case, need to justify your choice convincingly!
any insight for practitioners?
NEW EXPLANATORY VARIABLES
What explanations/variables have been looked at?
e.g., see papers citing Berger and Ofek (1995), such as the following in the shared folder on OneDrive:
Martin and Sayrak (2003)
Hoechle et al (2013)
How about using more recent data to re-examine old explanations with mixed findings? What explanations/variables can you newly propose?
Can you find quantifiable measures for the new explanations?
ALTERNATIVE METHODOLOGIES
For example,
Berger and Ofek (1995) use asset/sales/EBIT multiples
Custodio (2014) uses q ratio, which is equivalent to B&O¡¯s asset mulitiple
Any relevant metrics (e.g., another type of valuation multiples) not considered before? explain why a proposed metric has an edge over previously used metrics!
You need to understand the R code sufficiently well to be able to
replace the q ratio used to define the excess value by another valuation multiple
EXPECTATIONS
Very, very occasionally, there are reports with marks in low to mid 80¡¯s But most reports fall in the 40¡¯s to low 70¡¯s mark range
see the feedback on two reports resubmitted for the resit that still earned below 40:
report 1 and report 2
see a report (of a different theme group) that earned below 40
HELP ME TO GIVE HIGHER MARKS
I always want to get higher marks
but I need to let the internal second marker and the external examiner know the rationales for giving the marks.
For example,
I had given a low 80¡¯s to a student but the second marker independently gave a lower mark, leaving the final mark to be a high 70¡¯s.
The external examiner had questioned a case where I gave a low 40¡¯s
She believed it should be a fail.
OTHER THINGS BEING EQUAL,
Which of the reported tables below deserve higher marks?
CUT-AND-PASTED SOFTWARE OUTPUT (WITH DISTRACTING UNNECESSARY DETAILS)
ORIGINALLY PREPARED TABLES (WITH THREE DECIMAL PLACES)
WHAT CAN LEAD TO LOW MARKS?
For example,
in a report from a theme group on accounting restatements, it
repeatedly misspelt ¡®restatement¡¯ as ¡®reinstatement¡¯.
had a literature review that discusses not a single article with ¡®restatement¡¯ in the title; The quality of this report was in line with the student¡¯s low participation in the module:
Never attended any of the 5 meetings, not even the first one Submitted a very brief project proposal. I cautioned the student:
¡°Many details need to be filled in before one can tell if the project is likely to be workable. Please do not underestimate the effort required to complete the project and the risks involved if it is not taken seriously enough.¡±
Never submitted the milestone Methodology chapter; no response to submission reminders.
TIME TO SPEND ON THE PROJECT
No less than what you would spend on an equivalent taught module (full-year/one-term) Actually, should be more
because of the independent-learning nature of the FYP/ARP and the fact that you are an inexperienced student researcher
PRIOR STUDIES IN THE SHARED FOLDER
Must you read them all? Short answer: No Long answer:
The materials are there to help a student get started.
One may even go beyond those materials, depending on your aspiration and opportunity cost
So everyone should make her/his own judgment
In any circumstances, at least be familiar with the details in
Berger and Ofek (1995), Campa and Kedia (2002), and Custodio (2014)
HOW TO READ PRIOR LITERATURE
Should read broadly and assess the credibility and relevance of different reference sources to decide what to cite in the project report
Avoid articles from lower-quality journals (often with open access) found in the Internet
By referring to the studies cited in a study, you can map out the chronological development of the related literature
Citations on the WEB OF SCIENCE linked page in the library catalog is your friend
The Citation count indicated is an (imperfect) measure of credibility and relevance
In most cases, simply reading a study¡¯s introduction (often ~5 pages) would already let you understand a large part of its essence without getting into the fine details.
Before reading the introduction, read the abstract (often ~250 words) to decide whether it is worth reading further.
BARRIERS TO SUCCESS
Bad time management
too much time on trying to find the best objectives too little on actually working out the analysis
Unrealistic / naive planning
wishfully plan to spend only an afternoon to collect the data for analysis
actually need substantially more time to learn where to find and how to use new variables No justification to choices made in the research process
significant results can¡¯t be guaranteed but how to make choices is fully controllable Insufficient attention to details
all kinds of inconsistency in the research process and the project report eg, research cited in the main text not included in reference list
Zotero can help you prevent such inconsistency
(see videos: get-started, insert-citations; MS Word plugin)
WHAT R CAN DO FOR YOU
Diver_CleanData.rmd
CLEAN DATA
do sanity checks and an initial clean-up on the raw data downloaded from WRDS
apply various filtering criteria to construct the sample for analysis
save the cleaned data as segmentHist_cleanExample.rds and firm_cleanExample.rds
files in R¡¯s native dataset format (.rds)
CONSTRUCT MEASURES AND RUN REGRESSIONS
Diver_Measures.Rmd
construct the required variables (e.g., q ratio, excess value)
write the final data as the file firm_xvExample.csv
run several regressions and present the results side-by-side as columns in a single table
EXPORTING THE EXCESS VALUE VARIABLES
The code files were originally written for replicating some diversification discount results in Table
III of Custodio (2014)
Simple changes to the codes should let you create the Excess Value variables for measuring the
diversification discount (in terms of q ratios) for a different sample period
To export the file firm_xvExample.csv with the Excess Value variables (e.g., the asset- and sales-
weighted xv_aw and xv_sw),
click on the check box next to the .csv file in the lower-right panel of your copy of the shared project on RStudio Cloud
then click the ¡®More¡¯ drop-down menu of the panel (make sure the panel is wide enough to show the option) and select ¡®Export¡¯ to download the file
you can use another software (e.g., Stata) to read in the file and work with the Excess Value variables
BASE R AND R PACKAGES
Base R refers to the capability of R before loading any additional package The demo codes also use the R packages below:
{magrittr} allows piping together different operations into one sequence, like the different steps of a factory production line
{tidyverse} provides functions to work on a dataframe (if you like, a ¡®worksheet¡¯), such as
filter(), select(), mutate(), left_join(), group_by(), etc
{lubridate} provides functions to work on date-type variables, such as
ymd(), year(), month(), day()
{stargazer} allows the results of multiple regression models to be reported in a single table neatly
RSTUDIO INSTALLED IN YOUR PC (VS. RSTUDIO CLOUD)
R and RStudio are free software; can be installed locally in your pc
Please google how to install R and RStudio in windows for related instructions
RStudio¡¯s free-tier account
Limited to 1GB memory and only 25 hours of usage per month
Thus, will restrict the total number of years of data you can work with
However, you can use the limited memory more efficiently
by first removing from the downloaded data the segment-level data variables that will not be used at all before you use RStudio Cloud to read in the .csv data file
LET¡¯S GO TO THE CODES
The DiverDiscount project shared on RStudio Cloud:
https://rstudio.cloud/project/1747338
You must save a permanent copy to your own account to continue to work with the shared project.
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com