ACCT30001 Firm and Financial Statement Analysis Tutorial Six

TUTORIAL SIX
Tutorial Six beginning 17th April 2023

TOPIC: Valuation using Earnings Models

This weeks questions are about valuation using the earnings model. The first question address
the application of the model. Question 2 to 4 considers what are the sources of the most
uncertainty in undertaking valuation and we assess some empirical evidence as to whether
past earnings are useful for forecasting future earnings.

Question One: Use the Earnings Valuation Model to value Apple Inc as at March 2023
Apple Inc designs, manufactures and markets smartphones, personal computers, tablets,
wearables and accessories and provides variety of related services such as an App store.
The following table gives the firm's past earnings per share and analysts forecasts of future
earnings for the years 2023 and 2024.

Future Estimates
Past 1 2 3
2020A 2021A 2022A 2023E 2024E 2025E
Past and Predicted Earnings Per Share 3.31 5.67 6.15 5.96 6.59

US Country Risk premium is 5.94%.

https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/ctryprem.html
Us Twenty-Year Treasury rate is 1.5%. Make an assumption regarding the terminal growth
rate.
a. Value the firm as at March 2023 and compare it to the current market valuation.
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

Solution
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

See attached excel spreadsheet.

• Basic approach we will be forecast out for two years, 2023 and 2024, as these are
the only years we have analysts forecasts obtained from Yahoo Finance
• We need a terminal or continuing value as at end of 2024 and therefore we need a
forecast of earnings for 2025. We assume earnings will grow from 2024 to 2025 at
the rate of 4%. Thus our 2025 estimated earnings is $6.59 * 1.04 = $6.85
• We will assume long-term growth of Apple is 4%
• Our estimated cost of equity is 8.03%. See calculation below.

Valuation of Apple Inc as at March 2023 Future Estimates

Past 1 2 3
2020A 2021A 2022A 2023E 2024E 2025E
Past and Predicted Earnings Per Share 3.31 5.67 6.15 5.96 6.59 6.85
Continuing value 169.90
Discount Factor 1.08 1.17 1.17
Present Value 5.52 5.65 145.57

Estimated Price (5.52+5.65+145.57) 156.7

Assumptions
Cost of equity 8.03%
Terminal growth rate 4.00%

Continuing/Terminal Value = 6.85/(0.0803-0.040) = 169.90

Discount Factor = (1.083)t = for 2024 = (1.083)2 = 1.17

Cost of Equity
Beta 1.1
Market risk premium 5.9%
Risk free rate (20-year US Treasury) 1.5%

8.03%

Cost of equity = Risk Free Rate + (Market-Risk Premium * Beta) = 1.5% + (1.1*5.9%) =
8.03%
Sources of Cost of Equity Inputs
Beta 1.1 Yahoo Finance estimates at 1.3. This appears too high. Arbitary adjustment to 1.1
Market risk premium 5.9% Prof Damodaran's website*
Risk free rate (20-year US Treasury)
1.5% Yield on US 20 year government bonds

*Prof Damodaran's website (http://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/ctryprem.html)

Question Two
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

Our lecture example provide an estimated valuation of Woolworths at March 2023

(using earnings forecasts based on some simple trend analysis. There is a significant
difference between the current share price of $36.92 and our estimate of $32.54.
Review the valuation excel spreadsheet and our valuation assumptions and estimates
to determine the main reasons for uncertainty that explains the differences between the
market price and the estimated valuation. As part of this consider the attached AFR
article on economist forecasts of GDP growth for December 2022.
The impact of a valuation input on valuation uncertainty is a function of both the magnitude
of the impact of the input on valuation and the uncertainty associated with estimating that
specific input.
There are three primary inputs into all valuation models:

• Forecasts of periodic payoffs in next few years

• Forecast of terminal value based on forecast growth rate (g)
• The discount rate r

Of these three inputs it is the estimates of g and r that have the greatest impact on the
magnitude of a valuation and thus are the main reasons for the difference between market
price ($36.92) and our valuation ($32.54). So estimates of g and r have a very large impact
on the magnitude of a valuation. Estimates of g and r are also very uncertain. Thus, in turn it
is the uncertainty associated with estimating g and r that are the primary determinants of
valuation uncertainty.
As an illustration of the magnitude of the impact of g and r on valuation assume we change
each of the above inputs in the Woolworths valuation by 0.5%. The impact of the change in
input on valuation is as follows:
• Increase forecast earnings for 2023 from $1,421.40 to $1,428.50
– Valuation shift from $32.54 to $32.545
• Shift terminal value growth rate from 3.50% to 4.00%
– Valuation shifts from $32.54 to $37.02
• Shift Discount rate from 7.39% to 7.89%
– Valuation shifts from $32.54 to $37.63

As an illustration of the uncertainty associated with forecasting g consider the forecasts by

economists (in the attached AFR article) of GDP as at April 2022 for the year-end December
2022. The forecasts range from a minimum of 2.5% to a maximum of 6.20%. This range
implies forecasting GDP just for this year is highly uncertain. In a valuation model we are
attempting to forecast long-run growth which will be even more uncertain.
Estimating discounts rates is also very uncertain.
• To estimate r (CAPM) we need to known both Beta and the Market Risk Premium
• To estimate Beta we need to measure how strongly a firms performance varies with
the economy
• This requires a long-time series as we need both good and bad states of the
economy (booms and recessions) to measure how a firms performance varies with
the economy
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

• However over a long time period the firm’s business model would have changed
making the estimate severely biased. Many firms have only existed for a short-time
period
• A short-time period would be more relevant and less biased but we may not have
enough variation in the economy (e.g both good and bad states of the economy) to
obtain a reasonably precise estimate of Beta

Question Three
Predicting Earnings using Time-Series Properties (for valuation)

Assume current profitability is modeled primarily as a linear function of lagged profitability.

Consider the results reported below from the estimation of the following regression which
examines the association between current profitability (ROE in year t) and lagged
profitability (ROE in year t-1).
ROEt =α +βROEt-1 +ε

This regression was estimated across a sample of large Australian firms across the period
from 2013 to 2020. I report results for both the full sample and then five sub-samples based
the level of ROE. Specifically, all companies were categorized into five portfolios from
highest ROE (winners) to lowest ROE (losers). The regression was then estimated for each
portfolio. This is conceptually similar to Q3 from last week (Tutorial Five).
Answer the following two questions:
Q4a) Explain in simple plan language what the magnitude of the coefficient β implies
about the use of the most recent current period earnings to predict future earnings?

Q4b) For those firms earning a ‘normal ROE’ for example about 10 to 12% the
estimate coefficient on ROE is reasonably close to 1 (see portfolios 3 and 4). Does
this imply future earnings are easy to predict? Use some of the statistical summary
output to support your answer.
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

Results from estimation of the following regression: ROE t =α +βROE t-1 +ε

Parameter Estimate t Value R-Squared

Panel A All Firms
Intercept 0.0416 12.86
ROEt-1 0.4874 23.05 0.3963

Panel B Firms in Five portfolios from Lowest ROE (Portfolio 1 basically loss firms) to Highest ROE (Portfolio 5 extreme winners)
ROE Portfolio ROE
Portfolio Parameter Estimate t Value R-Squared Min Max
1 Intercept -0.0433 -6.01
1 ROEt-1 0.2806 20.34 0.1732 -1.2632 0.0099
2 Intercept -0.0078 -2.18
2 ROEt-1 0.7353 28.6 0.3521 0.0103 0.0737
3 Intercept -0.0222 -7.39
3 ROEt-1 0.9439 30.49 0.3695 0.0739 0.1244
4 Intercept -0.0397 -9.99
4 ROEt-1 1.0899 34.72 0.4422 0.1245 0.2074
5 Intercept 0.0000 0
5 ROEt-1 0.4868 14.32 0.1887 0.2077 1.1057

The β represents the mean percent amount that ROE in t changes for a percent change in
ROE in period t-1. If current period earnings was fully persistent, from one period to the next,
then the β coefficient would = 1. If current period earnings was complete noise so that it is
completely uninformative about the future earnings then the coefficient would be 0.
Therefore, a β coefficient of 0.48 implies that current period earnings are moderately
persistent into the immediate next period. Put simply, a significant amount of current period
earnings, approximately 48 percent, will most likely persist into the next period. The past
matters for predicting the future (and therefore it is important that we precisely measure the
past).
Note it is important to appreciate that this OLS regression estimate of 0.48 likely understates
the persistence of earnings for a “typical firm” for two closely related reasons. First, this OLS
estimate of 0.48 is an average across all firms. Some firms will have higher persistence and
some firms will have lower persistence. Second, OLS parameter estimates tend to be more
influenced by observations in the tails of the distribution (this is because OLS regression
minimizes the sum of squared deviations between an observed outcome and a within sample
prediction). An extreme ROE is likely to have lower persistence than more normal levels of
ROE. Therefore for a firm with a more typical level of ROE the persistence is likely to be much
higher.
To understand this variation in persistence of ROE, the firms are formed into five portfolios
from highest to lowest ROE in Year t-1. The regression ( ROEt+1 =α +βROEt +ε ) is then
estimated within each portfolio. The results are reported in Panel B. There are two main
observations.
First, extreme losses (Portfolio 1) have very low persistence. Specifically, the coefficient of
0.28 implies that only 28 percent of the reported loss persists into the next period. For
example if the ROE was a negative 17 percent then our best estimate of next periods
earnings would be a negative ROE of 4.7 percent (0.17 *0.28 = 4.7 percent). The intuition for
the low persistence of extreme losses is as discussed last week the extreme loss firm have
strong incentives to take actions to return to profitability.
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

Second, firms with normal levels of earnings ( Portfolio 3 and 4) have extremely highly
persistent earnings. Specifically, the coefficient of 0.94 (portfolio 3) implies that 94 percent
of the reported profits persists into the next period. For example if the ROE was 12 percent
then our best estimate of next periods earnings would be an ROE of 9.4 percent (0.12 *0.94
= 10.8 percent).

These regression estimates of the persistence of earnings reflect the same concepts that we
discussed in last weeks tutorial in Question 3. The figure is replicated below and as we can
see the extreme losses and extreme winners revert to the mean. In contrast those firms with
normal level of earnings tend to on average have highly persistent earnings. The extreme
profits and losses revert to the mean which drives the β to be less than 1. The firms in the
middle of the distribution of ROE have highly persistent ROE of closer to 1.

Mean-Reversion of ROE
0.2000

0.1500

0.1000

0.0500

0.0000
1 2 3 4 5 6
-0.0500

-0.1000

-0.1500

Q4b) For those firms earning a ‘normal ROE’ for example about 10 to 12% the
estimate coefficient on ROE is reasonably close to 1 (see portfolios 3 and 4). Does
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

this imply future earnings are easy to predict? Use some of the statistical summary
output to support your answer.
The answer is no.
Future earnings are very difficult to predict. The difficulty of prediction is a function of the
uncertainty of the variable we are predicting (in our case ROE). The magnitude of the
uncertainty can be measured by the variation of future ROE. What factors explain the
variation in future ROE? Some firms in the future will have high ROE and other will have
low ROE. What explains this variation?
In our very simple model, we have simply examined the impact of a single factor being past
earnings (ROE) in period t-1 on future earnings in t. There are however many other factors
that will also affect the variation of future earnings in period t in addition to past earnings
(ROE) in period t-1.
The most significant of which are new economic shocks in period t. Put simply the earnings
we observe in the current period will be influenced by past earnings but the current period
earnings will also be impacted by new economic shocks in the current period. For example
assume we are predicting ROE for Woolworths in 2022 and the ROE for Woolworths in
2021 is 12%. This past earnings of 12% will have some predictive power for current period
earnings (as for example some of the pre-existing customers persist) but Woolworths
earnings in 2022 will be also impacted by new shocks in 2022 such as a new product
launch, actions by Coles, industry-wide shocks to the supermarket industry such as
customers using Uber.
The adjusted R-Square provides some insight into the magnitude of the importance of these
other factors in explaining the variation of next period earnings. The R2 is the percentage of
the variation in the dependent variable y that is explained by x. Thus the R2 can range from
0 to 100%. Therefore, in our regression model the R2 provides a measure of the
percentage of the variation in ROE that is explained by past ROE. We will assume what is
not explained by the past is due to new economic shocks.
Thus, the adjusted R-Square of 0.39 (see Panel A) implies the immediate past can explain
39% of the variation in future earnings. Therefore 61% of variation in future earnings that is
not explained by the past earnings and is due to new economic shocks.

Question Four
The Value-Relevance of Earnings for Explaining Share Returns
Consider the results reported below from the estimation of the following regression which
examines the explanatory power or importance of unexpected earnings for explaining annual
share returns
Returns = α + βEarnings + βUnexpectedEarnings + ε
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

Where Returns is the Annual Share Return of a firm. Earnings is the annual level of
earnings of a firm measured as ROE and UnexpectedEarnings is unexpected earnings
measured as the difference between this years and last years earnings (ROEt- ROEt-1).
This regression was estimated across a sample of large Australian firms across the period
from 2013 to 2020.
In QM1 and Econometric classes you would have learned than in an OLS regression model
of y = α + βx the R2 is the percentage of the variation in the dependent variable y that is
explained by x. Thus the R2 can range from 0 to 100%. Therefore, in our regression model
the R2 provides a measure of the percentage of the variation in Returns that is explained by
Earnings and UnexpectedEarnings.
Consider the R2 from the regression output below and answer the following questions
(a) Do you consider this to be a low, normal or high R2 measured against what you
consider to the objective of financial reports
(b) What are the main reasons that the R2 is at the reported level
(c) What are the implications of the magnitude of the R2 for policies and standard-setting
associated with financial reporting and disclosure more generally.

Regression of Return on
Earnings
Parameter Estimate t Value
Intercept 0.0535 6.87
Earnings 0.0535 11.42
Unexpected 0.0535 6.94
Earnings

R-Square 0.1028
Observations 2556

Solution
(a) Do you consider this to be a low, normal or high R2 measured against what you
consider to the objective of financial reports

Assume an objective of financial reports is to provide information for valuation. We can

consider the annual return to be a measure of the change in the valuation of a share across
a year.
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

The R-Squared of 10% implies that current period earnings only explains 10% of the
variation in annual returns (and thus change in valuation). The degree of explanatory of a
model for a variable can range from 0% to 100%. Prima facie then as 10% is at the lower-
end of the potential explanatory power it possibly could be considered low given certain
assumptions.

(b) What are the main reasons that the R2 is at the reported level

There are two potential explanations for a low R2.

First, as we have discussed accounting has some degree of measurement error and biases
such as the non-recognition of intangibles which may for example bias earnings downwards.
Therefore these imperfections will lower the R2 as the errors in earnings will lower its value-
relevance for the share markets in measuring economic value-added.
Secondly, as we have discussed and shown that the primary determinants of the value of a
firm is the expected long-term growth rate g and the discount rate r. Therefore most of the
variation in valuation (share prices) is due to the markets changing expectations of growth
(g) for a company and change in the discount rate risk (r) of a company.
Of the two explanations the second is highly likely to be most important. Even assuming we
could perfectly measure the past economic performance of a firm the expectation of the
future performance (g) will change during the year (for example a company announces the
launch of a new product) implying accounting earnings will not provide all the information
necessary to value a firm.

(c) What are the implications of the magnitude of the R2 for policies and standard-setting
associated with financial reporting and disclosure more generally.

Some commentators have argued that the low R2 implies financial reports should be
changed to aim for a high R2. Clearly as we have discussed financial reports can be
improved and this may give rise to a partial increase in the ability of earnings to explain
share returns and thus an increase in R2
However this commentary, of aiming for a high R2 misunderstands the primary reason for
the low R2 being estimates of future growth (g) and discount rates (r). How should we
(being the financial economic system) provide information to estimate g and r?
 ACCT30001 Firm and Financial Statement Analysis Tutorial Six

As a starting point given the uncertainty associated with estimating g and r it is arguable that
the financial accounting reporting system should not attempt to value, recognize and record
this estimated future growth (g) within the financial reports. Rather the valuation and
estimation of g and r should left to investors. The capacity of investors to estimate g could
be facilitated by:
1. Improving the measures of historical past performance as an input into estimating the
future. If we understand the past then we are able to better forecast
2. Enhance disclosure (to be discussed in Week 11)
3. Appreciate and understand that estimating g will always be very uncertain as we do
not know the future. Therefore rather than attempting to report g we should continue
to develop the financial system and products that enable this risk to be shared.