© 2004 LifeCycle Returns, Inc. All Rights Reserved - 1 - VALUATION MODELS: ACCURACY AND PREDICTIVE CAPABILITY (A WORK-IN-PROGRESS) Financial Management.

© 2004 LifeCycle Returns, Inc. All Rights Reserved

- 1 -

VALUATION MODELS:VALUATION MODELS:ACCURACY AND PREDICTIVE ACCURACY AND PREDICTIVE

CAPABILITYCAPABILITY(A WORK-IN-PROGRESS)(A WORK-IN-PROGRESS)

Financial Management Association International

New Orleans, Louisiana

October 9, 2004By

Rawley Thomas

President

LifeCycle Returns, Inc.

[email protected]


- 2 -

Sources:Financial Statements and Price

Data – CapitalIQCalculations – LCRT Platform

THE CONCEPT OF VALUE MANAGEMENTTHE CONCEPT OF VALUE MANAGEMENT A strong correlation exists between enterprise value

and the spread between CER (cash economic return) and cost of capital

Econometric DCF models can be built and validated against historical data to quantify this correlation and identify the key operating drivers which most significantly impact value

Managements can use these models to analyze their own corporate performance and to improve their decision making and value generation

Portfolio managers can use these models for buy / sell decisions


- 3 -



IN A CAREFULLY IN A CAREFULLY CONTROLLED CONTROLLED

EXPERIMENT IN AN EXPERIMENT IN AN ECONOMICS ECONOMICS

LABORATORY, VERNON LABORATORY, VERNON SMITH et. al. SMITH et. al.

DEMONSTRATES DEMONSTRATES SIGNIFICANT SIGNIFICANT

DIFFERENCES OF TRADED DIFFERENCES OF TRADED PRICES FROM KNOWN PRICES FROM KNOWN

INTRINSIC VALUESINTRINSIC VALUES

Vernon L. Smith, Gerry L. Suchanek, and Arlington W. Williams, “Bubbles, Crashes, and Endogenous Expectations in Experimental Spot Asset Markets,” in Vernon Smith, Papers in Experimental Economics, Cambridge University Press, Cambridge, 1991, pp. 339-371, chart from p. 352.


- 4 -



NOTICE THE SAME PATTERN IN AN INTRINSIC VALUE CHART NOTICE THE SAME PATTERN IN AN INTRINSIC VALUE CHART WHICH ENABLES US TO VISUALIZE THE MEASUREMENT OF WHICH ENABLES US TO VISUALIZE THE MEASUREMENT OF

ACCURACY OF A DCF MODEL PRICE ACCURACY OF A DCF MODEL PRICE LEVEL LEVEL FOR HPFOR HPUSING ONLY ACTUAL REPORTED FINANCIAL DATA AND THE SAME GLOBAL USING ONLY ACTUAL REPORTED FINANCIAL DATA AND THE SAME GLOBAL

PARAMETERS ACROSS THE ENTIRE UNIVERSE TO DRIVE A MECHANICAL LIFE PARAMETERS ACROSS THE ENTIRE UNIVERSE TO DRIVE A MECHANICAL LIFE CYCLE FORECAST OF CASH FLOWS FOR EACH COMPANYCYCLE FORECAST OF CASH FLOWS FOR EACH COMPANY

Notice the large high low variation around the intrinsic valuations,

providing opportunity for profitable trading

The Absolute “Tracking” Error Intrinsic Value vs. Actual Equals the Absolute Geometric Mean Error

Between The Intrinsic Value Red Line And Closing Prices at Fiscal Year + 3 Months

(hollow circles) Over the Number of Years

LCRT

Residual Income

LCRT15.2

Residual Income65.2


- 5 -



FOR HEWLETT-PACKARD, A FREE CASH FLOW MODEL FOR HEWLETT-PACKARD, A FREE CASH FLOW MODEL DISPLAYS LOWER ACCURACY THAN LCRTDISPLAYS LOWER ACCURACY THAN LCRT

(OF COURSE, H-P IS ONLY A SAMPLE OF ONE)(OF COURSE, H-P IS ONLY A SAMPLE OF ONE)

LCRTFree Cash Flow

LCRT15.2

Free Cash Flow23.5

Absolute “Tracking” Error


- 6 -



BEHAVIORAL EXPLANATIONS OF BEHAVIORAL EXPLANATIONS OF VIOLATIONS OF INSTANTANEOUSLY VIOLATIONS OF INSTANTANEOUSLY EFFICIENT MARKET’S HYPOTHESISEFFICIENT MARKET’S HYPOTHESIS

People employ significantly different and inconsistent models of fundamental valuation, relying on various forecasts

Depending on the weights of all the classes of people buying and selling a stock at any point in time, the actual price will diverge significantly from the long term intrinsic value

Strongly held academic beliefs in instantaneous market efficiency impede empirical research to show otherwise

Price event studies only demonstrate that the market reacts in the correct direction, but not necessarily by the correct amount

Robust, accurate DCF models of intrinsic valuation are required to empirically test instantaneous market efficiency


- 7 -



SEVERAL PROFESSORS ARE SEVERAL PROFESSORS ARE COLLABORATING WITH LCRT TO COLLABORATING WITH LCRT TO

DETERMINE THE ROBUSTNESS AND DETERMINE THE ROBUSTNESS AND ACCURACY OF TRADITIONAL MODELS AS ACCURACY OF TRADITIONAL MODELS AS

NULL HYPOTHESESNULL HYPOTHESES

Sally Webber of NIU and Doug Clinton of NIU

– EVA® or Residual Income Models

– Feltham-Ohlson Model

Adam Gehr of DePaul

– Dividend Discount Models


- 8 -



TESTED MODELSTESTED MODELS

8 X EBITDA

Feltham-Ohlson

Residual Income

Free Cash Flow

LCRT


- 9 -



COMPARISON OF FELTHAM-OHLSON COMPARISON OF FELTHAM-OHLSON AND FREE CASH FLOW PERPETUITYAND FREE CASH FLOW PERPETUITY

Feltham-Ohlson Based on market value of

equity/ operating assets regressed against return on assets, change in return on assets, and growth rate in assets

From Jing Liu and James A. Ohlson, “The Feltham-Ohlson Model: Empirical Implications,” Journal of Accounting, Auditing and Finance, 2000, v15 [3, Summer], pp. 321-331, especially p. 326-327.

Programmed with the aid of Sally Webber, Accounting Professor, Northern Illinois University

Free Cash Flow Perpetuity Based on growing free cash

flow for T years and capitalizing the terminal year’s free cash flow into perpetuity

Free cash flow = income after taxes + depreciation and amortization – non-operating items after tax – normalized capital expenditures – working capital additions

The terminal year’s cash flow is capitalized by a CAPM nominal discount rate less a nominal growth rate

From specifications by Dan Van Vleet of Willamette


- 10 -



COMPARISON OF RESIDUAL COMPARISON OF RESIDUAL INCOME AND LCRTINCOME AND LCRT

Residual Income From PV of growing excess

residual income (EVA®) for T years plus release of capital at terminal value employing a CAPM cost of capital

Bennett Stewart, The Quest for Value, Harper Business, 1991, especially p. 324-325.

Programmed with the aid of Sally Webber, Accounting Professor, Northern Illinois University

LifeCycle Returns (LCRT) From PV of net cash flows for

50+ years using a market derived discount rate

Net cash flows derive from fading growth rates and cash economic returns applied to constant dollar gross investment less replacement assets less growth in gross investment

See Bartley J. Madden, CFROI Valuation: A Total System Approach to Valuing the Firm, Butterworth-Heinemann, Oxford, 1999 and LCRT.com


- 11 -



Income

+ Depreciation and Amortization

+/- Inflation Adjustments

+/- Finance / Non-Operating Expense Adjustments

Assets

+Accumulated Depreciation

+Inflation Adjustments

+/- Finance / Non-Operating Balance Sheet Adjustments

Cash Economic Return (CER)*

Constant Dollar Gross Asset Base

Super Sector CER Fade-To

Company Cash Economic Return (Fade From)

Company Cash Economic Return Fade-To

CER Fade Rate

Super Sector Growth Fade-To

Company Past CapEx Real Growth Rate and Real Sustainable Growth Rate (Fade From)

Super Sector Gross Asset Growth Fade Rate

Company Specific Real Investors’ Discount Rate, Adjusted for Leverage, Asset Mix, and Life

= Enterprise Intrinsic Value

+ Cash – Debt

= Equity Intrinsic Value

⁄ Number of Shares

= Intrinsic Value Per Share

LCRT FRAMEWORKLCRT FRAMEWORKAccounting

Constant Dollar Cash

Economic Cash Real Rates of Return on Un-

Depreciated Gross Assets (Reflecting Real

IRR of all Projects Currently in Place)

Valuation

Asset Mix

Asset Life

Life Cycle

Valuation

Model

Super SectorsIndustrials

UtilitiesFinancials

(*Real ROE)


- 12 -



Income $206

A: Eliminate Non-Operating Special Extraordinary Items After Tax 33

Items (-) Non-operating Expense After-Tax (16)

B: Translate to Cash Non-Cash Charges 333

C: Restate for Inflation Inflation Gain on Non-Fixed Assets 14

D: Eliminate Leverage After-Tax Interest (Debt and Operating Leases) 134 $781

$206 Rentals – Principal Payments 77 Current Dollar

Income E: Capitalize Expenses (-) Advertising and R & D After Tax (0) Gross Cash Flow

Assets Total Assets $5,825 Current Dollar

$5,825 A: Eliminate Non-Operating (-) Non-Operating Assets (137) Investor Gross

Items (-) Purchase Goodwill (1,531) Cash

Receivables Reserve 23 Investment

B: Translate to Cash Invest. LIFO Reserve 141 $5,704

Accumulated Depreciation 1,580

C: Restate for Inflation Inflation Adjustments to Land, Gross Plant and Deferred Taxes 249

D: Eliminate Leverage Gross Leased Property from Operating Leases 1,202

E: Capitalize Expenses Capitalized Advertising, R & D 0

F: Capital Owner Cash Invest. (-) Operating Non-Interest Bearing Liabilities (1,648)

CASH ECONOMIC RETURN EXAMPLE:CASH ECONOMIC RETURN EXAMPLE:ACCOUNTING TO CASHACCOUNTING TO CASH

SUPERVALU– 2001 ($Millions)SUPERVALU– 2001 ($Millions)


- 13 -



CASH ECONOMIC RETURN EXAMPLE:CASH ECONOMIC RETURN EXAMPLE:CASH TO ECONOMICSCASH TO ECONOMICS SUPERVALU– 2001 ($ MILLIONS)SUPERVALU– 2001 ($ MILLIONS)

Current Dollar Gross Cash Flow

$781Non-Depreciating

Asset Release

$727

($5,704)

Current Dollar Investor Gross

Cash Investment

Economic Life: 11.55 Years

Cash Economic Return - IRR: 9.09% Years IRR

11 8.62

12 9.48

11.55 9.09


- 14 -



0

5

10

15

20

25

0 1 2 3 4 5 6 7 8 9 10

Year

Ca

sh

Ec

on

om

ic R

etu

rn

LIFE CYCLE OF CASH ECONOMIC RETURNS (CER)LIFE CYCLE OF CASH ECONOMIC RETURNS (CER)

Firm Fading CER*

CER Momentum Effect

Firm CER Fade-To

Super Sector CER Fade-ToFirm CER Fade-To Difference

Firm Discount Rate, Adjusted for Leverage, Asset Mix, and Asset Life

Cash Flows continue for 50 Years =>

Super SectorsIndustrials

UtilitiesFinancials

(*Real ROE)


- 15 -



TRADITIONAL ACCOUNTING MEASURES FIRST TRADITIONAL ACCOUNTING MEASURES FIRST UNDERSTATE AND THEN OVERSTATE ECONOMIC UNDERSTATE AND THEN OVERSTATE ECONOMIC

RETURNS AS ASSETS AGERETURNS AS ASSETS AGE(ASSUMING CONSTANT OUTPUT = CONSTANT DOLLAR LEVEL ANNUITY)(ASSUMING CONSTANT OUTPUT = CONSTANT DOLLAR LEVEL ANNUITY)

(A DESIRED ANNUAL PERFORMANCE MEASURE REFLECTS THE PROJECT IRR)(A DESIRED ANNUAL PERFORMANCE MEASURE REFLECTS THE PROJECT IRR)

NOTE: The Annual CER

each and every year precisely equals the IRR of the project.

-$10,000

PROJECT

$1,740

Life = 8 Years

IRR = 8.00%

Annual Performance Measures of Project

Year

1 2 3 4 5 6 7 8

Income 490 490 490 490 490 490 490 490

Depreciation 1,250 1,250 1,250 1,250 1,250 1,250 1,250 1,250 Gross Cash Flow 1740 1740 1740 1740 1740 1740 1740 1740

Gross Plant 10000 10000 10000 10000 10000 10000 10000 10000

Accumulated Depreciation 1250 2500 3750 5000 6250 7500 8750 10000

Net Plant 8750 7500 6250 5000 3750 2500 1250 0

Return on Net Assets =

RONA = Income/Net Plant 5.60% 6.53% 7.84% 9.80% 13.07% 19.60% 39.20% ∞

Cash Economic Return

(CER) 8.00% 8.00% 8.00% 8.00% 8.00% 8.00% 8.00% 8.00%

Difference -2.40% -1.47% -0.16% 1.80% 5.07% 11.60% 31.20% ∞

Return on Gross Assets 17.40% 17.40% 17.40% 17.40% 17.40% 17.40% 17.40% 17.40%


- 16 -



LCRT’S RESEARCH METHODOLOGY CONTRASTS LCRT’S RESEARCH METHODOLOGY CONTRASTS SHARPLY WITH THE TRADITIONAL VALUATION SHARPLY WITH THE TRADITIONAL VALUATION

APPROACHAPPROACH

Traditional Approach Forecasts 3-10 Years of

Cash Flows Applies Perpetuity or

Multiple for Terminal Value Discounts to Present

(“plan valuation”) Implicitly assumes the

structure and parameters of the terminal valuation are robust and accurate or “plugs” the parameters to explain current price

LCRT Methodology Employs only actual data

to empirically test robustness and accuracy of “spot intrinsic valuation” models and parameters

Extends the best models to use as terminal values in traditional “plan intrinsic valuations”

May eventually test the best models with forecast security analyst data


- 17 -



ACROSS A UNIVERSE OF 20,000ACROSS A UNIVERSE OF 20,000++ COMPANY- COMPANY-YEARS, TRACKING ERROR IMPROVEMENT YEARS, TRACKING ERROR IMPROVEMENT CHARTS MEASURE THE COMPARATIVE CHARTS MEASURE THE COMPARATIVE ACCURACY OF:ACCURACY OF: Models

Methodologies

Parameters

Tracking error equals the % absolute difference between the Model Intrinsic Value and the actual stock price at Fiscal Year + 3 Months

LOG2 of % Absolute Model Error versus

Actual Price - Fiscal Year +3 Months 1994-2002

63 79

Cumulative % of Universe


- 18 -



ADVANTAGES OF THE LCRT ADVANTAGES OF THE LCRT TRACKING ERROR IMPROVEMENT TRACKING ERROR IMPROVEMENT

CHARTSCHARTS Don’t depend on the distribution and don’t

assume Gaussian Normality

Cover the entire universe and all company years without “fudging” outliers with elimination or winsorization

Summarize very large amounts of time series and cross sectional data on one simple chart


- 19 -



LCRT BETA’S FOR THE RESIDUAL INCOME MODEL HAVE LCRT BETA’S FOR THE RESIDUAL INCOME MODEL HAVE SLIGHTLY BETTER TRACKING ERRORS THAN MEDIAN SLIGHTLY BETTER TRACKING ERRORS THAN MEDIAN

INDUSTRY BETA’S, BUT LESS THAN A 1%* DIFFERENCE EXISTS INDUSTRY BETA’S, BUT LESS THAN A 1%* DIFFERENCE EXISTS BETWEEN MEDIAN INDUSTRY BETA’S AND BETA’S = 1.00BETWEEN MEDIAN INDUSTRY BETA’S AND BETA’S = 1.00

LOG2 of % Absolute Model Error versus

Actual Price - Fiscal Year +3 Months 1994-2002

59 63

Cumulative % of UniverseThe low sensitivity of the tracking error results to employing the median industry beta over beta = 1.00 adds additional empirical evidence against the usefulness of CAPM beta as a measure of risk.

* 0.5% = 100%(63.4/63.1 -1)


- 20 -



EMPIRICAL EVIDENCE:EMPIRICAL EVIDENCE: LCRT’s MODEL IS 28-67% MORE ACCURATE THAN OTHER MODELS (at 50 LCRT’s MODEL IS 28-67% MORE ACCURATE THAN OTHER MODELS (at 50thth Percentile) and FREE CASH FLOW AND LCRT INTRINSIC VALUES PERFORM Percentile) and FREE CASH FLOW AND LCRT INTRINSIC VALUES PERFORM

THE BEST TO SEPARATE “WINNERS” FROM “LOSERS” THE BEST TO SEPARATE “WINNERS” FROM “LOSERS”

LOG2 of % Absolute Model Error versus Actual Price -

Fiscal Year +3 Months 1994-200247 63 79

67

Cumulative % of Universe

Sources:Financial Statements and Price Data – CapitalIQCalculations - LCRT’s PlatformConstant Dollar Gross Investment > $100 Million20,957 Company-Years; 1994-2002; Industrials

Sources:Financial Statements and Price Data – CapitalIQCalculations - LCRT’s PlatformConstant Dollar Gross Investment > $100 Million, Panel Data from 1994-2002, 17,697 Company-Years


- 21 -



USING THE LCRT DCF MODEL, 57% OF THE USING THE LCRT DCF MODEL, 57% OF THE APPROVED LIST IS OVER-VALUED AND 72% (100-APPROVED LIST IS OVER-VALUED AND 72% (100-28) IS MORE OVER-VALUED THAN THE UNIVERSE28) IS MORE OVER-VALUED THAN THE UNIVERSE

Any one of the following three hypotheses could be true:

– 1. Approved List may pick more “losers” than “winners”

– 2. OR Approved List may pick more “winners” than “losers”

– 3. OR Employing both the Approved List and LCRT’s valuation may pick more “winners” than “losers”

Empirical Tests to distinguish which Hypothesis of the three is most true

– Translate each dimension of the beliefs used to produce the Approved List into testable valuation models

– Back test each model’s accuracy and predictive capability

– Test combination of Approved List implicit valuation models and LCRT

Feedback from the empirical results will improve the stock selection PROCESS

Approved List

Universe

Median N Approved List -11.3 84Universe -6.2 1,000

57

28 = 100 – 72


- 22 -



NEXT GENERATION LCRT MODELNEXT GENERATION LCRT MODEL(About 70% Complete)(About 70% Complete)

Place all risk and other effects in the cash flows, so a singlesingle discount rate applies to ALLALL firms in each super sector for each year (22,000-115,000+ company-years 1994-2003)

Treat equity as an option on the cash economic returns, growth, capital structure, and restructuring potential of the operating assets (using only disclosed historical data to drive the model’s cash flow forecast)

– Refine company CER Fade-To’s and Fade Rates to reflect empirical realities across the entire CER spectrum from boundary to boundary, using option and tangent functions

– Refine asset growth fade rates to reflect market expectations– Employ an option pricing function to quantify the deadweight loss of

bankruptcy– Employ an option pricing function to describe equity and debt holder

pressure to sell assets– Divide the intrinsic value employing one uniform discount rate for all

firms into its debt and equity components– Describe start-up CER’s with an impulse function


- 23 -



FINAL THOUGHTSFINAL THOUGHTS

We practitioners and academics still have a long way to go to improve the robustness & accuracy of DCF valuation models and their predictive capability!– WSJ 9-29-04 page 1 article suggests new model where

donor demands 5 best scientists collaborate to find a cure much faster than individually (author’s picks – Tom, Joel, Bennett, Aswath, Ed Altman, RT)

Most people assume that tracking errors arise primarily from discontinuities between future cash flow forecasts and historical results. LCRT empirical results strongly suggest otherwise!

Stay tuned.


- 24 -



BACKUP SLIDESBACKUP SLIDES


- 25 -



THE RELATIVE WEALTH CHART STRATEGICALLY THE RELATIVE WEALTH CHART STRATEGICALLY COMPARES WEALTH CREATED TO CASH COMPARES WEALTH CREATED TO CASH ECONOMIC RETURNS (CER’S) ABOVE THE ECONOMIC RETURNS (CER’S) ABOVE THE DISCOUNT RATE AND TO ASSET GROWTHDISCOUNT RATE AND TO ASSET GROWTH


- 26 -



THE INTRINSIC VALUE CHART COMPARES PRICES THE INTRINSIC VALUE CHART COMPARES PRICES TO INTRINSIC VALUES TO DETERMINE IF THE TO INTRINSIC VALUES TO DETERMINE IF THE

FIRM IS UNDER OR OVER VALUEDFIRM IS UNDER OR OVER VALUED

Notice the large high low variation around the intrinsic valuations,

providing opportunity for profitable trading

The Absolute “Tracking” Error Intrinsic Value vs. Actual Equals the Absolute Geometric Mean Error

Between The Intrinsic Value Red Line And Closing Prices at Fiscal Year + 3 Months

(hollow circles) Over the Number of Years


- 27 -



THE PREDICTIVE INTRINSIC VALUE CHART THE PREDICTIVE INTRINSIC VALUE CHART ESTIMATES CONVERGENCE OF ACTUAL ESTIMATES CONVERGENCE OF ACTUAL

PRICE TOWARD INTRINSIC VALUEPRICE TOWARD INTRINSIC VALUE


- 28 -



THE COMPARATIVE INTRINSIC VALUE CHART COMPARES THE THE COMPARATIVE INTRINSIC VALUE CHART COMPARES THE TRACKING ERROR ACCURACY OF TWO MODELS TO TRACKING ERROR ACCURACY OF TWO MODELS TO

DETERMINE WHICH TO RELY ON FOR SELECTING STOCKSDETERMINE WHICH TO RELY ON FOR SELECTING STOCKS

LCRT

Residual Income


- 29 -



THE COMPARATIVE INTRINSIC VALUE THE COMPARATIVE INTRINSIC VALUE CHART ALSO QUANTIFIES VALUE EFFECTS CHART ALSO QUANTIFIES VALUE EFFECTS

OF NON-RECURRING ITEMS WHICH MAY OF NON-RECURRING ITEMS WHICH MAY NOT BE ONE TIME EVENTSNOT BE ONE TIME EVENTS


- 30 -



ADDING ANALYSTS’ SALES AND EPS ADDING ANALYSTS’ SALES AND EPS ESTIMATES ENABLES COMPARISONS OF ESTIMATES ENABLES COMPARISONS OF

PLAN VALUES FROM THE FORECASTS PLAN VALUES FROM THE FORECASTS AGAINST SPOT VALUES FROM HISTORYAGAINST SPOT VALUES FROM HISTORY


- 31 -



MOST ASSETS PRODUCE A NEARLY LEVEL USEFUL OUTPUT UNTIL FAILURE, INSTEAD OF THE STRAIGHT LINE OR THE

DECLINING BALANCE CURVE REFLECTING DEPRECIATED PLANT

Output

Time

(2) Most Assets Produce Nearly Level Output…

Until Failure

(1) Constant Output = Constant Dollar Level Annuity

Economic Life

(3) Straight Line Depreciation Net Plant

(4) Accelerated D

epreciation Net Plant

Failure (One Horse Shay)

(Economic Value Added Im

plicit Assumption)


- 32 -



CASH ECONOMIC RETURN REFLECTS THE CASH ECONOMIC RETURN REFLECTS THE AVERAGE INTERNAL RATE OF RETURN OF AVERAGE INTERNAL RATE OF RETURN OF

ALL THE PROJECTS IN PLACEALL THE PROJECTS IN PLACE

Cash Economic Return

Existing Projects


- 33 -



GLOSSARY OF KEY MEASUREMENT TERMSGLOSSARY OF KEY MEASUREMENT TERMS

Measurement Types – Maximum of 10 Years– Robustness - % of company years for each firm where the model

calculates a valid answer – Ideal is 100% Valid answers can be negative Rates of change (e.g. EPS growth rate) where the divisor is

negative are not calculated and treated as zero robustness for that year

– Accuracy – geometric mean % error between the model value and the actual value – averaged over 10 years maximum; non-robust company years are excluded; absolute error or signed error as separate measures – like golf, lower scores are better

Measurement Applications – Maximum of 10 Years– Explanatory Price Level: Model Price versus Actual Price– Explanatory % Capital Gain Return (CG) – Annual Model CG

versus Actual CG Concurrently – Fiscal Year -9 Months to Fiscal Year +3 Months to allow for disclosure lags’ effect on prices

– Predictive % Capital Gain (CG) – Annual Predicted Model CG versus Actual CG – Fiscal Year +3 Months to Fiscal Year +15 Months to allow for disclosure lags’ effect on prices


- 34 -



A BETTER EXPLANATORY MODEL IS A BETTER EXPLANATORY MODEL IS MORE PREDICTIVEMORE PREDICTIVE

Accuracy: % Explanatory Price Level

1009080706050403020100

Acc

ura

cy:

% P

red

ictiv

e C

ap

ital G

ain

100

90

80

70

60

50

40

30

20

10

0

R2 = 0.473N = 2,752

OLS

Least AbsoluteDeviation

LCRT Intrinsic Value Model

These results support the intuition of HOLT’s clients, who in the late 1980’s said, “I only employ the model for buy/sell decisions when it tracks well.”

“Few strokes separate the best from the worst professional golfers.”

Like golf, lower scores indicate more accuracy.

This chart violates the instantaneously efficient market hypothesis. It represents an “anomaly” consistent with behavioral finance theory.


- 35 -



LCRT ACKNOWLEDGEMENTS (1)LCRT ACKNOWLEDGEMENTS (1) “We all stand on the shoulders of giants” Consider receiving the insights and wisdom of radicals in the

minority Despite what the communists believe, property rights and democratic

rule of law matter (Adam Smith, Karl Marx, F.A. Hayek, Milton Friedman)

Despite what the socialists believe, high marginal tax rates seriously reduce freedom, incentives, and economic efficiency (Arthur Laffer)

Supply side matters – laws, rules, regulations, and structure significantly impact freedom and economic efficiency (Jude Wanniski)

Seek to understand anomalous empirical behavior, especially traditional outliers. Re-evaluate all your underlying assumptions (Thomas Kuhn)

The stock market may not be instantaneously efficient (Vernon Smith)


- 36 -



LCRT ACKNOWLEDGEMENTS (2)LCRT ACKNOWLEDGEMENTS (2) The Central Limit Theorem fails when the underlying distribution does

not possess a finite variance. Employ least absolute deviation instead of least squares (Robert Blattberg and Thomas Sargent)

Brownian motion fails. Draws from distributions are not independent. Where you start matters. Memory exists in the System and leads to Fat Tailed Stable Distributions instead of Gaussian Normal ones (Benoit Mandelbrot, Edgar Peters)

Replacing the Gaussian assumption with a Stable one eliminates option “volatility smiles” (Stanley Miles)

The CER Fade-To is not the same for all firms. This imperfect regression toward the mean relates to the starting position, memory in the System, and the underlying fat tailed Stable Distributions (Rawley Thomas)

Use an impulse function to describe the life cycle of a start-up firm (George Box, Gwilym Jenkins, Rawley Thomas)

The necessary specialization of all professions often breeds a myopic inability for all of us (the author included) to view the panorama of all the patterns affecting life and economic transactions between free individuals


- 37 -



LCRT PHILOSOPHICAL LCRT PHILOSOPHICAL RESEARCH INSIGHTSRESEARCH INSIGHTS

Get the Big Picture Right before Refining the Little Details (LCRT Included)

Better to be approximately correct than precisely wrong

Permutation production on panel data with binary search on three choices enables narrowing the range of reasonable model parameter values – shortens elapsed research time from years to weeks


- 38 -



EMPIRICAL VALUE AUDIT OF EMPIRICAL VALUE AUDIT OF VALUATION MODELSVALUATION MODELS

Translate conceptual intrinsic value models into testable intrinsic value models, using only disclosed historical data without any analyst overrides or interventions

Determine estimation procedures for all company years and model permutations Specify permutations of ranges of company drivers and model parameters to test Program model and planned permutations Select sample Measure robustness as % of company years where the models calculate versus all years Quantify accuracy as the % difference between model intrinsic values and actual stock prices Compare robustness and accuracy to other models with the cumulative % tracking error on the

whole sample Measure and compare predictive capability of the models to separate “Winners” from “Losers”

as prices oscillate around intrinsic values. Review distributions, deciles, and years. Refine model

– Identify anomalies and outliers of the models Graph model errors against key model drivers Over-sample tails of each value driver Review individual firms at the extremes and tails for data errors, algorithm

imperfections, and model structural problems– Correct errors, imperfections, and problems– Re-Test models


- 39 -



UNDER (OVER) INTRINSIC VALUATION VERSUS UNDER (OVER) INTRINSIC VALUATION VERSUS CER ILLUSTRATES EMPIRICAL VALUE AUDIT CER ILLUSTRATES EMPIRICAL VALUE AUDIT

ANALYSIS OF OUTLIERS AND PATTERNSANALYSIS OF OUTLIERS AND PATTERNSCER Fade-To

N = 26,018, 1994-2003Industrials, C$GI > $100M

R2 = 0.341

MSFT, Yahoo

AAPL, GY,AMESQ


- 40 -



PREFERRED EMPIRICAL SEQUENCE OF PREFERRED EMPIRICAL SEQUENCE OF PARAMETER DERIVATIONS IN DCF MODEL PARAMETER DERIVATIONS IN DCF MODEL

SPECIFICATIONSSPECIFICATIONS

Growth should be ranked 8th to 9th in conceptual and empirical importance behind other more important value drivers

Proper order for deriving parameters for key value drivers– Cash Economic Return (CER)– Uniform market derived discount rate for all companies for each year– CER Fade-To– CER Fade Rate– Gross Debt to Value of Existing Assets– Deadweight Cost of shifting asset ownership from equity to debt

holders - Bankruptcy– Market Value of Debt– Restructuring of Assets Potential– Growth Fade-To– Growth Fade-Rate


- 41 -



INDIVIDUAL SECURITY ANALYSIS NATURALLY OCCURS INDIVIDUAL SECURITY ANALYSIS NATURALLY OCCURS AFTER THE AFTER THE STRUCTURESTRUCTURE AND SUPPORTING AND SUPPORTING EMPIRICAL EMPIRICAL RESEARCHRESEARCH OF THE VALUATION MODEL SPECIFICATION OF THE VALUATION MODEL SPECIFICATION

Restatement of historical results to better reflect economic reality

Algorithms on such issues as operating lease capitalization, pension funds, post retirement benefits, and executive stock options

Override Assumptions

Forecast material discontinuities of past and future economic performance based on competitive strategic assessments

Compare “spot intrinsic values” from historical data and “plan intrinsic values” from forecasts to current price to quantify under or over valuation


- 42 -



SECURITY ANALYSIS PROCESS COMPARISONSSECURITY ANALYSIS PROCESS COMPARISONS

TRADITIONAL LCRT

Apply Empirically Validated LifeCycle Valuation Model and Market Derived Discount Rate to

historical statements to produce “Spot Intrinsic Values” for every year

Analyze historical statements to determine if they accurately reflect Cash Economic Returns on

operating assets and asset growth rates

Analyze historical statements to separate operating from financial cash flows; Compare accounting returns to CAPM costs of capital

Beat market benchmarks by purchasing under-valued firms and selling short over-valued firms

Forecast cash flows, often from margins and capital turnover

Apply perpetuity terminal valuation (NOT Empirically Validated) to forecast to produce

“Plan Intrinsic Value”

Forecast free cash flows, often from margins, capital turnover, and estimated CapEx

Apply LifeCycle terminal valuation to forecasts to create “Plan Intrinsic Value”

Compare “Plan Value” from forecast to current price

Compare “Spot Historical Intrinsic Values” and “Plan Intrinsic Value” from forecast to current

price


- 43 -



LCRT PERCEPTION:LCRT PERCEPTION:STRATEGIC POSITIONING AND EMPHASISSTRATEGIC POSITIONING AND EMPHASIS

ONONINDIVIDUAL FIRM SECURITY ANALYSISINDIVIDUAL FIRM SECURITY ANALYSIS

VERSUSVERSUSVALUATION AND STOCK SELECTION QUANTITATIVE MODELSVALUATION AND STOCK SELECTION QUANTITATIVE MODELS

0102030405060708090

100110

0 10 20 30 40 50 60 70 80 90 100 110

Valuation and Stock Selection Quantitative Models (%)

Ind

ivid

ua

l Fir

m S

ec

uri

ty

An

aly

sis

(%

)

Zacks, First Call

Morgan Stanley

Merrill Lynch

CSFB HOLT

LifeCycle Returns

SchwabRetail

“Sweet Spot” ?

Hypothesis: Moving from heavy individual security analysis in the upper left toward more model based valuations in the lower right may enable the security

analysis function to add more value and become more scaleable, IF the valuation model employed is

more robust, accurate, and predictive. Analysts spend more time on material valuation issues and

strategic forecasts.

© 2004 LifeCycle Returns, Inc. All Rights Reserved - 1 - VALUATION MODELS: ACCURACY AND PREDICTIVE CAPABILITY (A WORK-IN-PROGRESS) Financial Management.

Documents

financial data

lcrt free cash flow

financial statements

intrinsic value chart

enterprise value

value managements

concept of value management

historical data