SSRN 4864198

R E S E A R C H PA P E R
Should Your Stock Portfolio

Consider Your Career?
Philip Straehl
Chief Investment Officer, Americas
Robert ten Brincke

Director, Investment Planning and Optimization
Carlos Gutierrez Mangas

Senior Quantitative Researcher, Systematic Strategies
For Financial Professional and Institutional Investor Use Only. This document or its contents may not be made available to the general public or any other person without the prior written consent of
Morningstar Investment Management. Please see the disclosures for approved audience definitions.
In Memoriam:
Ari Schenker
We dedicate this paper to the memory of Ari Schenker, whose research was integral in shaping
the ideas presented in this study.
Abstract
We incorporate nontradable wealth, such as firm-level human capital and restricted stock units,
in a stock-level portfolio optimization and show that minimizing idiosyncratic risk at
the total portfolio level leads to a significant improvement in out-of-sample portfolio efficiency.
Based on data from 2013 to 2022, with a sample of 263 US firms, we find that total
wealth optimizations significantly reduce idiosyncratic portfolio risk, resulting in significant
improvements in risk-adjusted return at the total portfolio level compared with a cap-weighted
index. By contrast, simple company or sector exclusions are ineffective in improving total
portfolio outcomes. Additionally, we show that total wealth optimizations materially reduce
correlation with firm-level human capital and restricted stock units. Overall, our study
provides evidence that investors may reduce portfolio risk by customizing their stock portfolios
to account for the unique risk embedded in their nontradable wealth.
For financial advisor use only. Not for public distribution.

R E S E A R C H PA P E R SHOULD YOUR STOCK PORTFOLIO CONSIDER YOUR CAREER? 3
Direct indexing has taken the investment industry by storm. Cerulli (2022), a leading industry
research firm, projects direct-indexing assets to grow by 12.3% per year over the next five years,
significantly eclipsing other product categories such as exchange-traded funds and mutual
funds. In anticipation of this growth, major investment management firms have been racing
to acquire direct-indexing providers or build out their own direct-indexing offerings. In addition
to the well-established tax benefits of security-level tax-loss harvesting, the allure of direct
indexing is that it allows for personalization at the individual security level. While individually
tailored investment offerings have commercial appeal, direct-indexing solutions often lack robust
economic frameworks to help guide investors’ personalization decisions. Existing solutions leave
it up to the investors and their advisors to customize portfolios based on subjective preferences
around environmental, social, or governance) metrics; risk factors; or sector and security exclu-
sions, often without guidance on how these choices impact broader portfolio efficiency. While
the idea of personalizing a solution to an investor’s needs is conceptually appealing, it only
makes sense to customize a portfolio to the extent that it helps the investor reach his goals. In
other words, personalization is only a means to an end, not the goal, per se. Anecdotal evidence
suggests that only a minority of direct-indexing investors presently opt to personalize their
holdings, likely because of the lack of guidance investors receive on how to personalize their
portfolios. Deviating from market-cap weights is a form of active management, which means
personalization decisions should be grounded in rigorous economic frameworks, not based on
ad hoc assumptions.
In this paper, we develop a comprehensive theoretical framework to motivate security-level

portfolio customizations based on the unique risks embedded in an investor’s nontradable
wealth. Financial assets, like stocks and bonds, often only represent a small portion of an
investor’s total economic worth, with nontradable assets such as restricted stock units, human
capital, real estate, or other illiquid assets, such as ownership in a business, making up the bulk
of an investor’s net worth. These nontradable assets are exposed to unique risks that create a
hedging need in financial asset portfolios. For instance, an executive in an oil company has
significant exposure to the fortune of a single company (or its industry), such as unvested stock
grants and future bonuses, which need to be considered when building a financial asset
portfolio. Some of this risk may be mitigated by deviating from the direct-indexing portfolio’s
benchmark index and choosing a different mix of tradable assets. In this sense, when it comes
to choosing the optimal portfolio of financial assets, no portfolio is an island. That is, portfolio
choice must consider the unique risks embedded in the investor’s nontradable wealth.
So, what is this risk? Asset pricing theory (for example, Sharpe [1964], Treynor [1961], Lintner
[1965]) posits that investors should diversify idiosyncratic or nonmarket risks in their portfolios
because these nonmarket risks are not priced. In other words, the portion of the return of a
security that is not correlated with the pricing kernel of a given asset pricing model (for example,
CAPM beta) is not systematically rewarded with higher returns and therefore, in theory, does not
carry a “risk” premium. It follows that investors should look to diversify these idiosyncratic risks
in their portfolio. We note that our use of the term “idiosyncratic“ is informed by the asset
pricing literature rooted in the capital asset pricing model. As such, the risks we are seeking to

diversify are idiosyncratic in the sense that they are not “priced” in a CAPM sense, but these
idiosyncratic returns may still be correlated with other assets. For example, one would
expect the nonmarket return of two energy companies to be positively correlated, even if
this risk is not rewarded with a higher risk premium. Current direct-indexing solutions only
address the reduction of the risk of outside wealth indirectly by means of exclusions.
Beyond these theoretical reasons to diversify idiosyncratic portfolio risk, there are obvious
practical reasons to limit portfolio exposure to idiosyncratic factors such as the risk associated
with one’s company of employment. Consider the recent bankruptcy of Silicon Valley Bank, for
example. The Wall Street Journal1 reported that the SVB stock made up 18% of the retirement
plan assets of the SVB employee. SVB employees not only lost their jobs (human capital) but
also took a major hit to their net worth (financial capital). Clearly (with the benefit of hindsight!),
SVB employees’ total wealth, as measured by both their financial and human capital, was overly
exposed to the firm-specific risk of SVB, decimating the total economic worth of employees.
While employer-specific risk is hard to eliminate given the relative size of human capital in the
average investor’s total wealth balance sheet, a better diversified mix of financial capital
portfolio could have reduced the economic impact of SVB’s bankruptcy on employees’ total
economic worth.
In previous research, Blanchett and Straehl (2015) incorporated nontradable assets, such
as industry-specific human capital, real estate, and pension wealth, into a portfolio optimization
and provided evidence that an investor’s optimal asset-allocation mix varies significantly
based on the composition of the investor’s outside wealth. The study contributed to a growing
body of research (for example, Viceira 2001; Wallmeier & Zainhofer 2006; Ibbotson, Milevsky,
Chen, & Zhu 2007; Kyrychenko 2008; Blanchett & Guillemette 2019; Idzorek & Kaplan 2023) that
understands the portfolio choice problem in the context of an investor’s broader economic
worth. Yet, previous studies, including Blanchett and Straehl (2015) and Eiling (2013), focused
on portfolio implications at the broad asset-class or industry level. We focus on the idiosyncratic
portion of the portfolio choice problem and assume an appropriate systematic component (for
example, CAPM beta) has been determined exogenously. This assumption is reasonable in the
context of a typical financial advice process where an investor’s risk profile, along with the broad
asset mix, is often determined before the more granular security is defined.
Human capital represents the value of a person’s future earnings, which depends on a person’s
skills, occupation, age, industry, or employer. Human capital often makes up a significant
portfolio of total economic worth. For example, Heaton and Lucas (2000) estimated that human
capital makes up 48% of household wealth, whereas financial assets only represent 6.8%.
While previous research, such as Eiling (2013) and Blanchett and Straehl (2015), modeled human
capital at the industry level and explored the portfolio implications across broad asset-class
or industry portfolios, this paper studies human capital at the firm level and explores portfolio
implications at the stock level, extending the literature at a more granular level.
1 https://www.wsj.com/articles/when-svb and-first-republic-collapsed-so-did-their-employees-investments-59ce4715
Our focus on the portfolio impact of firm level is motivated further by the growing evidence that
labor income (that is, the return on human capital) is increasingly correlated with financial
assets. Eisfeldt, Falato, and Xiaolan (2022) estimate that stock-based compensation represents
36% of total compensation in high-skilled manufacturing jobs in recent years, up from 7% during
the 1980s. Eisfeldt et al. (2023) further report that 80% of stock-based compensation goes to
employees who are not high-ranking executives in the company, indicating that understanding
the impact of stock-based compensation on portfolio choice is relevant to employees below
the C-suite. Our sample indicates that stock-based compensation as a share of staff expense
rose from 4.2% in 2007 to 5.2% in 2022, peaking at 7.5% in 2021. This shift in compensation
policy impacts both the nontradable financial assets (for example, accumulated unvested stock
grants), as well as human capital. We develop a novel approach to modeling firm-level human
capital based on company-level compensation data and discount rates, which are among our
study’s central contributions.
….portfolio choice must consider the

unique risks embedded in the investor’s
nontradable wealth…
In the optimization section of our study, we contrast a novel total wealth optimization
framework that aims to minimize the idiosyncratic risk at the total portfolio level with simple
company and sector exclusions. The total wealth optimizations directly incorporate the
idiosyncratic correlation between an investor’s financial capital and nontradable wealth such as
restricted stock units and firm-level human capital. Our empirical analysis, covering the
period from 2013 to 2022 and comprising 263 US firms, shows that total wealth optimizations
significantly improve an idiosyncratic portfolio’s risk and risk-adjusted returns. We also
demonstrate that simple exclusionary approaches, such as company or sector exclusions, are
ineffective in diversifying portfolio risk. Overall, our study lends support to the idea that
investors may benefit from customizing their financial asset portfolio based on the unique
risks embedded in their nontradable assets.
Our paper starts with a discussion of our model of firm-level human capital and then
describes the optimization frameworks used in the study. Next, we present the results of our
empirical study of 263 US firms, contrasting the total wealth approach against sector and
company exclusions. Finally, we examine the efficacy of the optimization framework based on
the impact of the covid-19 lockdown on leisure firms during the onset of the global pandemic.

Modeling Firm-Level Human Capital
Intuitively, human capital is akin to a discounted dividend model, where a person’s expected
total compensation over their career is the stream of future dividends in the model. Broadly
speaking, total compensation is composed of cash wages, benefits (for example, health
insurance), and stock compensation. We adjust this expected compensation for mortality risk
and discount it using a firm-level rate. The set of equations (1), (2), and (3) articulates this idea:
HCi,t = Non.Stocki,t + Stocki,t (1)
R
q R-n w i,t (1+g) R-n
Non.Stocki,t = Σ
n (1+r i,t,R-n ) R-n (2)
Stocki,t= Stock.Pctgi,t-1 × Non.Stocki,t-1 × (1 + returni,t) (3)
Where:
HCi,t is the human capital of a person working for firm i at time t
Non.Stocki,t is the present value of mortality-adjusted wages
Stocki,t is the present value of the stock compensation
qR-n is the mortality weight for a given age
g is a nominal constant growth rate
wi,t is the level of wages
R-n is the number of years remaining in a work life
ri,t,R-n is the average yield on a 10-year maturity US zero-coupon bond plus the credit
default swaps estimated par spreads with tenor R-n
Stock.Pctgi,t-1 is the percentage of nonstock compensation that is delivered in stock
returni,t is the total return for stock i
To estimate human capital, we obtain data from four different sources: (i) we proxy wages using
average hourly earnings from the Bureau of Labor Statistics; (ii) we calculate stock compen-
sation as a percentage of staff expense using data from Compustat; (iii) the average 10-year
Treasury bond yield is from the Federal Reserve Bank of St. Louis Economic Data; and (iv) the
credit default swaps par spreads are from the Markit Credit Default Swaps database.

Wages
Through the Current Employment Statistics survey, the BLS compiles data on employment,
hours, and earnings. The average hourly earnings are available at the industry level using the
North American Industry Classification System, mostly since 2006. We chose these series to
proxy wages because average hourly earnings exclude stock compensation that would eliminate
the risk of double-counting any form of stock compensation in our estimate of human capital.
We work with 75 average hourly earnings time series, which are mapped either to a NAICS
two-digit or three-digit code. We link these time series to a firm using the firm’s NAICS code at
every point in time. We test for seasonality, and if found, we remove it.
For illustration, Figure 1 displays two of our 75 series: the leisure sector (both seasonally
and nonseasonally adjusted) and the utilities sector. We also include the most representative
inflation index—Consumer Price Index for All Urban Consumers—and the Employment Cost
Index. Except for the disruptions caused by the covid pandemic, labor costs are
fairly smooth.2
Figure 1: Inflation, Labor Cost, and Average Hourly Earnings Indexes

170
160 CPI
150 ECI
Index (December 2006=100)
140 Leisure_NSA
130 Leisure_SA
120 Utilities
110
100
90
12/2006 12/2008 12/2010 12/2012 12/2014 12/2016 12/2018 12/2020 12/2022
2 The labor market clears through fluctuations in quantity (that is, amount of labor hours) and not prices; consequently, prices have a smooth trend
driven by inflation and growth in real wages.
Table 1 presents the summary statistics for the average earnings series included.
Table 1: Summary Statistics: Inflation, Employment Cost Index, and
Average Hourly Earnings
Variable CPI ECI Avg Hourly Leisure Utilities
Earnings
N 193 193 14,475 193 193
N.Nan — — — — —
N.Obs 193 193 13,650 190 190
miss.Pctg 0.00% 0.00% 6.00% 2.00% 2.00%
N.Pos 179 193 12,483 162 178
N.Neg 14 — 1,167 28 12
N.Zero — — — — —
Min -2.00% 1.40% -11.10% -2.20% -2.80%
Mean 2.40% 2.70% 2.80% 3.00% 3.10%
Std.Dev 2.10% 1.00% 2.40% 3.80% 2.00%
Max 8.90% 5.70% 25.50% 18.60% 9.00%
P1 -1.40% 1.40% -2.90% -2.10% -2.00%
P2 -1.10% 1.40% -1.90% -1.70% -1.00%
P5 -0.10% 1.60% -0.70% -1.10% -0.20%
P10 0.20% 1.60% 0.20% -0.50% 0.40%
P20 1.10% 1.80% 1.10% 0.50% 1.30%
P50 2.00% 2.60% 2.70% 2.30% 3.10%
P80 3.50% 3.20% 4.50% 4.10% 4.80%
P90 5.30% 3.50% 5.70% 8.60% 5.60%
P95 7.40% 5.10% 6.80% 10.90% 6.20%
P99 8.50% 5.70% 10.10% 16.20% 7.20%
P98 8.30% 5.30% 8.60% 14.80% 7.00%
Sample: December 2006–December 2022

Stock Compensation
In the estimation of human capital, we use stock compensation as a percentage of staff

expense or its estimate when not available.3,4 The estimate for staff expense is the ratio of the
total staff expense for an industry k in year t divided by the total revenue for industry k and
year t multiplied by the total revenue of firm i.
Σ k,t Staff Expense
⟩
Staf f Expensei,k,t = Revenuei,k,t (4)
Σ k,t Revenue
×
To estimate stock compensation as a percentage of total staff expense when not available,
we follow a similar approach:
Σk,t Stock Compensation

⟩
Stock Compensationi,k,t = Revenuei,k,t (5)
Σk,t Staff Expense
×
In recent years, stock compensation has become a more prominent part of the compensation
package offered by firms (Table 2). Back in 2007, the average stock compensation as percentage
of staff expense was around 4.25%; that share has steadily increased since then to reach a
maximum of 7.5%. The drop observed in 2022 seems to be due to an incomplete dataset.5
3 The source is the annual Compustat file; the variables that define this ratio are stock compensation (stkco) and staff expense (xlr). Staff expense
represents salaries, wages, pension costs, profit sharing and incentive compensation, payroll taxes, and other employee benefits.
4 We use estimates, as Compustat’s coverage is limited for both staff expense and stock compensation. The coverage for stock compensation is
approximately 90% starting for the last years in the sample, and the coverage for total staff expense is around 33% of the Compustat dataset.
5 Notice the material drop in the number of firms.

Table 2: Trend in Stock Compensation
Stock Compensation
Year Number of Firms Percentage Staff Expense Percentage Revenue
Mean Median Mean Median
2007 406 4.20% 2.60% 1.10% 0.50%

2008 670 4.20% 2.30% 1.20% 0.50%
2009 679 4.10% 2.00% 1.30% 0.50%
2010 669 4.30% 2.00% 1.40% 0.50%
2011 657 4.20% 2.10% 1.30% 0.60%
2012 630 4.50% 2.20% 1.50% 0.60%
2013 615 4.50% 2.40% 1.60% 0.70%
2014 613 5.20% 2.70% 1.80% 0.80%
2015 608 5.30% 2.90% 2.00% 0.90%
2016 600 5.30% 3.10% 1.90% 1.00%
2017 589 5.70% 3.30% 2.30% 1.00%
2018 587 5.90% 3.40% 2.50% 1.00%
2019 584 6.10% 3.50% 2.50% 0.90%
2020 591 6.40% 3.50% 3.10% 1.00%
2021 527 7.50% 3.90% 4.40% 1.10%
2022 380 5.20% 3.70% 1.70% 1.10%
Empirical analysis supports the notion that stock compensation is both a reward–and a
risk-sharing mechanism. To illustrate this hypothesis, at every quarter in our sample, we sort
companies based on their return on assets, create deciles, and then produce the
average stock compensation through time. Figure 2 presents the median stock compensation
by profitability decile. Companies with very low profitability have very high stock compensation
(a mean stock compensation of around 11% and median stock compensation around 3.4%),
and highly profitable companies have high stock compensation (around 3.5% median compen-
sation and 7.5% average stock compensation).

Figure 2: Stock Compensation by Profitability

Median Stock Compensation (2007-2022) Mean Stock Compensation (2007-2022)
4% 12%
10
3
8
2 6
% Staff Expense
4
1
2
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Deciles by Return on Assets
Beyond profitability and using a similar approach, we find that companies with low earnings
yield, low leverage, and large market capitalization pay a greater proportion of their staff
expense in equity shares. (see Figure 3).
Figure 3: Relation Between Stock Compensation, Leverage, Earnings Yields, & Market Value
Mean Stock Compensation (2007-2022) Mean Stock Compensation (2007-2022)
12% 9%
8
10
7
8 6
5
6
4
3
% Staff Expense
4
2
2 1
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Deciles by Debt to Assets Deciles by Earnings Yield
Mean Stock Compensation (2007-2022)

7.0%
3
% Staff Expense
1
0
Deciles by Market Value (Size)

We conduct a regression analysis to more accurately quantify the impact of these drivers,
assess their statistical significance, and evaluate the collective influence of their correlations.
Our stock compensation model is as follows:
Stock Compensationi,t
(6)
= β0 + Industry FE + Year FE + β1 ROAi,t—1 + β2Returni,t—12 t
+ β3Mkt. Valuei,t+ β4Leveragei,t—1+ β5Earnings Yieldi,t+ εi,t
In this model, stock compensation for a given firm is a function of its industry sector, the stock
compensation trend (that is, year effects), its lagged profitability6, its lagged leverage, the
last-year stock returns, its equity market value, and its earnings yield. The results from this
estimation are in Table 3.
Table 3: Relation Between Firm Fundamentals, Time, Industry, and Stock Compensation
Term Estimate Standard Error t-Statistic p-value
Intercept 0.318 0.02 20.8 0%
Energy -0.049 0.02 -3.0 1%
Technology -0.131 0.02 -6.0 0%
Real Estate -0.152 0.02 -6.9 0%
Healthcare -0.157 0.02 -8.6 0%
Financials -0.179 0.01 -13.3 0%
Materials -0.189 0.01 -22.6 0%
Consumer Discretionary -0.208 0.01 -16.5 0%
Consumer Staples -0.215 0.02 -10.9 0%
Communication -0.225 0.01 -19.3 0%
Industrials -0.228 0.01 -27.3 0%
Lagged ROA -0.116 0.02 -6.5 0%
Lagged 12-months returns 0.002 0.00 0.4 71%
Mkt. Value (Size) 0.000 0.00 0.1 96%
Debt to Assets -0.117 0.03 -4.6 0%
Earnings Yield -0.014 0.01 -1.1 30%
Year Fixed Effects Y
Adjusted R-Squared 31%
N.Obs 9,405
Sample: 2008-2022
Source: Compustat & author’s calculations (estimations).
6 We lagged profitability and leverage by one fiscal year to avoid forward-looking bias.
There are two takeaways from the regression: First, two fundamentals have material and
statistical significance (ROA and leverage reduce stock compensation), and second, there is
significant heterogeneity in stock compensation across industries. Energy, a high-risk and
cyclical sector, has substantial stock compensation. This is also true of the information tech-
nology sector, where startups share their risk and future success with staff. On the other
hand, the stock compensation for the consumer staples sector (low risk and uncorrelated to
the business cycle) is 17% less than that of energy.
Discount Rates
Human capital’s risk is closely intertwined with the risk of the firm—growth opportunities,
prestige, ability to transfer to other positions, and the likelihood of experiencing a layoff depend
on a firm’s growth trajectory, its role in the industry, and its prestige. The likelihood of a firm’s
survival should be reflected in its credit spreads,7 which we use to discount the future nonstock
compensation flows. We source the credit spreads at the firm level from Markit’s Credit Default
Swaps database. In our model, for every firm in the sample, we need a spread for each year
of a person’s remaining working life.8 We encounter two challenges: First, Markit’s database
does not carry a complete and continuous set of tenors for every firm and date, and second,
there is incomplete coverage for the equity market of CDS spreads.
We address the missing tenors by interpolating with a smooth spline; at every date in the
sample, we pool the last 30 days of data for firms that have at least eight tenors available and
estimate a smooth spline. After estimating the smooth spline, we forecast the par spread
for 30 yearly tenors. Our CDS spread database contains around 52.5 million firm tenor-day
observations with 30 tenors per firm and 712 unique firms. The sample starts in June 2008 and
ends in December 2022. Figure 4 presents the estimated median spreads for the year-end
dates in our sample after interpolation.
7 Credit spreads may also embed liquidity but, by and large, should reflect the likelihood of a firm’s ability to meet its contractual obligation and
continue its normal operations.
8 Our baseline assumes a 45-year-old person who will retire at 65, which would imply a term structure with yearly tenors.
Figure 4: Estimated Median CDS Spreads (Sample Year-End Dates)
4.5% 12/31/2008
12/31/2009
4.0
12/31/2010
3.5 12/30/2011
12/31/2012
3.0
Estimated Median CDS Spread

12/31/2013
2.5 12/31/2014
12/31/2015
2.0
12/30/2016
1.5
12/29/2017
1.0 12/31/2018
12/31/2019
0.5
12/31/2020
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 12/31/2021
Tenor 12/30/2022
We address the second challenge by estimating spreads as a function of a firm’s credit risk.
To accomplish this for each tenor, we estimate a one-quarter rolling regression with the
par spread as the dependent variable and zi,t—a proxy to1the distance to default—as the
z = (ROAi,t + Mkt. Vali,t) ×
independent variable. i,t
σ(ROAi,t) (/ )
zi,t = (ROAi,t + Mkt. Vali,t) × (1/σ(ROA ))i,t

(7)
t
ROAi t
ROAi
ROAi,t = 1/2( Σ ) + 1/2( Σ ) (8)
i=t—4 4 i=t—12 12
t
ROAi t
ROA
Intuitively, zi,t is a standard
ROAi,t = 1/2(
score that tells us how) many
+ 1/2( Σ )
negative ROAi standard deviations Σ
4 12
we are away from default. Notice that zi,t has a negative relation with credit risk: the smaller it
i=t—4 i=t—12
is, the closer a firm is to default. ROAi,t is calculated with a quarterly frequency and is the
sum of a four-quarter average and a 12-quarter average with equal weights. Mkt. Vali,t is the
market value of equity as a percentage of total assets. σ(ROAi,t ) is the standard deviation
of the quarterly ROA series calculated using a 40-quarter rolling standard deviation. Then the
regression model to estimate the missing par spreads is:
spreadi,t, k = β0,k + β1,k i,t zi,t (9)

To make sure our model produces intuitive estimates, at every date we group zi,t into five
levels of risk as detailed in Table 4. Figure 5 displays the median spreads time series for the
one-year tenor. There are two observations regarding the grouping: First, the groups are
not equal-sized, and second, the credit risk is highly nonlinear. The increase in risk from very
low to low is materially smaller than the increase in risk from high to very high. The motivation
for the different size of the risk groups is because credit risk is a tail phenomenon. In the
credit risk spectrum, there is a large proportion of firms that bear very small or no credit risk;
it is only the very extreme of the credit risk distribution that captures most of the risk.
Table 4: Risk Groups by Z-Score and Observed Median Spread

Risk Z-Score Percentile Median Spread
Very High <5 1.92%
High 5—15 0.49%
Medium 15—30 0.19%
Low 30—80 0.16%
Very Low > 80 0.13%
Figure 5: 1-Year Tenor Median Spread by Risk Group

8
6
Log (Par Spread), Basis Points
Very High
5
High
4
Medium
3 Low
Very Low
2
0
03/06/2006 03/06/2010 03/06/2014 03/06/2018 03/06/2022
9 This same behavior is observed in the distribution of credit ratings. There is a very small proportion of entities that obtain the highest
credit rating (that is, AAA in the Moody’s scale), and there is also a very small proportion of entities that receive the worst credit rating (that is, C).
Facilities rated A, BBB, and BB account for most of the rated population.
As a final check on the intuitiveness of our framework, we calculate the correlations between
our estimate of human capital and the stock returns by level of stock compensation and
level of credit risk. Figure 6 presents the correlation heat map. Intuitively, human capital is a
portfolio with a substantial proportion in a bond that is defined by the nonstock compensation
and a smaller proportion in the stock of the company. Consequently, we should see that
a greater stock compensation implies a greater correlation with stock returns. Moreover, we
could expect that companies with high credit risk have greater correlation between stock
returns and human capital.10
Figure 6: Correlation Human Capital Changes vs. Stock Returns

(Full Dataset, Monthly Frequency)
Credit Risk Level Correlation
1 2 3 4 5 6 7 8 9 10 0.1
10 50% 52% 54% 56% 56% 60% 59% 58% 58% 59% 0.2
9 41% 48% 52% 54% 59% 59% 59% 57% 53% 28% 0.3
Stock Compensation Level
8 26% 34% 39% 47% 53% 55% 56% 53% 46% 23% 0.4
7 19% 24% 33% 41% 47% 51% 50% 48% 37% 11% 0.5
6 13% 19% 28% 32% 39% 42% 42% 39% 37% 30% 0.6
5 12% 14% 26% 32% 41% 46% 39% 38% 9% 8%
4 7% 18% 24% 32% 38% 36% 36% 38% 26% 19%
3 7% 14% 23% 29% 41% 41% 41% 34% 14% 12%
2 5% 12% 24% 31% 38% 41% 37% 10% 19% 2%
1 1% 10% 21% 24% 31% 34% 30% 27% 14% 5%
The heat map quantifies the correlation between human capital and a firm’s stock returns—it
is the most important driver. The relation between credit risk and correlation is more complicated;
it may seem counterintuitive that for the highest levels of credit risk, the correlation is
low compared with lower levels of credit risk. The cause for this pattern is the relation between
standard deviation of stock returns and credit risk; a correlation is the covariance between
two series divided by their respective standard deviations, and, as the credit risk increases, the
standard deviation of a stock return grows faster than the covariance between stock and
human capital driving down the correlation. Now that we have a complete framework, we can
posit the circumstances that will deliver the greatest benefits for an optimized total wealth
10 As the risk of default increases, fixed-income securities exhibit greater correlation to common stock, which is the lowest-rank security in the capital
structure and the first whose value is wiped out in the event of a default.
approach. These include firms that have a high level of stock compensation, risky startups,
companies with: a high level of idiosyncratic risk, volatile industry sectors, and companies with
material credit risk.
Optimization Framework
Any optimization in this paper pertains to a broker account and one or two additional accounts:
an RSU account and a human capital account. The broker account is the primary focus of the
optimization; it is the account to which changes are made whenever the portfolio is rebalanced.
We assume that the broker account houses the individual shares used to create the direct-in-
dexing portfolios. Conversely, the RSU and human capital accounts consist only of nontradable
assets. The assets in these accounts may or may not play a role in deciding the broker account’s
portfolio, but the assets in them remain untouched by the optimization.
The broker account is the only account in which the composition may be adjusted, so
its size compared with the other accounts is a major factor. We choose the sizes of the accounts
to approximate the situation of a 45-year-old investor of medium income. We arrive at
this representative total wealth composition based on an individual whose annual income
of USD 100,000 at age 25 grows at 1% (real) until the retirement age of 65. The individual is
assumed to earn 15% of their income in stock-based compensation with a five-year vesting
period. Human capital is the present value of the 45-year-old’s future savings discounted
at a 1% inflation-adjusted interest rate and based on a savings rate of 10%.
Table 5: Accounts Relevant to the Optimization
Age Tradable? Size in % $ Amount

Broker Yes 45 $270k
RSU No 15 $90k
Human Capital No 40 $240k
The optimizations come in three flavors: island, exclusion, and total wealth. All of these are
direct-indexing strategies and rely on the presence of a benchmark index. We have chosen the
Morningstar US Target Market Exposure (TME) Index as the benchmark index, which represents
a substantial portion of the large- and mid-cap US stocks. It is therefore assumed that our
hypothetical investor’s risk/return preferences are aligned with the TME Index.

Island
The island optimization is simplest and serves as a baseline. It minimizes the tracking error for
the broker account without consideration for the RSU and human capital accounts.
The objective function is shown in the equation (10) below, where w is a vector of portfolio
weights, b is a vector of benchmark index weights, T denotes the transpose, Σ is the asset-by-
asset covariance matrix, and 1 is a vector of ones. The constraints ensure that all weights
sum to 1 and that no shorting is allowed. Without any additional constraints, the optimal portfolio
is such that w = b.
minw (w — b)T ∑(w — b) (10)

s.t. wT 1 = 1
wi ⩾ 0
The island optimization reflects an investor who chooses a direct-indexing benchmark index
without consideration for any outside sources of wealth or any additional constraints.
Exclusion
Many direct-indexing solutions offer the exclusion of an asset or a group of assets as

customization. We consider two common exclusion variants: company and sector. Each of these
is a refinement of the island strategy above and introduces just one additional constraint.
For a company exclusion, a single asset is forcibly excluded from the solution by setting wi = 0
for the company of interest. An investor working for Apple would exclude Apple from its
broker account, despite its presence in the TME Index. With only imperfect substitutes available,
the resulting tracking error is now no longer 0%.
A sector exclusion is much broader and excludes the sector in which the investor’s company
is active. It relies on a vector s where an entry is 1 if asset i operates in the sector of interest and
0 otherwise. The constraint then requires that s Tw = 0. It removes any asset active in the
GICS sector of interest from the portfolio by forcing its weight to 0. The sector of interest is the
sector in which the company of interest operates: An investor working for Apple would exclude
the information technology sector.
These exclusion variants mitigate some of the risk stemming from idiosyncratic components and
reduce the risk associated with one company or the sector in which it operates. But particularly,
the sector exclusion variant is sufficiently broad that it may affect the resulting broker portfolio in
other ways, too. Note also that the company of interest may not be in the chosen index,
rendering the company exclusion variant useless.

Total Wealth
The narrowly focused optimizations presented above represent typical offerings by direct-in-
dexing providers. None of these considers outside sources of wealth explicitly, though the
exclusion variants do so implicitly. An explicit consideration of outside wealth is the goal of the
total wealth optimizations presented here.
The total wealth optimization aims to minimize the variance (rather than the tracking error)
for all three accounts simultaneously. But of course, such an optimization would simply lead
to a minimum variance portfolio, which is likely incompatible with the investor’s risk/return
preferences, which we assume has been determined exogenously. Because of our focus on
idiosyncratic risk, we constrain the systematic risk level to remain in line with the systematic
risk of the benchmark. We take CAPM’s (predicted) β as the systematic component and
require that the resulting portfolio’s β remains unchanged, where the TME Index itself serves
as a proxy for the market. The objective function used here is similar in spirit to the one
used by Blanchett and Straehl (2015) as the asset-class level. Blanchett and Straehl minimize
the variance of total wealth for a given level of expected return.
Note that the vector w now encompasses all accounts rather than just the broker account. A
matrix M is a matrix with M(i,i) = 1 if asset i is tradable and 0 otherwise. M is used to mask11 the
beta vector and covariance matrix in the constraints (and so these constraints only apply to
the broker account assets): βM = Mβ and ΣM = M Σ M . The RSU and human capital account
weights are dictated by their fixed initial position weights f.
minw wT ∑w (11)
s.t. wT βM = 1 broker account beta constraint
(w – b)T ΣM(w – b) ⩽ ω broker account tracking error constraint
wT 1 = 1
w i= f i if i is untradable
wi ⩾ 0
The predicted β of a portfolio is given by bb Σw

T
Σb and is therefore derived from the covariance

T
matrix. To calculate the predicted β for a single asset as required for the vector β, set the vector
w such that there is a single entry 1 for the asset in question (and 0 in any other position).
Note that we also require that the tracking error remains below a certain maximum level ω, for
which we have taken levels of 1%, 3%, and 5%. These different levels represent the three variants
of the total wealth optimizations. This guardrail constraint is a tacit acknowledgement that there
are limits to the extent to which the covariance matrix captures the riskiness of assets. The tracking
error constraint is not linear, and the resulting optimization is no longer a quadratic program.
11 M does nothing but zero out values in β and Σ for untradable assets.
Although perhaps not immediately obvious, this problem formulation is related to the minimiza-
tion of tracking error. Consider an asset pricing model rp = α + βrb + ϵ, where rp represents
the portfolio return, β the beta with the benchmark b, rb the benchmark return, and ϵ the
idiosyncratic return. Minimizing the tracking error ω means minimizing active variance, which
amounts to minimizing the variance of rp – rb= α+(β – 1) rb + ϵ.
minw Var(a + (β – 1)rb + ϵ) (12)

Var((β – 1)rb + ϵ)
(β – 1)2Var(rb) + Var ϵ) + 2(β – 1)Cov(rb ,ϵ)
(β – 1)2Var(rb) + Var ϵ)
This breaks the minimization of the tracking error into two components: a systematic component
and an idiosyncratic component. A tracking error of 0% is obtained only if β = 1. Under this
model, the minimization of total variance in the presence of a β = 1 constraint is equivalent
to the minimization of the tracking error. Absent any outside wealth, this minimizes the broker
account’s tracking error with the benchmark.
But how does outside wealth affect the total variance? Consider the outside wealth
ro = αo+ βorb + ϵo. The problem minimization of variance amounts to the minimization
of variance across two accounts Var (rp + ro). By expanding this and using the fact that
βp = 1 and Cov(rb ,ϵ) = 0, this expression can be simplified.
Var(αp + βprb + ϵp + αo + βorb+ ϵo) (13)

= (1 + 2βo+ βo2 )Var(rb) + Var(ϵp) + Var(ϵo) + 2Cov(ϵp ,ϵo)
Note that the terms βo, Var(rb), and Var(ϵo) are fixed because they come from outside wealth
and therefore play no role in variance minimization. The only relevant terms are the idiosyncratic
variance in the broker account Var (ϵp) and the covariance of the idiosyncratic components
Cov(ϵp ,ϵo). In other words, the minimization of variance under this βp= 1 constraint exploits
the correlation between the idiosyncratic components of wealth in both accounts. Without such
correlation, the outside wealth does not affect the optimal allocation in the broker account.12
12 This may be shown through simple examples. Consider a broker account with uncorrelated assets x and y that is optimally weighted in terms of
minimum variance. Now add outside wealth in the form of only x. As viewed over the total wealth, the inclusion of additional x means there is
now too much x, and the broker account must adapt accordingly. Alternatively, consider that the outside wealth is some asset z that is uncorrelated
with x or y. Without covariance terms, the total variance is just the sum of variances, but z is untradable and overall variance is still minimized
by minimizing the broker account’s variance. Therefore, an uncorrelated z does not change the optimal allocation within the broker account and is
irrelevant to the minimization of total wealth variance.
Under the pricing model above, the idiosyncratic terms for any pair of individual assets may still
be correlated due to underlying factors that are “unpriced” (because an investor may diversify
them away), but nevertheless explain the joint movement of assets. An example is sector or
industry membership. The formulation above therefore aims to lock in the “priced” part of the
portfolio and reduce the idiosyncratic variance present due to the correlation between the
outside wealth asset and the “idiosyncratic” risk of the broker account assets.
Risk Model
The importance of tracking error in all these optimizations, as well as the presence of
a predicted β constraint in the total wealth variant, puts a heavy burden on the covariance
matrix Σ. Σ comes directly from the Axioma US4 Medium Horizon factor model. We
understand that this model is widely used in practice and covers a significant fraction of the
US equity asset universe.
To run optimizations for human capital assets, we must extend the risk model to include them.
To do so, we use the change in human capital as the human capital return. A constrained
linear regression then minimizes the sum of squares between the returns13 of the human capital
asset and its equity counterpart and estimates a multiplier 0 ≤ mi ≤ 1. The equity counterpart—
Apple’s human capital would have Apple shares as its proxy—therefore serves as a proxy
asset, and the human capital asset’s risk is some multiple m of the proxy’s risk. The values of
the resulting proxy multipliers depend on the chosen time window, but in recent years, these
range from almost 0 for some assets to 0.75 for others, and average approximately 0.15 to 0.20.
Human capital assets are therefore composite assets in the risk model, consisting of mi parts
equity proxy and 1 – mi parts cash.
13 Daily returns were obtained from CRSP and FactSet.

Total Wealth Optimization Results
Having defined the total wealth optimization frameworks applied in this paper, we move to
testing the efficacy of the optimization framework based on historical data. Specifically,
we run optimizations from January 2013 through December 2022, reconstituting the portfolio
quarterly.14 Our sample includes 263 firms that had continuous membership in the Morningstar
US Total Market Exposure Index over the analysis period and for which we have human capital
data based on the methodology described above. See the firm-level human capital summary
statistics for the 263 firms in our empirical analysis sample in Table 6.
Table 6: Human Capital Summary Statistics for Total Wealth (263 Firms) Sample
Stock Return Human CDS Spread CDS Spread Z-Score Z-Score (Δ%)
Capital (Δ%) (Δ%)
Stock Return 15.2% -20.3% -0.4% 0.6% 79.7%
Human Capital (Δ%) 15.2% -76.8% -4.5% 3.1% 26.7%
CDS Spread (Δ%) -20.3% -76.8% 2.5% -1.4% -39.9%
CDS Spread -0.4% -4.5% 2.5% -89.2% -0.9%
Z-Score 0.6% 3.1% -1.4% -89.2% 0.9%
Z-Score (Δ%) 79.7% 26.7% -39.9% -0.9% 0.9%
We evaluate the efficacy of the total wealth optimization framework with regard to the bench-
mark (that is, island) and the company and sector exclusions. We consider two types of outside
wealth scenarios:
• RSU Only: We assume that the individual’s outside wealth is composed of restricted stock
units of the company of employment.
• RSU & Human Capital: We incorporate both human capital and RSUs as the nontradable
outside wealth assets.
The optimizations are performed based on the Axioma Portfolio Optimizer. Assets are rolled over
to the next reconstitution date by applying the asset’s period returns15 to the positions using
returns data available in the APO. The returns of the human capital asset is equal to its growth
rate. We start the backtest based on the wealth proportions highlighted above and
allow the relative portions to change over the course of the 10-year period based on their total
return. Because human capital grows at a slower rate than stocks (the TME Index had an
average return of 12.4% over the period), the human capital asset proportion for the average firm
declines on a relative basis (human capital on average grows by 3.9% in our sample).
14 Each reconstitution occurs on the last trading day of March, June, September, and December.
15 Accounting for certain corporate actions such as (reverse) splits, though not spinoffs. A preemptive sell was used for assets that are no longer listed
at the next reconstitution date.
Table 7: Total Return Summary Statistics: Combined RSU and Broker Portfolio
Total Wealth (TE<1%) Total Wealth (TE<3%) Total Wealth (TE<5%) Company Exclusion Sector Exclusion Island
Return Summary Stats
Arith. Return (Ann) 12.64% 12.67% 12.66% 12.61% 12.49% 12.61%
Geo. Return (Ann) 11.81% 11.85% 11.83% 11.76% 11.62% 11.76%
StDev (Ann) 16.12% 16.13% 16.16% 16.21% 16.24% 16.22%
Max DD -24.44% -24.47% -24.50% -24.46% -24.63% -24.46%
Skew (0.90) (0.85) (0.85) (0.92) (0.92) (0.92)
Kurt 2.17 2.15 2.16 2.19 2.21 2.19
Sharpe Ratio 0.79 0.79 0.79 0.78 0.77 0.78
Calmar Ratio 0.53 0.53 0.53 0.53 0.52 0.53
Corr. (Broker/RSU) 55.27% 53.72% 53.46% 57.17% 56.37% 57.33%
Diff. Relative to Island Portfolio
Arith. Return (Ann) 0.03% 0.07% 0.05% 0.00% -0.12%
Geo. Return (Ann) 0.05% 0.10% 0.07% 0.00% -0.14%
StDev (Ann.) -0.10% -0.09% -0.06% -0.01% 0.02%
Max DD 0.02% -0.01% -0.04% 0.00% -0.17%
Skew 0.02 0.07 0.07 (0.00) (0.00)
Kurt (0.02) (0.04) (0.03) 0.00 0.02
Sharpe Ratio 0.01 0.01 0.00 0.00 (0.01)
Calmar Ratio 0.00 0.00 0.00 (0.00) (0.01)
Corr. (Broker/RSU) -2.06% -3.61% -3.86% -0.16% -0.96%
T-Stat Relative to Island Portfolio
Arith. Return (Ann) 1.81 1.98 1.29 (0.77) (5.64)
Geo. Return (Ann) 2.85 2.47 1.61 (0.15) (5.81)
StDev (Ann) (6.96) (3.17) (1.94) (5.03) 3.55
Max DD 0.60 (0.11) (0.69) 1.88 (6.09)
Skew 6.47 10.05 9.97 (6.31) (2.32)
Kurt (1.68) (2.28) (1.51) 7.11 2.34
Sharpe Ratio 4.18 2.56 1.58 4.47 (5.92)
Calmar Ratio 1.09 1.19 0.59 (0.46) (6.32)
Corr. (Broker/RSU) (27.60) (21.88) (20.04) (8.36) (11.41)
Note: The table shows the total return summary stats for total wealth portfolios that are composed of RSUs and broker portfolios across 263 firms. In the total wealth optimizations,
the broker account is optimized to minimize the variance of total wealth subject to a broker account level beta = 1 and varying (that is, 1%, 3%, and 5%) tracking error constraints.

RSU-Only Scenario
In the first optimization scenario, we consider the case of an individual with concentrated
positions in a restricted stock unit. For the 263 firms in our sample, we assume that the individual
has USD 90,000 in restricted stock and USD 270,000 in the broker account (that is, direct-in-
dexing portfolio) at the beginning of the period.
Table 7 shows the summary statistics for the optimized portfolios alongside the exclusions and
compares the results with the results of the island portfolio for our 10-year sample period. The
historical analysis shows that the total wealth optimized portfolios exhibit favorable return and
risk statistics. The total wealth portfolios with 1% and 3% tracking error constraints have
statistically significantly higher geometric returns and lower standard deviations compared with
the island (that is, TME Index) portfolio. The combined improvement and risk and return also
leads to statistically significant improvement in the Sharpe ratios. The 5% tracking error
constrained version shows directionally similar improvements, which are not significant,
suggesting that the greater leeway the optimizer has does not translate into out-of-sample
improvements in risk-adjusted returns. By comparison, the company and sector exclusions are
less advantageous in diversifying the risk of the RSU account. While the company exclusion
leads to a statistically significant reduction in standard deviation of 1 basis point, the impact is
not economically meaningful. Sector exclusions, on the other hand, lead to a statistically
significant deterioration in risk-adjusted returns as shown by the increased standard deviation
and lower returns, suggesting that removing an entire economic sector is not a useful tool to
diversify portfolio risk. Finally, Table 7 shows that both the total wealth and exclusion-based
optimization portfolio led to statistically significantly lower correlations with the RSU stock. The
correlation reduction for the total wealth portfolios is both economically and statistically
meaningful, declining monotonically from 2.06% for the 1% constraints to 3.86% for the 5%.
Overall, the total return statistical highlights that the optimizing concentration risk directly leads
to better risk-adjustment returns. While the total portfolio return lens is important, the focus of
our paper is to reduce idiosyncratic portfolio risk, which is shown in the next section.
Table 8 shows the summary returns based on the market beta-adjusted returns,16 which allows
one to more directly observe the diversification benefits the adjusted portfolios provide in
relation to investing in a benchmark. Through this lens, the risk-adjusted return improvements
from the total wealth optimization approach are more meaningful, especially as it relates to the
risk side. Notably, the improvement in (idiosyncratic) max return drawdown is material at 1.11%
on average for the 1% tracking error case and 1.66% on average for the 3% tracking error case.
Again, we observe a drop-off in efficiency beyond the 3% tracking error point.
16 We use the predicted market beta estimate from the Axioma risk model to calculate the beta-adjusted returns.
Table 8: Idiosyncratic Return Summary Statistics: Combined RSU and Broker Portfolio
Arith. Return (Ann) 0.39% 0.43% 0.41% 0.36% 0.25% 0.36%
Geo. Return (Ann) 0.26% 0.30% 0.28% 0.22% 0.11% 0.22%
StDev (Ann) 4.69% 4.67% 4.67% 4.89% 4.88% 4.91%
Max DD -13.50% -12.94% -13.00% -14.55% -14.96% -14.60%
Skew (0.10) (0.10) (0.09) (0.07) (0.08) (0.07)
Kurt 0.51 0.53 0.55 0.48 0.47 0.48
Sharpe Ratio 0.03 0.05 0.05 0.02 0.00 0.02
Calmar Ratio 0.06 0.07 0.07 0.05 0.04 0.05
Corr. (Broker/RSU) -0.99% -2.59% -2.84% 1.24% 0.43% 1.43%
Arith. Return (Ann) 0.03% 0.07% 0.05% 0.00% -0.11%
Geo. Return (Ann) 0.04% 0.08% 0.06% 0.00% -0.11%
StDev (Ann) -0.22% -0.24% -0.23% -0.02% -0.02%
Max DD 1.11% 1.66% 1.60% 0.06% -0.36%
Skew (0.02) (0.02) (0.02) 0.00 (0.00)
Kurt 0.04 0.06 0.08 0.00 (0.00)
Sharpe Ratio 0.01 0.03 0.03 0.00 (0.02)
Calmar Ratio 0.00 0.01 0.01 0.00 (0.01)
Corr. (Broker/RSU) -2.41% -4.02% -4.27% -0.19% -1.00%
Arith. Return (Ann) 1.91 2.08 1.39 (0.67) (5.34)
Geo. Return (Ann) 2.39 2.25 1.51 0.04 (5.24)
StDev (Ann) (16.28) (9.42) (8.33) (10.37) (1.61)
Max DD 10.77 8.16 7.37 8.81 (3.12)
Skew (2.47) (1.43) (1.10) 0.79 (0.31)
Kurt 1.35 1.13 1.45 1.23 (0.23)
Sharpe Ratio 3.05 4.35 4.11 1.68 (4.09)
Calmar Ratio 1.04 2.50 2.51 1.41 (3.54)
Corr. (Broker/RSU) (26.05) (20.35) (18.84) (7.87) (9.55)
Note: The table shows the idiosyncratic return summary stats for total wealth portfolios that are composed of RSUs and broker portfolios across 263 firms. In the total wealth
optimizations, the broker account is optimized to minimize the variance of total wealth subject to a broker account level beta = 1 and varying (that is, 1%, 3%, and 5%) tracking error
constraints.
The total wealth optimization still significantly eclipses the simple exclusionary approaches
based on company and sector exclusions. The latter, as is the case in the total return case,
exhibits statistically significant deterioration relative to investing in the TME Index (that is,
island) portfolio, suggesting that investors seeking to diversify concentrated stock positions are
better off investing in the market-cap-weighted index than removing the economic sector
wholesale. The correlation benefits, measured as a correlation between the broker return and
the nonmarket return of the RSU stock, are also more pronounced for the idiosyncratic
return case. We observe a 4.27% reduction in the correlation coefficient for the 5% tracking
error optimizations, which is both statistically and economically meaningful.
RSU & Human Capital Scenario
In this section, we consider a scenario where an individual has nontradable assets in restricted
stock units (USD 90,000) along with firm-level human capital (USD 240,000). The total wealth direct-
indexing portfolio (USD 270,000) is constructed to minimize the variance of total wealth and
compared with the island portfolio and the simple exclusions-based approach.
Table 9 shows the total returns summary statistics for the combined human capital, RSU,
and direct-indexing portfolio for the 10-year sample period. The results are comparable to the
RSU-only case, where we observe marginally higher returns along with lower standard
deviation for the total wealth portfolios compared with the island portfolio. We see a statistically
significant improvement in the Sharpe ratio for the 1% and 3% tracking error case, with the
5% tracking error case showing a less pronounced increase in risk-adjusted return. The average
impact of the company exclusions is immaterial with the return and risk statistics matching the
benchmark, even as some of the small deviations are statistically significant. Sector exclusions
lead to suboptimal risk-adjusted return outcomes in relation to the benchmark, notwithstanding
the statistically significant reduction in correlations with the human capital and RSU assets. That
said, the total wealth optimizations exhibit lowest correlations against the RSU and human
capital assets, decreasing monotonically with higher tracking errors. The market beta-adjusted
returns, which isolate the idiosyncratic component of returns from the systematic ones, are
shown in Table 10.

Table 9: Total Return Summary Statistics: Combined Human Capital, RSU, and Broker Portfolio
Arith. Return (Ann) 9.90% 9.91% 9.90% 9.88% 9.79% 9.88%
Geo. Return (Ann) 9.36% 9.37% 9.35% 9.32% 9.23% 9.32%
StDev (Ann) 12.92% 12.93% 12.97% 12.98% 12.98% 12.98%
Max DD -20.13% -20.27% -20.32% -20.04% -20.14% -20.04%
Skew (0.90) (0.86) (0.85) (0.92) (0.92) (0.92)
Kurt 2.55 2.51 2.51 2.56 2.58 2.56
Sharpe Ratio 0.78 0.78 0.77 0.77 0.76 0.77
Calmar Ratio 0.51 0.50 0.50 0.51 0.50 0.51
Corr. (Broker/HC) 7.86% 6.88% 6.69% 8.58% 8.28% 8.63%
Corr. (Broker/RSU) 55.24% 53.20% 52.77% 57.14% 56.34% 57.30%
Arith. Return (Ann) 0.03% 0.03% 0.02% 0.00% -0.09%
Geo. Return (Ann) 0.04% 0.05% 0.02% 0.00% -0.10%
StDev (Ann) -0.06% -0.05% -0.02% 0.00% 0.00%
Max DD -0.09% -0.23% -0.28% 0.00% -0.10%
Skew 0.01 0.06 0.07 (0.00) (0.00)
Kurt (0.01) (0.05) (0.05) 0.00 0.01
Sharpe Ratio 0.01 0.01 0.00 0.00 (0.01)
Calmar Ratio (0.00) (0.01) (0.01) 0.00 (0.01)
Corr. (Broker/HC) -0.77% -1.75% -1.93% -0.05% -0.35%
Corr. (Broker/RSU) -2.06% -4.10% -4.53% -0.15% -0.96%
Arith. Return (Ann) 2.00 1.09 0.51 (0.53) (5.64)
Geo. Return (Ann) 2.65 1.34 0.60 0.03 (5.76)
StDev (Ann.) (6.75) (2.44) (0.70) (6.08) (0.31)
Max DD (3.68) (4.55) (5.00) 3.44 (4.82)
Skew 5.25 10.88 11.16 (5.32) (0.81)
Kurt (1.30) (2.88) (2.37) 6.02 1.71
Sharpe Ratio 4.59 2.08 1.01 5.01 (5.56)
Calmar Ratio (1.63) (2.91) (3.46) 1.27 (6.01)
Corr. (Broker/RSU) (11.44) (11.04) (10.12) (5.59) (4.65)
Corr. (Broker/RSU) (27.59) (23.09) (21.21) (8.36) (11.41)
Note: The table shows the total return summary stats for total wealth portfolios that are composed of human capital, RSUs, and broker portfolios across 263 firms. In the total
wealth optimizations, the broker account is optimized to minimize the variance of total wealth subject to a broker account level beta = 1 and varying (that is, 1%, 3%, and 5%)
tracking error constraints. Performance is gross of fees. If fees were added, results would have been lowered.

Table 10: Idiosyncratic Return Summary Statistics: Combined Human Capital, RSU, and Broker Portfolio
Arith. Return (Ann) 0.90% 0.92% 0.90% 0.88% 0.80% 0.88%
Geo. Return (Ann) 0.70% 0.71% 0.70% 0.67% 0.60% 0.67%
StDev (Ann) 5.50% 5.49% 5.51% 5.62% 5.61% 5.63%
Max DD -14.07% -13.83% -13.89% -14.70% -14.95% -14.74%
Skew (0.19) (0.12) (0.10) (0.19) (0.20) (0.19)
Kurt 1.76 1.52 1.50 1.79 1.81 1.78
Sharpe Ratio 0.12 0.13 0.14 0.11 0.10 0.11
Calmar Ratio 0.09 0.10 0.11 0.08 0.08 0.08
Corr. (Broker/Idio.HC) -19.14% -19.86% -19.97% -18.49% -18.71% -18.44%
Corr. (Broker/Idio.RSU) -0.99% -3.00% -3.38% 1.24% 0.43% 1.43%
Arith. Return (Ann) 0.02% 0.04% 0.02% 0.00% -0.08%
Geo. Return (Ann) 0.03% 0.04% 0.03% 0.00% -0.08%
StDev (Ann) -0.13% -0.13% -0.12% -0.01% -0.02%
Max DD 0.66% 0.91% 0.85% 0.03% -0.21%
Skew (0.00) 0.07 0.09 (0.00) (0.01)
Kurt (0.01) (0.26) (0.28) 0.01 0.03
Sharpe Ratio 0.01 0.03 0.03 0.00 (0.01)
Calmar Ratio 0.01 0.02 0.02 0.00 (0.01)
Corr. (Broker/Idio.HC) -0.70% -1.42% -1.53% -0.05% -0.27%
Corr. (Broker/Idio.RSU) -2.42% -4.42% -4.81% -0.19% -1.00%
Arith. Return (Ann) 2.04 1.40 0.79 (0.46) (5.59)
Geo. Return (Ann) 2.50 1.47 0.81 0.26 (5.49)
StDev (Ann) (15.00) (6.31) (4.89) (12.14) (1.58)
Max DD 10.44 6.26 5.15 8.07 (2.88)
Skew (0.25) 4.05 4.64 (3.58) (1.62)
Kurt (0.53) (3.95) (3.72) 4.56 1.65
Sharpe Ratio 4.70 4.36 4.16 2.95 (3.66)
Calmar Ratio 3.88 4.56 4.65 0.51 (3.43)
Corr. (Broker/Idio.HC) (10.47) (8.94) (8.05) (5.28) (3.61)
Corr. (Broker/Idio.RSU) (26.14) (20.52) (18.79) (7.87) (9.55)
Note: The table shows the idiosyncratic return summary stats for total wealth portfolios that are composed of human capital, RSUs, and broker portfolios across 263 firms. In the
total wealth optimizations, the broker account is optimized to minimize the variance of total wealth subject to a broker account level beta = 1 and varying (that is, 1%, 3%, and 5%)
tracking error constraints. Performance is gross of fees. If fees were added, results would have been lowered.

In line with the RSU-only scenario, the total wealth optimization significantly improves risk-adjusted
returns of the combined total wealth portfolio compared with the island case. Both Sharpe and
Calmar ratios are significantly higher than the benchmark across all three tracking error cases.
As observed in the RSU-only scenario above, the improvements in risk-adjusted returns appear
to decline beyond the 3% tracking error constraint, with the 5% tracking error case exhibiting
less attractive risk-adjusted returns. Correlations with firm-level human capital and RSUs are
materially lower than the benchmark, suggesting that the total wealth optimizations are effective
in diversifying idiosyncratic stock and career risk from the human capital component.
Taken together, the results presented in this section lend considerable support to the idea that
adjusting equity portfolios based on the unique risk in an individual’s nontradable wealth,
such as human capital and restricted stock, leads to material improvements in the investor’s
total wealth outcomes. We demonstrate that the stock-level total wealth optimization framework
introduced in this paper leads to statistically and economically significant improvement in
risk-adjusted returns, especially when isolating the impact on idiosyncratic returns. For investors
with concentrated positions in an asset, reducing the risk associated with the idiosyncratic risk
is of primary concern. Not only does asset pricing theory posit that this risk is not compensated
with higher returns, but it also exposes investors to the risk of idiosyncratic drawdowns. We
find that even with a moderate tracking error of 3%, the total wealth optimized can reduce the
idiosyncratic maximum return drawdown by 166 basis points on average across our sample of
US large-cap companies.
Case Study: Impact of Covid Lockdowns on Leisure Companies

To further illustrate how the total wealth framework developed in this study may help mitigate
idiosyncratic shocks to investors’ nontradable wealth, we analyze the efficacy of pursuing
the total wealth optimization approach for a select group of companies that were significantly
impacted by the onset of the covid-19 pandemic in March 2020. The government lockdowns
during the initial stages of the pandemic put certain business models on the brink of bankruptcy,
leading to precipitous declines in stock prices. Few businesses were more impacted than
the leisure industry, with hotels, cruise lines, and casino companies’ core businesses coming
to a virtual standstill. Employees in these firms, especially the ones with share ownership,
took significant hits to their personal balance sheets, with company managements laying off
or furloughing people to preserve capital, impacting the value of workers’ human capital.
In the empirical analysis in this section, we explore to what extent the total wealth optimization
approach can mitigate the business model-specific risks of leisure companies.
As in the previous section, we evaluate the ability of total wealth portfolios to provide better
risk-adjusted returns by minimizing the variance of total wealth and compare their performance
with the TME Index and the simple exclusionary approach. We use the same wealth asset
assumptions as in the previous section based on a 45-year-old individual with USD 90,000 in
RSUs, USD 240,000 in firm-specific human capital, and USD 270,000 in a direct-indexing broker
account. The portfolios are optimized daily from March 2020 through the end of May 2020.

Table 11 shows the return and risk summary statistics for the hotel & cruise line and casino
companies in our sample. Our sample consists of three companies in each group. Hotels & cruise
lines include Expedia (EXPE), Marriot (MAR), and Royal Caribbean (RCL), while casinos are
made up of Las Vegas Sands (LVS), MGM Resorts (MGM), and Wynn Resorts (WYNN). Table 11
highlights that the stocks of leisure companies came under significant stress during the first half
of 2020, with hotel & cruise line and casino stocks losing more than 50% during the height of
selloff. Our estimate of firm-level human capital was also impacted, declining around 10% over
the period. While the overall market recovered quickly, posting a flat return, leisure stocks only
partly recovered their losses at the end May 2020, ending the period down more than 20%.
Table 11: Summary Statistics March 2, 2020–May 29, 2020

Hotels & Casino Island
Cruise Lines (TME Index)
RSU Return -23.35% -21.29% 0.32%
Max DD -57.33% -54.97% -28.99%
Human Capital Return -7.84% -6.56%
Max DD -10.21% -9.28%
Total Portfolio Island Return -6.96% -5.96%
Statistics
(RSU, Human Max DD -20.79% -19.46%
Capital, and
Broker Combined)
Company Exclusion Return -6.95% -5.97%
Max DD -20.78% -19.46%
Sector Exclusion Return -7.52% -6.52%
Max DD -21.01% -19.54%
Total Wealth Return -6.61% -4.19%
(TE < 1%)
Max DD -20.55% -18.89%
(TE < 3%)
Max DD -20.06% -18.07%
(TE <5%)
Max DD -20.06% -18.10%
Note: The table shows the average return and drawdown of hotel & cruise line and casino companies from March 2, 2020, to May
29, 2020. RSUs and human capital show the average return of hotel & cruise line (EXPE, MAR, and RCL) and casino (LVS, MGM, and
WYNN) companies in our sample. The Total Portfolio Statistics show the household-level return and risk statistics based on different
portfolio construction approaches for the broker portfolio.

The combined household-level risk and return stats show that the total wealth portfolios provided
superior risk-adjusted performance during this period. All three total wealth variants have
higher returns and lower drawdowns than the island portfolio. The company exclusion has only
marginal impact (that is, ^1 basis point) on risk and returns, while sector exclusions lead to
suboptimal return, with both returns and risk deteriorating relative to the benchmark. The 3% and
5% tracking error total wealth portfolios see the biggest improvement in risk-adjusted perfor-
mance, even as the reward from the additional tracking error from 3% to 5% is limited.
Figure 7 and Figure 8 show the daily time series of the difference in the drawdown between
total wealth/exclusions-based broker portfolios compared with the island (that is, TME Index)
broker portfolio for hotels & cruise lines and casinos, respectively. Figure 7 confirms that total
wealth broker portfolios exhibit materially improved drawdowns compared with exclusions-
based approaches. In the 5% tracking error case, the broker portfolio has a 3.48% lower
maximum drawdown than the benchmark. Company exclusions have no material impact on
the benchmark, and the sector exclusion (excluding the consumer discretionary sector)
leads to larger drawdowns. Figure 8 shows that, for casinos, the improvements are even more
pronounced, with the 5% and 3% total wealth tracking error portfolios improving broker
portfolio drawdowns by 4.41% and 3.95%, respectively. The marginal benefit of taking on
additional tracking is reduced in relation to the hotel & cruise line sample, suggesting portfolio
risk mitigation can be achieved with moderate tracking error in the casino example.
Figure 7: Hotels & Cruise Lines Broker Portfolio Drawdown Compared With Island
4% Total Wealth
(TE<1%)
3%
Total Wealth
(TE<3%)
2%
Total Wealth
(TE<5%)
1%
Company
0% Exclusion
Sector
-1% Exclusion
-2%
03/02/20 03/16/20 03/30/20 04/13/20 04/27/20 05/11/20 05/25/20
Portfolios referenced are hypothetical, do not represent actual decision-making, and may not reflect the impact material economic
and market factors might have had on Morningstar’s decision-making process. Future performance or returns can differ significantly
from the backtested performance shown. Performance is gross of fees, if fees were deducted, the performance results would have
been lowered.

Figure 8: Casino Broker Portfolio Drawdown Compared With Island

5% Total Wealth
(TE<1%)
4%
Total Wealth
3% (TE<3%)
Total Wealth
2%
(TE<5%)
1% Company
Exclusion
0%
Sector
-1% Exclusion
-2%
03/02/20 03/16/20 03/30/20 04/13/20 04/27/20 05/11/20 05/25/20
Portfolios referenced are hypothetical, do not represent actual decision-making, and may not reflect the impact material economic
and market factors might have had on Morningstar’s decision-making process. Future performance or returns can differ significantly
from the backtested performance shown. Performance is gross of fees, if fees were deducted, the performance results would have
been lowered.
Overall, the covid case study lends further support to the notion that the total wealth optimiza-
tion approach at moderate levels of tracking error (that is, 3%) can reduce the drawdown risk at
the broker account and the combined total portfolio level. The case study highlights that even
during a period of marketwide stress, when cross-asset correlations tend to rise, the total wealth
approach can mitigate the impact of idiosyncratic shocks by investing in firms that are not
exposed to the same economic drivers. Additionally, we add to the evidence that simple exclu-
sionary approaches are ineffective in mitigating idiosyncratic portfolio risk. Their impact is either
not economically meaningful (that is, company exclusion) or they are shown to deteriorate the
risk-adjusted returns (that is, sector exclusions) compared with the cap-weighted benchmark.

Conclusions
In the past, the investment industry has failed to provide investors with effective tools to mitigate
idiosyncratic risk embedded in their broader economic net worth. Rather than addressing
this need, the industry’s predominant focus has been on creating strategies that aim to achieve
positive excess returns compared with a generic index or replicating an index in a cost-effective
way. This misses the fact that financial asset portfolios are not managed in a vacuum and
that investors’ total economic balance sheets are often overly exposed to an idiosyncratic risk
factor such as the fortunes of a single firm. The issue has grown in importance with the
increasing use of stock-based compensating, meaning that investors increasingly hold restricted
stock units and that their future income (that is, human capital) is more directly linked to
single-company outcomes. In this study, we set out to develop a framework to mitigate idiosyn-
cratic portfolio risks at the individual security level. We provide evidence that a portfolio that
seeks to minimize idiosyncratic risk in an investor’s total wealth leads to economically and
statistically significant improvements in risk-adjusted returns. Our research makes several
important contributions to the portfolio literature:
1. We develop a novel framework to model human capital at the firm level, incorporating
information on stock-level compensation and default risks from the credit default swap
market to create a daily series of human capital returns at the firm level. The firm-level
human capital model allows us to incorporate the idiosyncratic firm risk into a portfolio
optimization process and evaluate the efficacy of portfolios in diversifying the firm-level
career risk.
2. We develop a total wealth optimization framework that seeks to minimize idiosyncratic
portfolio risk at the individual firm level by incorporating RSUs and firm-level human capital.
Across a sample of 263 firms, we provide evidence that a portfolio around a concentrated
restricted stock position leads to materially better risk-adjusted returns at the total portfolio
level, reducing idiosyncratic portfolio drawdowns by an average 1.66% compared with a
cap-weighted benchmark. Similarly, we document material increases in risk-adjusted returns
when jointly incorporating human capital and restricted stocks. Additionally, total wealth
optimized portfolios meaningfully reduce correlations between the restricted stocks and the
firm-level human capital series, suggesting that the total wealth approach effectively
diversifies idiosyncratic risks. By contrast, we show that simple company and sector
exclusions are ineffective in improving total portfolio outcomes.
3. Finally, in a case study on the impact of government lockdowns on leisure companies, we
show that total wealth optimized portfolios significantly outperform their cap-weighted
benchmark, suggesting that our approach is effective in mitigating idiosyncratic portfolio
shocks even during periods of elevated market stress.

The results of this paper have important practical implications on how to build “personalized”
investment solutions that help investors meet their financial goals. While existing direct-indexing
solutions often present investors with many ways to tailor portfolios, they generally lack a
rigorous economic framework to help investors know how to customize their portfolio to improve
investment outcomes. The total wealth framework developed in this study fills this void and can
form the basis for investment solutions that seek to customize portfolios based on the unique
risks embedded in an investor’s nontradable wealth.

22 West Washington Street
Chicago IL 60602 USA
References
Blanchett, D.M., & Straehl, P.U. 2015. “No Portfolio Is an Island.” Idzorek, T.M., & Kaplan, P.D. 2023. Lifetime Financial Advice:
Financial Analysts Journal, Vol. 71, No. 3, P. 15. A Personalized Optimal Multi-Level Approach. Charlottesville,
VA: CFA Institute Research Foundation.
Blanchett, D.M., & Guillemette, M. 2019. “Do Investors
Consider Nonfinancial Risks When Building Portfolios?” Kyrychenko, V. 2008. “Optimal Asset Allocation in the Presence
Financial Analysts Journal. of Nonfinancial Assets.” Financial Services Review, Vol. 17, No.
1, P. 69.
Cerulli. 2022.”The Case for Direct Indexing: Differentiation
in a Competitive Marketplace.” Cerulli Associates. Lintner, J. 1965. “The valuation of risk assets and the selection
Dec. 1, 2022. https://www.cerulli.com/resource/ of risky investments in stock portfolios and capital budgets.”
white-paper-the-case-for-direct-indexing The Review of Economics and Statistics, Vol. 47, No. 1, P. 13.
Eiling, E. 2013. “Industry-Specific Human Capital, Idiosyncratic Mauboussin, M.J., & Callahan, D. 2023. “Stock-Based
Risk, and the Cross-Section of Expected Stock Returns.” Compensation: Unpacking the Issues.” Counterpoint Global
Journal of Finance, Vol. 68, No. 1, P. 43. Insights, Morgan Stanley.
Eisfeldt, A.L., Falato, A., & Xiaolan, M.Z. 2018. “Human Sharpe, W.F. 1964. “Capital Asset Prices: A Theory of Market
Capitalists.” NBER Macroeconomics Annual. Available at Equilibrium Under Conditions of Risk.” The Journal of Finance,
SSRN: https://ssrn.com/abstract=3375849 Vol. 19, No. 3, P. 425.
Heaton, J., & Lucas, D. 2000. “Portfolio Choice and Asset Prices: Treynor, J.L. 1961. “Market Value, Time, and Risk.” Unpublished
The Importance of Entrepreneurial Risk.” Journal of Finance, manuscript, University of California, Berkeley.
Vol. 55, No. 3, P. 1,163.
Viceira, L. 2001. “Optimal Portfolio Choice for Long-Horizon
Ibbotson, R., Milevsky, M.A., Chen, P., & Zhu, K.X. 2007. Investors with Nontradeable Labor Income.” Journal of Finance,
Lifetime Financial Advice: Human Capital, Asset Allocation, Vol. 56, No. 2, P. 433.
and Insurance. Charlottesville, VA: Research Foundation
of CFA Institute. Wallmeier, M., & Zainhofer, F. 2006. “How to Invest over the
Life Cycle: Insights from Theory.” Journal für Betriebswirtschaft,
Vol. 56, No. 4, P. 219.
All investments involve risk, including the loss of principal. There can be no assurance that any looking statements. Such forward-looking statements involve known and unknown risks,
financial strategy will be successful. Customers should seriously consider if an investment is uncertainties, and other factors which may cause the actual results to differ materially and/or
suitable for them by referencing their own financial position, investment objectives, and risk profile substantially from any future results, performance, or achievements expressed or implied by
before making any investment decision. Morningstar Investment Management does not guarantee those projected in the forward-looking statements for any reason.“
that the results of their advice, recommendations, or objectives of a strategy will be achieved.
Morningstar Wealth is made available through Morningstar’s Investment Management Group
This commentary is for general educational and informational purposes only. The information, and includes such products as: (i) the Morningstar Wealth Platform; (ii) Morningstar Funds
data, analyses, and opinions presented herein are designed to provide general information about Trust, (iii) Morningstar OfficeSM, Morningstar’s RIA portfolio software service; (iv) Morningstar®
investing, to inform investors about developments in the broader financial ecosystem, and/or to ByAllAccounts®, Morningstar’s investment data aggregation service; and (v) Morningstar.com®,
help investors interpret market and regulatory shifts. This material is not an offer of our investment Morningstar’s individual investor site offering. Morningstar’s Investment Management
advisory services or an offer to buy or sell a security. Group consists of certain Morningstar subsidiaries that are authorized in the jurisdictions in
which they operate to provide investment management and advisory services. In the United
This commentary contains certain forward-looking statements. We use words such as “expects,” States, these subsidiaries are Morningstar Investment Management LLC and Morningstar
“anticipates,” “believes,” “estimates,” “forecasts,” and similar expressions to identify forward- Investment Services, LLC.
©2024 Morningstar. All Rights Reserved. Unless otherwise provided in a separate agreement, you may use this report only in the country in which its original distributor is based. The information, data,
analyses, and opinions presented herein do not constitute investment advice; are provided solely for informational purposes and therefore are not an offer to buy or sell a security; and are not warranted
to be correct, complete, or accurate. The opinions expressed are as of the date written and are subject to change without notice. Except as otherwise required by law, Morningstar shall not be responsible
for any trading decisions, damages, or other losses resulting from, or related to, the information, data, analyses, or opinions or their use. The information contained herein is the proprietary property of
Morningstar and may not be reproduced, in whole or in part, or used in any manner, without the prior written consent of Morningstar. To license the research, call +1 312 696-6869.

SSRN 4864198

Uploaded by

Copyright:

Available Formats

SSRN 4864198

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SSRN 4864198

Uploaded by

Copyright:

Available Formats

R E S E A R C H PA P E R

Should Your Stock Portfolio

Robert ten Brincke

Carlos Gutierrez Mangas

For financial advisor use only. Not for public distribution.

In this paper, we develop a comprehensive theoretical framework to motivate security-level

For financial advisor use only. Not for public distribution.

For financial advisor use only. Not for public distribution.

….portfolio choice must consider the

For financial advisor use only. Not for public distribution.

Modeling Firm-Level Human Capital

HCi,t = Non.Stocki,t + Stocki,t (1)

Stocki,t= Stock.Pctgi,t-1 × Non.Stocki,t-1 × (1 + returni,t) (3)

Non.Stocki,t is the present value of mortality-adjusted wages

Stocki,t is the present value of the stock compensation

qR-n is the mortality weight for a given age

g is a nominal constant growth rate

wi,t is the level of wages

R-n is the number of years remaining in a work life

returni,t is the total return for stock i

For financial advisor use only. Not for public distribution.

Figure 1: Inflation, Labor Cost, and Average Hourly Earnings Indexes

For financial advisor use only. Not for public distribution.

For financial advisor use only. Not for public distribution.

In the estimation of human capital, we use stock compensation as a percentage of staff

Σ k,t Staff Expense

Σk,t Stock Compensation

For financial advisor use only. Not for public distribution.

5 Notice the material drop in the number of firms.

Table 2: Trend in Stock Compensation

2007 406 4.20% 2.60% 1.10% 0.50%

For financial advisor use only. Not for public distribution.

Figure 2: Stock Compensation by Profitability

Mean Stock Compensation (2007-2022)

Deciles by Market Value (Size)

For financial advisor use only. Not for public distribution.

+ β3Mkt. Valuei,t+ β4Leveragei,t—1+ β5Earnings Yieldi,t+ εi,t

Source: Compustat & author’s calculations (estimations).

For financial advisor use only. Not for public distribution.

For financial advisor use only. Not for public distribution.

Figure 4: Estimated Median CDS Spreads (Sample Year-End Dates)

Estimated Median CDS Spread

zi,t = (ROAi,t + Mkt. Vali,t) × (1/σ(ROA ))i,t

spreadi,t, k = β0,k + β1,k i,t zi,t (9)

For financial advisor use only. Not for public distribution.

Table 4: Risk Groups by Z-Score and Observed Median Spread

Figure 5: 1-Year Tenor Median Spread by Risk Group

For financial advisor use only. Not for public distribution.

Figure 6: Correlation Human Capital Changes vs. Stock Returns

Credit Risk Level Correlation

5 12% 14% 26% 32% 41% 46% 39% 38% 9% 8%

4 7% 18% 24% 32% 38% 36% 36% 38% 26% 19%

3 7% 14% 23% 29% 41% 41% 41% 34% 14% 12%

2 5% 12% 24% 31% 38% 41% 37% 10% 19% 2%

1 1% 10% 21% 24% 31% 34% 30% 27% 14% 5%

For financial advisor use only. Not for public distribution.

Table 5: Accounts Relevant to the Optimization

Age Tradable? Size in % $ Amount

For financial advisor use only. Not for public distribution.