Risk Based and Factor Investing 1st

Edition Jurczenko
Visit to download the full and correct content document:
https://textbookfull.com/product/risk-based-and-factor-investing-1st-edition-jurczenko/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Index Fund Management: A Practical Guide to Smart Beta,

Factor Investing, and Risk Premia Fadi Zaher

https://textbookfull.com/product/index-fund-management-a-
practical-guide-to-smart-beta-factor-investing-and-risk-premia-
fadi-zaher/

Equity Smart Beta and Factor Investing for

Practitioners Khalid Ghayur

https://textbookfull.com/product/equity-smart-beta-and-factor-
investing-for-practitioners-khalid-ghayur/

Asset Management A Systematic Approach to Factor

Investing 1st Edition Andrew Ang

https://textbookfull.com/product/asset-management-a-systematic-
approach-to-factor-investing-1st-edition-andrew-ang/

Country Asset Allocation: Quantitative Country

Selection Strategies in Global Factor Investing 1st
Edition Adam Zaremba

https://textbookfull.com/product/country-asset-allocation-
quantitative-country-selection-strategies-in-global-factor-
investing-1st-edition-adam-zaremba/
Machine Learning for Factor Investing R Version Chapman
and Hall CRC Financial Mathematics Series 1st Edition
Guillaume Coqueret

https://textbookfull.com/product/machine-learning-for-factor-
investing-r-version-chapman-and-hall-crc-financial-mathematics-
series-1st-edition-guillaume-coqueret/

Real Estate Investing Market Analysis Valuation

Techniques and Risk Management 1st Edition Benedetto
Manganelli (Auth.)

https://textbookfull.com/product/real-estate-investing-market-
analysis-valuation-techniques-and-risk-management-1st-edition-
benedetto-manganelli-auth/

Investing in Hedge Funds A Guide to Measuring Risk and

Return Characteristics 1st Edition Turban Bali

https://textbookfull.com/product/investing-in-hedge-funds-a-
guide-to-measuring-risk-and-return-characteristics-1st-edition-
turban-bali/

Risk Return Analysis The Theory and Practice of

Rational Investing Volume One 1st Edition Harry
Markowitz

https://textbookfull.com/product/risk-return-analysis-the-theory-
and-practice-of-rational-investing-volume-one-1st-edition-harry-
markowitz/

The Forrester WaveTM Risk Based Authentication Q2 2020

1st Edition Andras Cser

https://textbookfull.com/product/the-forrester-wavetm-risk-based-
authentication-q2-2020-1st-edition-andras-cser/
Risk-Based and Factor Investing
 Quantitative Finance Set
coordinated by
Patrick Duvaut and Emmanuelle Jay

Risk-Based and
Factor Investing

Edited by
Emmanuel Jurczenko
First published 2015 in Great Britain and the United States by ISTE Press Ltd and Elsevier Ltd

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as
permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,
or in the case of reprographic reproduction in accordance with the terms and licenses issued by the
CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the
undermentioned address:

ISTE Press Ltd Elsevier Ltd

27-37 St George’s Road The Boulevard, Langford Lane
London SW19 4EU Kidlington, Oxford, OX5 1GB
UK UK
www.iste.co.uk www.elsevier.com

Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical treatment
may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and
using any information, methods, compounds, or experiments described herein. In using such information
or methods they should be mindful of their own safety and the safety of others, including parties for
whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any
liability for any injury and/or damage to persons or property as a matter of products liability, negligence
or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in
the material herein.

For information on all our publications visit our website at http://store.elsevier.com/

© ISTE Press Ltd 2015

The rights of Emmanuel Jurczenko to be identified as the author of this work have been asserted by him
in accordance with the Copyright, Designs and Patents Act 1988.

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
A catalog record for this book is available from the Library of Congress
ISBN 978-1-78548-008-9

Printed and bound in the UK and US

 Acknowledgements

I would like to thank Professor Patrick Duvaut from Telecom Paris-Tech and Dr.
Emmanuelle Jay from QAMLab for guidance throughout the publishing process. I
would also thank Ludovic Moulard and Harry Greene at ISTE for their editorial
work. I further thank Professors Serge Darolles and Gaelle Le Fol from University
Paris-Dauphine for providing the opportunity to edit this book and the
QuantValley/QMI Research Initiative for its support. Finally I also want to thank
Jérôme Teiletche, Head of Cross-asset Solutions at Unigestion, for his helpful
suggestions and comments.
 Preface

This book contains a collection of 20 exclusive new Chapters written by leading

academics and practitioners in the area of risk-based and factor investing (RBFI), a
term that encompasses both alternative non-return based portfolio construction
techniques and investing style risk premia strategies. The Chapters are intended to
introduce readers to some of the latest, cutting edge research encountered by
academics and professionals dealing with RBFI solutions. The articles deal with new
methods of building strategic and tactical risk-based portfolios, constructing and
combining systematic factor strategies and assessing the related rules-based
investment performances. Although numerous articles are technical in nature, this
book can assist portfolio managers, asset owners, consultants, academics, as well as
students wanting to further understand the science and the art of risk-based and
factor investing.

Emmanuel JURCZENKO
Ecole Hotelière De Lausanne, HES-SO
University of Applied Sciences Western Switzerland
September 2015
 1

Advances in Portfolio Risk Control

Spurred by the increased interest in applying “risk control” techniques in an asset allocation
context, we offer a practitioner’s review of techniques that have been newly proposed or revived
from academic history. We discuss minimum variance, “1/N” or equal-weighting, maximum
diversification, volatility weighting and volatility targeting – and especially equal risk contribution
or “risk parity”, a concept that has become a real buzz word. We start from a taxonomy of risk
control techniques. We discuss their main characteristics and their positives and negatives and
we compare them against each other and against the maximum Sharpe ratio (SR) criterion. We
illustrate their implications by means of an empirical example. We also highlight some key
papers from the vast and still growing literature in this field. All in all, we aim to provide a
practical and critical guide to risk control strategies. It may help to demystify risk control
techniques, to appreciate both the “forest” and “trees” and to judge these techniques on their
potential merits in practical investment applications.

1.1. Introduction

Recently, there has been increased interest in applying “risk control” techniques
in an asset allocation context. Some examples of techniques that have been newly
proposed or revived from academic history are “1/N” or equal-weighting, minimum
variance, maximum diversification, volatility weighting and volatility targeting –
and especially equal risk contribution or “risk parity”, a concept that has become a
real buzz word. In this chapter, we start from a taxonomy of risk control techniques.
We discuss their main characteristics and their positives and negatives, we compare

Chapter written by Winfried G. HALLERBACH*.

*Robeco Asset Management
2 Risk-Based and Factor Investing

them against each other and against the maximum Sharpe ratio (SR) criterion and we
illustrate their implications by means of an empirical example. We also highlight some
key papers from the vast and still growing literature in this field. All in all, we aim to
provide a practical and critical guide to risk control strategies that may help to
appreciate both the “forest” and “trees” and to judge these techniques on their actual
potential merits in practical investment applications. For an in-depth exposition,
comparison and evaluation of these strategies, we recommend [RON 14].

The main question in risk control is: “does it work?” Do risk control techniques
achieve the ex ante targeted risk balance or risk profile? Can we avoid hot spots
(pockets of risk concentration in a portfolio) and can we achieve diversification against
losses? Apart from this risk-budgeting perspective, a large part of the literature has
promoted risk control as a full-fledged investment criterion − suggesting that
controlling the risk dimension is sufficient to build a portfolio or an opportunity to
reap risk-adjusted outperformance. But, why would ignoring the return dimension ex
ante produce portfolios that are superior in terms of ex post risk-adjusted performance?

Several studies indicate that the historical outperformance of risk control

strategies can be linked to overweighting asset classes that in the rear-view mirror
have paired high historical risk premia with low-risk levels (as is the case for bonds,
e.g.), or to implicit exposures to factor premia [JUR 15]. However, focusing directly
on factor exposures, as is done in factor investing (see [ANG 14]), provides a much
more efficient and effective way to capture factor premia. Still, focusing only on risk
aspects when forming a portfolio can be a perfectly sensible heuristic (see [FIS 15])
or a starting point if we have only low confidence in ex ante risk premia estimates.
From the perspective of estimation risk, mis-estimation of risk premia has the
greatest impact on portfolio composition and especially risk premia are notoriously
hard to estimate ex ante. For example, suppose that ex ante we cannot meaningfully
differentiate between all assets’ SRs (so assuming that all SRs are equal), then
constructing a maximum diversification portfolio (MDP) gives the maximum Sharpe
ratio portfolio (MSRP). If, in addition to equal SRs, we cannot meaningfully
differentiate between asset correlations (so also assuming that all correlations are
uniform), then applying risk parity gives the MSRP. So, besides the risk-budgeting
dimension, also the potential relevance of risk control techniques in full-fledged
risk-return optimization is not to be underestimated.

This chapter is divided as follows. Section 1.2 introduces our empirical example
and provides some preliminaries. Section 1.3 outlines the MSRP within the familiar
mean-variance framework. Next, we discuss the risk control strategies. The main
skeleton of the taxonomy of risk control strategies has a cross-section and a time-
series branch. The objective of risk control in the cross-section is to control a
portfolio’s risk profile at a given point in time. The focus is across assets:
reweighting the portfolio constituents so as to obtain a desired risk profile. The main
 Advances in Portfolio Risk Control 3

cross-sectional risk control strategies are: 1/N, or the equally weighted portfolio
(section 1.4), the minimum variance portfolio (MVP) (section 1.5) and the MDP
(section 1.6). Next, we have risk parity, which comes in two flavors, the equal risk
contribution portfolio (ERCP) or “full” risk parity (section 1.7) and the inverse
volatility portfolio (IVP) or “naive” risk parity, implying volatility weighting in
cross-section (section 1.8). The objective of time-series risk control is to control the
portfolio risk level over time. There are two closely related time-series techniques:
volatility weighting over time, or adjusting the exposure to risky assets according to
the level of forecasted volatility, and volatility targeting, or volatility weighting with
the specific goal to achieve a prespecified level of portfolio volatility (section 1.9).
Each of these sections is organized according to a fixed format, starting with main
references, followed by the recipe to calculate the particular portfolio, its
characteristics and evaluation. Section 1.10 concludes with an overall evaluation.
Section 1.11 contains technical details.

1.2. The empirical example and preliminaries

We consider monthly returns in excess of the risk-free rate over the decade
January 2005 through December 2014 for a selection of US asset classes: equities,
treasuries (Tsies), investment grade (IG) corporates and high yield (HY) corporates.
The risk-free return comes from the Ibbotson “Stocks, Bills, Bonds and Inflation”
1
database. Equity is the market factor from Kenneth French’s database . The fixed
2
income series are taken from Barclays Capital Live . All returns are in USD. The
composition of the market capitalization weighted portfolio “Mkt Cap” is estimated
3
as per 2014Q4 . “EqWtd” is the equally weighted portfolio.

Table 1.1 shows the descriptive statistics. Over the past decade, fixed income
assets were the real winners in terms of risk-adjusted performance. This is not
surprising given the substantial tail wind from decreasing interest rates. Especially
Tsies paired a 3% average return with a relatively low level of risk. Equities showed
the highest volatility, but viewing the SR this was not matched by a proportionally
higher risk premium. Equities and Tsies were negatively correlated, providing hedge
opportunities (see the small negative correlation between Tsies and the market cap

1 The Ibbotson risk free rate and the equity market factor can be downloaded from
http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html.
2 Download from https://live.barcap.com/.
3 Sources are (1) Securities Industry and Financial Markets Association (SIFMA), US Bond
Market Outstanding, download from http://www.sifma.org/research/statistics.aspx, (2) World
Bank, year-end market capitalization of listed companies by country, download from
http://data.worldbank.org/indicator/CM.MKT.LCAP.CD and (3) Barclays Live (for the
relative IG and HY capitalizations).
4 Risk-Based and Factor Investing

portfolio). The highest correlation is between equities and HY, pointing at a strong
link between equity risk and credit risk. Credit risk is dominant in HY and the
negative correlation between interest rates and credit spreads manifests itself in the
negative correlation between Tsies and HY.

Equities Tsies IG HY Mkt Cap EqWtd

Market cap
weight: 53% 29% 14% 4% 100%

Return
statistics:

avge p.a. % 7.54 2.97 4.17 6.62 5.71 5.33

stdev p.a. % 15.11 4.16 6.04 10.54 8.46 6.78

Sharpe ratio 0.50 0.71 0.69 0.63 0.67 0.78

Correlations:

Equities –0.30 0.35 0.74 0.98 0.88

Tsies 0.44 –0.24 –0.11 –0.01

IG 0.63 0.53 0.73

HY 0.78 0.90

Mkt Cap 0.95

Table 1.1. Statistics of US excess returns (p.a.) over

the risk-free rate (January 2005 – December 2014)

1.2.1. Money allocation versus risk allocation

The money allocation in the market cap portfolio is given in Table 1.1. For the
risk allocation within the market cap portfolio, we compute the Ordinary Least
Squares (OLS) regression slope or beta of the assets against the market cap
portfolio. This beta represents the relative marginal contribution of the
corresponding asset to the overall portfolio risk (for details, see section 1.11). The
component risk contribution is given by the product of the portfolio weight and beta.
Hence, the betas can be interpreted as the adjustment factors to transform money
allocation into risk allocation (note that the weighted average value of beta is unity).
The risk allocation within the market cap portfolio is given in Table 1.2. From this
 Advances in Portfolio Risk Control 5

table, we see a nasty surprise: at first sight, the market cap portfolio appears to be a
properly diversified portfolio but in reality more than 90% of the portfolio risk is
due to equities (this was already forewarned by the high correlation between equities
and the market cap portfolio as shown in Table 1.1). The same finding is widely
reported for conventional 60/40 equity-bond portfolios in general, and for typical
“Yale-type” portfolios (where alternatives and/or commodities are added to main
holdings of equities and bonds).

Eq Tsies IG HY sum
Market cap weight 53% 29% 14% 4% 100%
Beta 1.74 –0.06 0.38 0.97
% Risk contribution 92% –2% 5% 4% 100%

Table 1.2. Risk attribution with respect to Mkt Cap portfolio

Although we focus on volatility as the risk measure, most of the results in this
chapter carry over to downside risk measures such as portfolio loss or value-at-risk
(VaR). Table 1.3 shows the average of the six largest monthly losses against the
risk-free rate on the market cap portfolio: equities contributed by far the most to the
realized losses. The excessive contribution of equities to (downside) risk within
portfolios that seem only moderately geared toward equities provided the impetus to
the research into risk control strategies (see also [QIA 05]). In the remainder of this
chapter, we use this empirical example to illustrate various risk control strategies.

Mkt Cap index Eq Tsies IG HY

–6.02 –5.39 –0.04 –0.40 –0.19
100% 90% 1% 7% 3%

Table 1.3. Absolute (in %) and relative contribution of assets to the average of the six
largest losses on the market cap portfolio (in terms of excess returns), 2005–2014

1.2.2. Implied risk premia and the implied Sharpe ratios

There is one additional perspective that we want to highlight – a perspective that

is helpful in evaluating risk control strategies vis à vis the MSRP. For each of the
portfolios that we discuss, we present the implied risk premia and the implied SRs of
the individual assets. Instead of using actual risk premia and the variance-covariance
matrix to calculate the MSRP, we reverse the process and assume that the reference
portfolio at hand actually is the MSRP. Together with the variance-covariance
matrix of excess returns, this allows us to derive the “imputed” risk premia
6 Risk-Based and Factor Investing

(pioneered by Sharpe [SHA 74]); together with the actual (historical) asset standard
deviations, we can then compute the implied SRs. Hence, given a particular
portfolio, these implied risk premia (or implied SRs) would make this reference
portfolio the MSRP. For more details, we refer the reader to section 1.11.

This reverse portfolio optimization is relevant when there is an uncertainty about

ex ante risk premia. After all, since the MSRP is the tangency portfolio to the mean-
variance efficient frontier without including risk-free borrowing and lending, this
portfolio is very sensitive to the input risk premia. Slight differences in these inputs
can result in very different (and sometimes “unrealistic” or extreme and hence
unacceptable) portfolios. At the same time, estimating ex ante risk premia is a very
difficult task. Reverse optimization can help since the assets’ implied risk premia,
derived from a reference portfolio such as the market capitalization weighted
portfolio or a risk control portfolio, serve as a sensible starting point. Depending on
the confidence placed in one’s ex ante views, next we can adjust the implied risk
premia accordingly. After reoptimization, the resulting portfolio is closer to the
original portfolio and is less extreme. This two-stage portfolio optimization process
is originally proposed by Black and Litterman [BLA 92] and extended by Haesen
et al. [HAE 14] who take the risk parity portfolio as the reference portfolio.

Table 1.4 presents the implied risk premia and the implied SRs of the market cap
portfolio. For equities, the implied risk premium is about 25% larger than the
historical risk premium. For IG, the implied risk premium is about half the historical
risk premium. So, if the market cap portfolio were the MSRP, equities would have
to offer a risk premium of 10% and IG of 2%. Conversely, if we felt confident in
extending the historical risk premia to the future, this implies that we should
increase the weight of IG and lower the weight of equities in order to increase the
SR of the market cap portfolio. For Tsies, the implied risk premium (and hence the
implied SR) is even slightly negative, reflecting the role of Tsies as a hedge in the
market cap portfolio. Because of the negative correlation of Tsies with equities (and
HY), their inclusion in the market cap portfolio would be justified even when their
risk premium were negative.

Eq Tsies IG HY Mkt Cap

Implied risk premium 9.96 –0.32 2.14 5.55 5.71
Implied Sharpe ratio 0.66 –0.08 0.35 0.53 0.67

Table 1.4. Implied risk premia (%) and implied

SRs within the market cap portfolio

In order to derive the implied risk premia from a portfolio different from the
market cap portfolio, we first calculate the relative risk aversion coefficient λ * as
 Advances in Portfolio Risk Control 7

implied by the market cap portfolio (see [SHA 74]): dividing the historical risk
premium on the market cap portfolio by the historical variance of the market cap
portfolio yields λ* = 8.0 . The implied risk premium on an alternative portfolio,
when assuming that this portfolio is mean-variance optimal, is then given by the
product of λ * and its historical variance. This portfolio risk premium is next
attributed to the assets comprised in the portfolio according to their relative risk
contributions (or betas) with respect to this portfolio (see section 1.11).

NOTE.– We use fairly conventional notation. The weight of asset i in portfolio p is

wi . Individual and portfolio risk premia or “rewards” (average returns in excess of
the risk-free rate) are denoted by ri and rp . Individual asset standard deviations or
volatilities are denoted by σ i and the portfolio volatility by σ p .The beta of asset i
with respect to portfolio p is βip (reflecting its relative marginal contribution to
portfolio volatility) and its correlation with the portfolio is denoted as ρip . Where
deemed necessary, technical details are mentioned in the main text; section 1.11
contains a general background and additional derivations.

1.3. Maximum Sharpe ratio portfolio (MSRP)

For a discussion of the SR, see [SHA 94]. For mean-variance portfolio theory
and for finding the composition of the MSRP, we refer to standard investment texts.

Recipe: Choose the portfolio weights to maximize the SR:

rp
max SR p = [1.1]
{w} σp

Characteristics: (1) For the MSRP, it is not possible to further increase its risk
premium without increasing its risk. This implies that for all assets the ratios of
marginal contributions to risk and reward are constant. An asset’s marginal
contribution to portfolio risk equals ∂σ p / ∂wi , whereas an asset’s marginal
contribution to the portfolio risk premium simply equals its risk premium,
∂ rp / ∂ wi = ri . Hence, for all assets within the MSRP, we must have:

1 ∂σ p 1 ∂σ p
= [1.2]
ri ∂wi rj ∂w j
8 Risk-Based and Factor Investing

Substituting the definition of beta from equation [1.23] in section 1.11, this
translates into ri / β ip = rj / β jp = rp (where the last equality follows from the fact
that the portfolio beta equals unity). Note that ri / βip is the Treynor’s [TRE 66] risk-
adjusted performance ratio. Hence, for each asset within the MSRP, its risk premium
should be equal to the product of (i) its beta with respect to the MSRP and (ii) the
risk premium of the MSRP:

ri = β ip rp [1.3]

This is the first-order condition of mean-variance optimality. (2) Since we can

rewrite the beta as the product of (i) the correlation with the portfolio and (ii) the
quotient of the asset and portfolio volatility, βip = ρip σ i / σ p , it follows that in the
MSRP the stand-alone asset SRs and the portfolio’s SR are related by
SRi = ρip ⋅ SR p . If SRi > ρip SRp , we can increase the SR of the portfolio by
increasing the weight of (or adding) asset i to the portfolio p. (3) Without any
additional constraints, the long-only MSRP can be a very concentrated portfolio. (4)
When all volatilities, correlations and risk premia are the same, then the MSRP is
the 1/N portfolio (which then also coincides with the ERCP and the MVP). After all,
diversification lowers risk but in the portfolio context all assets are perfect
substitutes. It is not possible to lower portfolio risk or increase the portfolio risk
premium by changing the weights. Hence, we end up with the equally weighted
portfolio.

Evaluation: Table 1.5 shows the historical statistics of the portfolios we consider.
Historically, the MSRP has the maximum SR. This is so by construction, since we
optimized the SR over the full historical sample period (in-sample). In practical
applications, we would use trailing historical windows (avoiding a look-ahead bias)
to periodically recalculate the weights. In this way, the out-of-sample properties of
the MSRP can be evaluated. The same argument applies to the other strategies that
we discuss below. Whether the MSRP indeed delivers the maximum SR ex post
depends on the quality of the inputs, especially the risk premia.

Cap Wtd MSRP 1/N MVP MDP ERCP IVP

avge p.a. 5.71 3.97 5.33 3.68 4.05 4.24 4.46
stdev p.a. 8.46 3.57 6.78 3.43 3.70 4.19 4.74
Sharpe
ratio 0.67 1.11 0.78 1.07 1.10 1.01 0.94

Table 1.5. Comparative historical excess return portfolio statistics, 2005–2014

 Advances in Portfolio Risk Control 9

In our example, the MSRP is indeed a concentrated portfolio, containing mostly

Tsies supplemented with HY and only 7% equities, see Table 1.6. Tsies dominate
because of their low volatility and negative correlation with HY. The smaller than
unity beta of Tsies reveals that Tsies are included as a diversifier; the larger than
unity betas of HY and equities show that these assets are included because of their
high average return. Slight changes in risk premia will change the composition of
the MSRP markedly. Note that the implied risk premia are different from the
historical risk premia as shown in Table 1.1. After all, the implied risk premia are
calculated from the derived risk aversion parameter λ* = 8.0 and the historical
portfolio variance (and hence are proportional to betas).

Eq Tsies IG HY sum
Weight 7% 74% 0% 19% 100%
Beta 1.90 0.75 1.41 1.67
% Risk contribution 13% 55% 0% 31% 100%
Implied risk premium 1.93 0.76 1.43 1.69 1.01
Implied Sharpe ratio 0.13 0.18 0.24 0.16 0.28

Table 1.6. Risk attribution with respect to MSRP,

and implied risk premia (%) and Sharpe ratios

1.4. 1/N or equal-weighting

Recipe: In equally weighted portfolios, each of the N assets is assigned a weight

of 1/N. In our example, each asset class gets a weight of 25% in the portfolio, with
monthly rebalancing [DEM 09].

Characteristics: (1) 1/N avoids concentrated positions – in terms of money

allocation. (2) Within equities, 1/N implies an exposure to the small-cap anomaly.
The market cap portfolio is tilted toward large cap stocks. The 1/N portfolio is tilted
toward small cap stocks and hence will capture a size premium. (3) Furthermore,
1/N implies a disciplined and periodical rebalancing of positions. By definition, the
market cap portfolio is a buy-and-hold portfolio. The 1/N portfolio, in contrast,
implies a “volatility pumping” effect: in order to maintain the 1/N allocation, we
have to buy (sell) out- (under-) performing assets. This is effectively a “buy low, sell
4
high” strategy, which profits from reversals . Depending on the frequency, the
rebalancing process implies portfolio turnover with the associated transaction cost
and exposure to potential illiquidity (since even the smallest market cap assets get a
weight of 1/N). (4) In Bayesian terms, the 1/N portfolio is the “uninformed prior”:

4 For the effects of rebalancing on portfolio returns, see [HAL 14].

10 Risk-Based and Factor Investing

the naively diversified portfolio that is optimal when we have no information to

discriminate between the attractiveness of assets. (5) It can be shown that when all
assets have the same volatility and when all pairwise correlations are the same, then
the 1/N portfolio is the MVP. In this case, the MVP also coincides with the ERCP,
see below. (6) From Tables 1.1 and 1.5, we see that the 1/N portfolio has a lower
risk than the market cap portfolio and a higher historical SR. This arises from
underweighting equities (with a lower SR) and overweighting IG and HY (with a
higher SR).

Evaluation: (1) Table 1.7 shows the 1/N portfolio statistics. It clearly shows that
equal money weights are very different from equal risk weights. Notably, Tsies act
as a strong diversifier (negatively correlated with equities and HY) and show
(virtually) zero risk contribution. Still, equity risk dominates in the 1/N portfolio,
accounting for about half of the portfolio volatility. (2) For equities, the implied risk
premium is 7.17% p.a. (which given historical volatility implies a SR of 0.47).
When we believe that the ex ante equity risk premium is below 7.17%, the weight of
equities should be lowered in order to improve the risk-adjusted portfolio
performance. Likewise, when we believe that the ex ante bond risk premium is
above –3 bps p.a., the weight of Tsies should be increased. Equivalent reasoning
applies to IG and HY.

Eq Tsies IG HY EqWtd
Weight 25% 25% 25% 25% 100%
Beta 1.95 –0.01 0.65 1.41
% Risk contribution 49% 0% 16% 35% 100%
Implied risk premium 7.17 –0.03 2.39 5.16 3.67
Implied Sharpe ratio 0.47 –0.01 0.40 0.49 0.54

Table 1.7. Risk attribution with respect to 1/N portfolio,

and implied risk premia (%) and Sharpe ratios

1.5. Minimum variance portfolio (MVP)

Haugen and Baker [HAU 91] show that market cap weighted portfolios are
inefficient (suboptimal) when there are market frictions and highlight the high-
relative performance of low volatility portfolios. Clarke et al. [CLA 06] extend
Haugen and Baker’s empirical research. Blitz and van Vliet [BLI 07] revive the
interest in the low volatility anomaly and provide possible explanations (behavioral
biases, leverage restrictions, and delegated portfolio management and
benchmarking).
 Advances in Portfolio Risk Control 11

Recipe: Choose the portfolio weights to minimize the portfolio variance:

max σ p2 = ∑ ∑ wi w jσ iσ j ρ ij [1.4]
{w} i j

The optimal portfolio is characterized by equal marginal contributions to

portfolio risk:

∂σ p ∂σ p
= [1.5]
∂wi ∂w j

Characteristics: (1) Note that marginal risk contributions are given by

∂σ p / ∂wi = σ ip / σ p = β ip ⋅ σ p , so all asset betas with respect to the MVP are
identical (and hence equal to unity). (2) Since an asset’s risk contribution equals
wi ⋅ ∂σ p / ∂wi ∝ wi , the risk contribution is proportional to the portfolio weight, so
risk allocation equals money allocation. (3) When all assets have the same volatility
and when all pairwise (imperfect) correlations are the same, then the MVP is the 1/N
portfolio: it pays to diversify over the assets but in the portfolio context, all assets
are perfect substitutes. (4) The MVP is the MSRP when all assets have the same risk
premium, ri = r (implying that all SRs SRi are proportional to 1/ σ i ). After all, in
this case, we have (see equation [1.2]):

1 ∂σ p 1 ∂σ p
= [1.6]
r ∂wi r ∂w j

Evaluation: (1) The MVP favors low volatility assets and low beta assets and
hence benefits from the low volatility anomaly. The MSCI Minimum Variance
Index and the S&P Low Volatility Index are examples of low-risk portfolios that are
designed to benefit from this anomaly. For more information on the low volatility
anomaly, see [BLI 07]. (2) Several studies have documented that MVPs also pick up
other priced anomalies. Clarke et al. [CLA 06] find that, in general, the MVP has a
substantially higher value (B/P) exposure than the market (since value stocks tend to
have low volatilities), which explains at least part of its higher average realized
return. Scherer [SCH 11] shows that the MVP not only loads significantly on the
Fama-French factors (large size and high value) but also finds that MVPs have a
negative beta bias (favor low beta assets) and favor assets with low residual
volatility. The latter effects crowd out the Fama-French factors in the sense that low
beta and low residual volatility alone can explain more of the variation in the MVP’s
excess returns than the Fama-French factors. This leads Scherer to conclude that low
beta and low residual volatility is a more efficient and effective way to capture the
low volatility anomaly than minimum variance. (3) When time passes and the MVP
12 Risk-Based and Factor Investing

is reoptimized, we will need to apply constraints on turn-over in order to mitigate

transaction costs. However, turnover constraints make the MVP a path-dependent
strategy. (4) The MVP is a concentrated portfolio. Assets with low volatility and/or
low correlations with other assets will carry a large weight. Conversely, assets with
high volatility and/or high correlations with other assets will carry a small or even
negative weight; when excluding short positions, these assets will not appear in the
MVP. This is illustrated in Table 1.8: IG is not included in the long-only MVP.
Table 1.8 confirms that when the assets comprised in the MVP have identical risk
premia, then the MVP is the MSRP. Note again that this only applies to assets that
are comprised in the MVP in the first place.

Eq Tsies IG HY MVP

Weight 5% 82% 0% 13% 100%

Beta 1.00 1.00 1.35 1.00
% Risk contribution 5% 82% 0% 13% 100%
Implied risk premium 0.94 0.94 1.27 0.94 0.94
Implied Sharpe ratio 0.06 0.23 0.21 0.09 0.27

Table 1.8. Risk attribution with respect to MVP, and

implied risk premia (%) and Sharpe ratios

(5) Table 1.8 also confirms that marginal risk contributions of MVP constituents
are identical (all betas equal unity) so the money allocation equals the risk allocation
in an MVP. (6) Table 1.5 shows that the MVP has the lowest historical risk premium
as well as the lowest volatility (by construction), yielding a SR slightly lower than
the MSRP and MDP. This lower volatility is achieved by overweighting Tsies at
82%, supplemented by positions in equities and HY which are negatively correlated
with Tsies. (7) Last but not least, the quadratic optimization underlying the MVP has
the property of being “error maximizing”, see [MIC 89]. This means that the
composition of the MVP is very sensitive to slight differences in variances and
covariances. When (part of) these differences are not real but due to sampling error,
this will propagate into portfolio composition. Again, note that we use the full
historical sample to calculate the weights of the MVP.

1.6. Maximum diversification portfolio (MDP)

The MDP is proposed by Choueifaty and Coignard [CHO 08].

24 Risk-Based and Factor Investing

1.11.3. Portfolio risk premium

The average portfolio return over the risk-free rate, the portfolio risk premium,
follows as:

rp = ∑ i∈ p wi ri [1.15]

The marginal and component contributions of asset i to the portfolio risk

premium are ri and wi ri , respectively.

1.11.4. Portfolio variance

The variance of portfolio excess returns is defined by the double sum:

σ p2 = ∑ i ∑ j wi w jσ ij [1.16]

By definition of the correlation ρij , the covariance σ ij can be expressed as

σ ij = ρijσ iσ j .

Since (1) the variance of a variable is the covariance of that variable with itself
and (2) the covariance is a linear operator (the covariance of a weighted sum is the
weighted sum of covariances), we can write the variance of the portfolio excess
return as:

var ( r%p ) ≡ cov ( r%p , r%p ) = cov ( ∑ w r% , r% ) = ∑ w cov ( r% , r% ) = ∑ w σ

i i i p i i i p i i ip , [1.17]

where σ ip is the covariance between the excess returns on asset i and the portfolio p.
So, although the portfolio variance is the quadratic sum of weights and covariances,
we can express the portfolio variance as the weighted sum of the covariances of each
asset with the portfolio: σ p2 = ∑ i wi σ ip .

1.11.5. Decomposing portfolio volatility

Dividing the previous expression by the portfolio volatility, we get:

σ ip
σ p = ∑ i wi = ∑ i wi σ p ρip [1.18]
σp
Another random document with
no related content on Scribd:
disturbances of bladder function have been described as vesical
crises, and recent French observers have observed so-called crises
clitoridiennes in female tabic patients which were characterized by
voluptuous sensations. All of these symptoms have these in
common: that they last but a short time, that their disappearance is
as rapid as their advent, and that they depend for their distribution on
the attitude of the disease in the cord. The vesical crises are more
apt to occur early than late in the disease and where the belt
sensation is in the hypogastric region. The gastric and enteric crises
are usually found when the belt is in the epigastric level, and the
bronchial, cardiac, and laryngeal crises when it is in the thorax and
neck. Some connection has been observed between the occurrence
of the lightning-like pains and these crises. Thus, a sudden cessation
of the former is sometimes the forerunner of the latter. It is also
found that one variety of these crises disappears to give way to
another: this is particularly noticed with the bronchial crises, which
often cease suddenly, to be followed by vomiting.
22 Krause, in a paper read before the Society of Neurologists at Berlin
(Neurologisches Centralblatt, 1885, p. 543), found either laryngeal crises or other
laryngeal symptoms, such as demonstrable ataxia of movement of the vocal cords, in
13 out of 38 cases. This proportion would be far too high for tabes in general; his
cases were probably advanced ones. He established the interesting fact that the
laryngeal crisis may sometimes be provoked by pressure with a probe on the superior
laryngeal nerve at its laryngeal entry-point. Jastrowitz claims to have determined the
existence of actual paralysis of the vocal cords with the crises, but Krause concludes
from the experimental inductibility of the crisis that it cannot be due to a posticus
paralysis. In a discussion on the subject Remak affirms that a unilateral paralysis of
the crico-arytenoideus posticus may be an early or the earliest symptom of tabes. It
seems, however, that in all cases where he determined such paralysis there had been
disturbances of function of other cranial nerves in unusual severity at about the same
time.

23 They must be extremely rare: they have not occurred in a single one of my 81
private cases, nor do I recall one in clinic or dispensary experience.
There is some analogy between the exceptionally-occurring mental
disturbance of tabic patients and the crises. Like them, it resembles
a disturbance of innervation, in this instance the centres regulating
the cerebral circulation appearing to be at fault. It also seems as if in
cases of this character the direction of the mental disturbance were
determined in some sense by the emotional relations of the
oblongata, for the insane outbreak usually consists in a brief but
extreme outburst either of a depressed and melancholiac or an
expansive maniacal or delirious outburst. It is a very rare occurrence,
and usually limited to the latest stages. Much more common is the
development of paretic dementia, but this is to be regarded rather as
a complication than an integral feature of tabes. Most tabic patients
retain their mental equilibrium to the last moment; some develop
truly heroic resignation to their terrible sufferings and gloomy
prospects; and a few, becoming irritable, petulant, and abnormally
selfish, show the effect of invalidism manifested equally with other
chronic diseases.

In addition to the visceral crises there are other disturbances of

innervation of a nutritive or trophic character which are characteristic
of tabes. There is scarcely one of the eruptions or other changes
which are attributable to central nervous disease which may not
occur at some one or other period of this disease. Thus, herpetic
eruptions may occur independently or in conjunction with lancinating
pains, in one case even corresponding to the area of the belt
sensation. Discoloration of the skin or hair, the latter turning gray or
white in circular24 or in irregular patches, spontaneous falling of the
nails, spontaneous hemorrhages in the matrix of the nail,
ichthyosis,25 purpura, diffuse œdemas, localized perspiration
sometimes associated with sudamina and accompanied by
ecchymoses, muscular atrophies, bullæ, and pemphigus-like
eruptions, have been noted by various observers. As yet, they are to
be regarded rather as clinical curiosities than as constant features,
as diagnostic criteria, or as presenting special indications for
treatment. The trophic disturbances affecting the bones and joints
and a peculiar ulcerative process known as the perforating ulcer of
the foot are better studied, and, ranking among the most formidable
and striking manifestations of tabes, merit special consideration. The
tabic joint affection, first described by Charcot as a trophic
manifestation of tabes, is usually located in the hip or knee, but
almost any joint, including the temporo-maxillary articulation, may be
its seat. The earlier or lighter forms consist in serous exudations with
or without active local manifestations resembling those of synovitis;
but the commoner and graver form consists in actual deformity of the
articular surfaces, such as enlargement or atrophy, sometimes
supplemented by the formation of osteophytes. In addition,
spontaneous dislocations occur in a manner which is often quite
surprising.26 As a type of such an occurrence the following may
serve: A tabic patient presented a synovial swelling of the left knee
of truly enormous dimensions; as soon as this became reduced a
distinct crepitus was elicited, and both the femoral and tibial articular
surfaces found enlarged. The whole process occupied only a few
days; the lateral ligaments were then found so relaxed that the leg
could be freely moved in every direction.
24 Rossolymmo, Archiv für Psychiatrie, xv. p. 727. I have never observed this
symptom as beautifully developed as in a case of spinal irritation, in which the
possibility of the existence of any organic disease could be positively excluded.
Bulkley has observed albinism.

25 In three cases by Ballet et Dutil, Progrès médicale, May 19, 1882.

26 Boyer (Revue de Médecine, 1884, p. 487) records a case where the hip became
luxated spontaneously as an early symptom.

The connection between the morbid process in the spinal cord and
these remarkable arthropathies is as yet unexplained; and as
Charcot's original proposition, that they can be always referred to
lesions in the anterior horns of the gray matter, has not been
sustained, some of the German clinicians, notably Strümpell, are
inclined to attribute them, in part at least, to ordinary results and
accidents due to the anæsthesia and its disturbing effect on
voluntary and automatic joint protection. A controversy arose which
was participated in by the Clinical Society of London,27 and evoked
opinions from nearly all the eminent neurologists of Europe. The
result of this thorough discussion has been to establish the joint
affection as a true tabic symptom due to the same morbid process
which underlies the less enigmatical symptoms of the disease. One
of the best reasons for regarding these joint lesions as of trophic
origin is the fact that they are associated with textural bone-changes
by which they become either unduly soft or brittle, and therefore
exposed to apparently spontaneous fracture. Cases of fracture of
almost every long and some of the short bones are recorded, among
the most remarkable being one by Dutil,28 where fracture of the
radius occurred without adequate cause in the initial period of tabes;
and one by Krönig,29 in which the patient, who had been tabic for
eight years, broke a lumbar vertebra while catching himself in the act
of falling down stairs.
27 The question was formally raised by Morrant Baker at the December meeting in
1884, and the discussion participated in by Charcot himself. Among the opponents of
Charcot's theory were Jonathan Hutchinson and Moxon, the latter of whom
administered a sound and well-merited criticism regarding some of the premature, if
not sensational, announcements of the distinguished French neurologist. He
compared the joint lesions of tabes to decubitus: just as the latter can be avoided by a
proper protection of the exposed parts, so the former would not occur if the joints
could be kept in a, surgically speaking, normal state. Moxon seems to have forgotten,
in suggesting this comparison, that there is a form of decubitus which will occur
independently of the greatest care and in spite of every measure taken to arrest it,
and which can be attributed only to an obscure but active perversion of nervous
nutritive control. Barwell, Paget, Herbert Page, Broadbent, and McNamara agreed
that the joint lesions are not of surgical or rheumatic origin, but essentially signs of the
nervous affection. Although Barwell's claim, that the rheumatic and tabic joint
diseases are essentially different, because the former is hyperplastic and the latter is
atrophic, is not borne out by all cases of tabic joint disease, some of which are
certainly hyperplastic, yet the other reasons advanced for regarding these affections
as distinct, and considering the joint affection of Charcot as a trophic disorder, far
outweighed those advanced by the opponents of this view. In fact, the only ground the
latter had to stand on was the fact that Charcot's asserted anatomical foundation was
found to be chimerical.

28 Gazette médicale de Paris, 1885, No. 25.

 29 Zeitschrift für klinische Medizin, 1884, vii., Suppl. H.

Another reason for considering the arthropathies as of central origin

is their chronological association with other signs whose trophic
origin cannot be disputed, such as unilateral sweats and œdemas
and visceral crises. They are also found to correspond in many
cases to the lancinating pains both in location and time.

The feet of an advanced tabic patient usually appear deformed; the

extremities of the bones, particularly at the metacarpo-phalangeal
articulation, are thickened, and the axis of the foot as a whole is
bent. It is a question how far this deformity may be due to persistent
faulty innervation of the muscles going to the foot and to ensuing
mechanical influences. A most pronounced deformity found in a few
cases has been designated by Féré as the tabic foot.30 It consists in
the thickening of the tissues on the inner side of the foot, obliterating
its arch in such a manner that in standing the plantar surface is
everywhere in contact with the floor. With this the tarso-metatarsal
and the metacarpo-phalangeal joints of the great toes appear greatly
swollen.
30 Pied tabétique—Revue de Médecine, 1884, p. 473.

There is less dispute concerning the nature of the so-called

perforating ulcer of the foot. This lesion occurs only in cases where
the central or peripheral nervous apparatus is diseased; thus, it may
occur with neuritis, and it is therefore unanimously regarded as due
to morbid nervous influences. In tabes it may follow an exacerbation
of the fulminating pains. It consists in a swelling of the tissues of the
foot; the skin becomes firm and thickened; deep sinuous ulcers are
then developed, showing but slight if any tendency to granulate, and
which extend down to the bone, the latter usually remaining
unaffected.31 In a few cases more remarkable accidents due to
disturbed nutrition have been noted. Thus, J. Hoffmann records a
subcutaneous rupture of the Achilles tendon, and in another case the
spontaneous discharge of all the teeth of the upper jaw within about
a week's time.32
 31 Like most of the manifestations of advanced tabes, the perforating ulcer may occur
exceptionally as a pre-ataxic symptom. Thus, Suckling (British Medical Journal, 1885,
April 4, p. 693) mentions its preceding the development of tabes proper by one year.

32 Berliner klinische Wochenschrift, 1885, No. 12. In this unique case the loosening
and falling out of the teeth preceded the tabes by a year. Demange, who observed the
same phenomenon in two cases, found that the ascending root of the fifth pair was
involved. The only analogous observation in my experience relates, like the bleaching
of the hair in circular patches, to a case of spinal irritation: here the gums and alveolar
borders underwent atrophy, exposing the roots of the teeth to beyond the normal
alveolar border: first the right upper row, then the right lower row, and finally the teeth
on the left side in the same order, showed this condition; only the most posterior fell
out.

COURSE.—Tabes dorsalis is to be regarded as an extremely chronic

affection. Its development, as previously stated, is insidious,33 and
the symptoms demonstrating the advent of the disease may be so
slight for many years that the patient does not regard himself as ill or
as likely to become so. In this condition he may remain for ten or
more years; and there are cases recorded where the exact signs of
the incipient stage were well marked, and yet no ascertainable
progress was made in a period of such long duration that the
inference naturally follows of a much longer prospective period in
which the patient would not have become ataxic. There may be
distinguished two forms of invasion: the more frequent is marked by
the development of the terebrating, fulgurating, and fulminating
pains, to which, sooner or later, the abolition of the pupillary and
certain other deep reflexes is added. In the vast majority of cases
they are found absent when the physician's attention is directed to
the question of the existence of spinal disease. In about one-fifth of
the cases pain is at no time a prominent feature of the disease, and
may be, as is established by the cases cited, entirely absent. In
these patients the first symptoms noted are muscle-tire on slight
exertion. Usually, the evidences of disturbed sensation and
innervation begin in one extremity or preponderate in it, but
symmetry is soon established. The same is noticed in the extension
of the morbid process: when the ulnar distribution is involved—which
is the first danger-signal on the part of the cervical enlargement of
the cord—usually one side is first involved, but the other soon
follows, and becomes affected in equal or nearly equal degree. In
the overwhelming majority of cases the disease manifests itself in
the lower extremities first, remaining limited to them for a long
period, and if extending to the upper extremities doing so in much
slighter intensity than in the lower. Gull, Leyden, and others have
observed cases where the arms were first and chiefly involved; but
this is extremely rare. Cases of a simultaneous involvement of the
upper and lower extremities are more frequently noted. The
symptoms detailed above are developed in varying order in different
patients, and the line has been already indicated which separates
the progressive from the episodial features of the disease. Even in
the ataxic period the patients may remain in a comparatively
stationary period for many months and years.34 The general bodily
nutrition is usually good, and the muscular strength, as such, rarely
affected to any noteworthy extent until the extreme phase of the
disease is reached. Even the formidable-appearing crises and
trophic disturbances, though occasionally fatal, are not always so.
The direct danger from the disease proper, so far as life is
concerned, is in the possible development of cystitis and pyelitis, of
the malignant bed-sore, and of delirium and coma, the latter of which
is occasionally found to be due to cerebral hemorrhage—a lesion
which some writers regard as possibly due to some impalpable
influence of the spinal disease on the cerebral circulation.35
33 Cases of acute locomotor ataxia have been described. In no instance are they pure
or typical, and the last case of the kind published in America, beginning, as it is
evident from the account, with signs of general myelitis, illustrates the impropriety of
utilizing one or a few symptoms as determining nomenclature.

34 Thirteen patients are now under my observation, or I have been able to obtain
reliable medical information concerning their condition, who have been in the ataxic
period since the date of my first examination, varying from a year to seven years ago.
Not one of these patients is materially much worse than at that time. Two only died,
both being bedridden at the time they came under observation. Of the first group,
 three have shown more or less lasting improvement in respect to special symptoms,
as will be detailed in connection with the treatment.

35 In a case of sudden death, with an asthma-like seizure, of a tabic patient under the
charge of T. A. McBride, I found an intense injection of one (the right) vagus nucleus,
the color of the ala cinerea and of the nucleus on section being almost black.

MORBID ANATOMY AND PHYSIOLOGY.—Our knowledge of the structural

disease underlying the symptoms which comprise the clinical picture
of tabes dorsalis is perhaps the clearest in the domain of spinal
pathology. Not alone the commencement of the morbid process and
every stage of its subsequent encroachment on the spinal cord, but
also its relations to special symptoms, have been demonstrated in
numerous cases by some of the most careful German and French
investigators.

In advanced cases of tabes the disease of the spinal cord is so

evident that it was noted even by earlier observers, the region of the
cord lying between the posterior nerve-roots being altered in color,
consistency, and dimensions. It is gray instead of white, hard instead
of soft, and considerably shrunken. The gray discoloration of the
surface appears to be complete in the lumbar and lower dorsal
region; it is less intense in the upper dorsal and cervical part of the
cord; and often shows a division into three streaks, of which one, in
the middle line, occupies both columns of Goll, while on either side
another extends in the inner side of the posterior root-entry. The
posterior roots themselves are frequently atrophied, and exhibit a
firmer consistency and darker color than in health. On making cross-
sections of the spinal cord it is found that the gray discoloration
extends inward, involving more or less of the posterior columns in
well-defined areas. Microscopic examination shows that the myelin
of the nerve-tubes is either destroyed or atrophied, and the axis-
cylinder is found to be affected in the same way, usually presenting
alterations in its refraction and diameter where it is not entirely
destroyed. In the maximum foci of disease the conducting elements
of the cord are nearly destroyed, and their place is occupied by a
firm connective substance made up of wavy bundles, enclosing here
and there a few atrophied axis-cylinders with wasted myelin-sheaths.
The blood-vessels participate in the morbid process, at least as far
as the larger vessels are concerned: these are sclerotic, their lumen
appears contracted, the adventitial sheath is hypertrophied and
occupied by a nucleated, fibrillar connective substance. The
capillaries are sometimes normal, or participate in the morbid
process to the extent of a thickening of the extra-endothelial sheaths
which is either hyaline or fibrillary, and may show inflammatory cell-
proliferation. It is evident that the increase of connective tissue in the
posterior columns is not merely a relative one, due to the
condensation of the normal neuroglia after the disappearance of the
normal nerve-tubes, but is the result of actual proliferation. Opinion is
somewhat divided as to whether this is to be considered as an
inflammatory interstitial proliferation or as a more passive
development of tissue compensating for the destroyed conducting
elements. My own observations incline me to the belief that in
syphilitic tabes the morbid process is essentially interstitial, while in
non-syphilitic tabes it is parenchymatous, the disease beginning in
the nerve-tubes proper. Perhaps the controversies between
pathologists as eminent as Schultze, Tuczek, Rumpf, Zacher,
Kahler-Pick, Rosenstein, and Adamkiewicz—who are rather evenly
divided into opposing camps, one maintaining the parenchymatous,
the other the interstitial, origin of the disease—may be solved when
we learn to distinguish between the non-syphilitic and syphilitic
cases, which is clinically, as yet, impossible. The difficulty of deciding
what is a parenchymatous and what is an interstitial disease of the
cord is considerably enhanced by the fact that the neuroglia, which
by some at least is regarded as an interstitial connective tissue, is
derived, like the nerve-substance proper, from the upper or neuro-
epidermic germ-layer of the embryo, and not from the mesoblast,
which furnishes all other connective substances in the body.

It is regarded as well established that the sclerosis of the areas of

the spinal cord which are affected is preceded by a stage of granular
degeneration.36 This is supported by the facts that the sclerotic areas
are surrounded by a zone in which sclerosis has not yet developed,
but which is in a state of granular degeneration, and that in more
advanced tabes this belt is also found sclerosed. In paretic
dementia, a disease whose complicating cord affection closely
resembles that of locomotor ataxia, a granular degeneration of the
posterior columns is very common in earlier periods, while in later
periods a sclerotic tissue is found in the same locality.
36 Westphal, Archiv für Psychiatrie, ix. p. 725.

I regard these observations as opposed to the assumption of

Strümpell that tabes is a system-disease of the cord in the sense in
which he employs that term. In his fascinating and suggestive thesis
on the systemic affections he cites the observations of Tuczek made
on the posterior sclerosis ensuing from ergotism as confirming his
view that special systems of nerve-fibres are separately vulnerable
to special morbid influences. The remarkable correspondence
between the topographical distribution of the lesion of ergotin tabes
described by Tuczek and the characteristic areas of sclerosis found
in the column of Burdach in ordinary tabes, seems to justify a very
different conclusion. Ergot exerts its morbid influence through a
disturbance of nutrition. Just as it produces gangrene by constriction
of the nutritive vessels of the fingers and of whole extremities, so it
produces interference with nutrition of those nerve-elements in the
cord which are most liable to suffer from general arterial constriction.
These are presumably those portions which are supplied by the
longest and narrowest arterial stems—a condition obtaining in
precisely those parts of the cord which are affected both in ergotin
tabes and in ordinary tabes.

It has also been supposed that the morbid process began in the
posterior roots and crept in with these, thence extending upward.37
This view is opposed by the fact that there is no constant relationship
between the root lesion and the cord lesion; the sclerosis of the root-
zones within the columns of Burdach I found to be absolute in at
least one case where the outer nerve-roots were not distinctly
affected.38
37 Takacs is the most recent defender of this view.
 38 The root lesion may, like that of the column of Goll, according to a minority of the
interpreters, be a secondary process, for in ergotin tabes (Tuczek) both are usually
intact.

In typical tabes the sclerotic process begins in a special triangular

field of the posterior column in the lumbar enlargement of the cord.
The innermost of the posterior rootlets run through this field, which
has been designated by the French school as the posterior root-
zones, and the rootlets become exposed to all the vicissitudes which
the conducting tracts are destined to undergo in consequence.
Throughout the remainder of the cord it is the involvement of the
root-zone which indicates the advance of the affection upward. At a
time when the lesion is comparatively slight in the cervical level,
manifesting itself by a slight grayish or reddish streak to the naked
eye, a faint, pale discoloration in chromic-acid specimens, a deeper
stain in carminized, and a pallor in sections stained by Weigert's
method, the affected part in the lumbar region has undergone great
shrinkage; scarcely a single myelin-tube is preserved to receive
Weigert's stain within the affected area; the latter involves nearly the
entire field between the cornua, and absorbs carmine deeply, so that
it is difficult to differentiate the gray and white substance in
carminized sections. A small part of the posterior column which most
closely adjoins the posterior (gray) commissure remains free in all
cases; so also does a small laurel-leaf-shaped field bisected by the
posterior median septum. Examination with higher magnifying
powers, aided by the modern improved methods of preservation and
staining, reveals that the gray substance of the posterior segment of
the cord is usually more or less diseased. In advanced cases the
gelatinous as well as the spongy part of the posterior gray horn, and
even the posterior gray commissure, undergo such considerable
atrophy that they may occupy but half the normal area. In some
cases the columns of Clarke appear to be involved quite early in the
disease.39
39 Leyden discovered the degeneration in the fibres which in part make up these
columns and correctly traced its origin to the posterior rootlets. Lissaner and Zacher
(Archiv für Psychiatrie, xv. p. 437) confirmed him, the latter finding, as Leyden was the
 first to discriminate, that the cells may escape while the fibres are materially
diminished. Krause claims that the columns of Clarke are constantly affected, but not
necessarily in their entire length, the higher levels appearing quite normal at a time
when the lower are severely involved (Neurologisches Centralblatt, 1884, p. 50).

The triangular field in the lumbar part of the posterior column, which
is one of the typical starting-points of the affection, contains those
ascending nerve-bundles which in their cephalic course emancipate
themselves from the column of Burdach and constitute the slender
columns of Goll. The result is that the degenerative process creeps
up these columns at the same time that it ascends in the root-zones
and deep portions of Burdach's columns. Some authorities regard
this as a mere extension by contiguity;40 others incline to consider it
a secondary degeneration. It may extend to the medulla oblongata,
becoming lost in the level where the nucleus of the column of Goll
terminates, and is accompanied, at least in those advanced cases in
which the upper extremities are involved, by a comma-shaped area
of degeneration in the adjoining part of the column of Burdach, which
similarly extends into the oblongata and terminates slightly more
cephalad. In typical advanced tabes, therefore, the cross-section of
the cord exhibits a characteristic distribution of the sclerosis in each
level. As this distribution is associated with certain constant
symptoms, it is permissible to attempt bringing certain features of the
lesion in relation with special features of the disease symptoms. The
posterior gray horns and the posterior white columns, together with
other fibre-systems connected with them, are much more
complicated in structural and physiological relations than the
corresponding anterior structures. The relations of the anterior
rootlets to the gray substance, and those of the motor ganglionic
elements to their controlling tracts, are comparatively simple; those
of the posterior roots are very intricate. They run up, in great part, at
an angle to the longitudinal tracts; a few pass in directly, and still
fewer dip to a lower level. The result is that a section of the cord
made in the longitudinal direction through the root-zones, so as to
pass from the root-entry to the anterior commissure, shows the
column and root-fibres to be woven into each other like a plait.
Trabeculæ of connective tissue, dragged in as it were with the
posterior roots, fill up the interstices of this labyrinth. They are
particularly dense in the lowest part of the lumbar enlargement of the
cord, constituting the so-called posterior processi reticulares. It is
reasonable to suppose that the overlapping of ascending and
descending root-fibres, associated with the presence of an extra
amount of connective tissue, imbedded as this fibre-maze is in that
part of the cord which is most distant from its lymphatic emunctories,
affords a favorable soil for slow inflammatory trouble. This is the
primary field of tabic sclerosis, and in it the disease may remain most
intense for years, extending but slowly and with diminishing intensity
upward, hand over hand, as it were, on the natural ladder which the
intertwined fasciculi and their matrix constitute. The longitudinal
tracts which lie in and near the root-zones belong to the so-called
short fibre systems, uniting the segments of higher and lower levels
of the cord with each other. As the sclerotic process ascends it
involves the caudal ends of these systems: they consequently
undergo secondary degeneration, and, shrinking in their turn, affect
the caudal part of the next system above in the same manner. The
morbid process in the column of Burdach may therefore be
considered as a combination of inflammatory and degenerative
changes, the inflammatory products causing a series of short
ascending degenerations, and the vulnerable path thus established
being followed by a cirrhotic condition in which the connective and
vascular structures participate actively. With regard to the reasons
for regarding the degeneration of the column of Goll and that of the
comma-shaped field near it as a secondary process due to the
cutting off of its apparent nerve-supply at the caudal end, and of the
posterior nerve-roots or their provisional terminations, they may be
stated in this way: When the lesion of the primary field is limited to
the lower lumbar or sacral part of the cord, the degeneration of the
column of Goll is limited to its postero-internal part; when the upper
lumbar and lower dorsal cord is involved, the entire tract is affected;
and when the cervical portion is diseased, the supplementary
comma-shaped area degenerates. In other words, the projection
tract of the sciatic nerve, as far as it is represented in Goll's column,
suffers in the first, that of the crural nerves in the second, and that of
the brachial nerves in the third instance. In all advanced cases of
tabes the affection of the column of Goll is in direct proportion to the
altitude of the lesion in the primary field. Symptomatically, it bears an
equally constant relation to the ataxia.41 No case is on record in
which these columns were totally degenerated without some motor
inco-ordination of the lower extremity having been observed during
life; and no case is recorded in which brachial ataxia had been a
marked and persistent feature in which the comma-shaped area—
area of the column of Burdach—was healthy.
40 It is held by them that the histological character of the change of the columns of
Goll is not different from that in the column of Burdach. Zacher (Archiv für Psychiatrie,
xv. p. 435) urges that it does not resemble true secondary degeneration, beginning in
the vessels and connective substance instead of the nerve-fibres. Schultze (ibid., xiv.
p. 386), on the other hand, recognizes a primary involvement of the nerve-fibres in
both of the areas of fascicular degeneration in tabes. The observation of intact axis-
cylinders by Babinski in the sclerotic fields is in conflict with the latter's claim, and the
various differences of observation and interpretation seem to be reconcilable only on
the assumption that there are two different modes of origin, both leading to nearly the
same results and occasionally combined in one and the same case.

41 Krause's case and others show that the ataxia of movement is not influenced by
lesion of the column of Clarke; but we are not informed as to the static equilibrium of
the patients in whose cords these columns were found intensely affected.

It is scarcely necessary to seriously consider the suggestion of

Strümpell, that the lesion of the column of Goll is in relation with the
bladder disturbance.42 Cases are on record by Wolff and others
where this lesion was intense and there was little or no bladder
disturbance.43
42 Archiv für Psychiatrie, xii.

43 The column of Goll is not present in those mammals which, like the porpoise, have
no developed hind limbs, but these animals have urinary bladders.

If the disease of the column of Goll were a primary systemic affection

independent of the disease of the root-field, it would be difficult to
understand why it, as well as the likelihood of finding a
corresponding degeneration of the direct cerebellar tract, increases
with the extent to which this field is involved. This occurrence
becomes quite clear when we remember that both the direct
cerebellar tract and that of Goll, being centripetal, are under the
trophic dominion of the posterior nerve-roots. The opinion is not
distinctly expressed, but implied in some writings, that the column of
Goll degenerates because of a general transverse cord lesion at a
low level; this is not the case in the tabic cord. There is a difference
in appearance between that part of the primary field which
corresponds to the column of Goll in the lumbar cord and the
surrounding sclerosis in early cases: it is more intensely
degenerated, more homogeneous in appearance, and more evenly
stained. The other part of the triangular field presents a more
trabecular appearance. In the specimen represented in the
accompanying figure this is easily recognized: the darker field
corresponds exactly with the ascending degeneration, which follows
compression of the cauda equina,44 and is the sciatic equivalent of
the column of Goll.
 FIG. 31.

Trans-section of Upper Lumbar Cord of a Patient moderately advanced in

Tabes: f, ventral or anterior fissure; g, caput gelatinosum; r r′, entry and
deep course of the posterior rootlets; d, degenerated field, including the
origin of the tracts which in higher levels form the columns of Goll; n, field
near the posterior commissure which remains free from degeneration,
both in the lumbar and cervical cord; s, sharply marked leaf-shaped field,
bisected by posterior septum, which, as claimed by Strümpell and shown
in this case, remains free from disease; z, zone of nearly normal
consistency around it; a, diseased field, suspected to be related to the
analgesia, not usually affected in early tabes; v, fibres running up and
down in front of the gelatinous substance; *, region where the tabic
process sometimes begins.

44 I have also found that this field corresponds to the column of Goll in its myelinic
development: the lumbar part of this column—designated as such by Flechsig—is an
entirely different tract, which enjoys a remarkable immunity from disease in tabes.

As illustrating the bearing of the lesion of the column of Goll on the motor ataxia I may
refer to two cases which happen to be related side by side by Strümpell (Archiv für
Psychiatrie, xii. p. 737, Cases 1 and 2). As far as the lumbar segment of the cord is
 concerned, the distribution of the lesion is similar; but in the one presenting marked
motor ataxia the triangular field was slightly diseased, and there was no upward
extension of the lesion in the column of Goll. In the other, with marked ataxia, the
triangular field was intensely diseased, and ascending degeneration (?) occurred in
the sciatic fields of the latter.

In support of the view that the comma-shaped area in Burdach's columns is

homologous with the fibres of the column of Goll, it is to be advanced that
degeneration of this field bears the same relation to symptoms in the distribution of
the ulnar nerve that degeneration of the column of Goll bears to sciatic and crural
symptoms. Where the initial pains and subsequent tactile and locomotor disturbance
were severe, this field was found affected, and most so in the side where the
symptoms had been most intense (Friedreich-Schultze's cured cases, Archiv für
Psychiatrie, xii. p. 234). This area has no direct connection with the root-fields.
Secondly, in a primary system disease of the column of Goll, associated with
degeneration of the nucleus of the column of Goll, described by Scoli, an irregular
encroachment of the column of Burdach was noted. Third, the innermost fibres of the
column of Burdach (those belting the nucleus in the oblongata) have the same
relation to the interolivary layer which the column of Goll has through its provisional
nucleus of termination.

While the evidence of high lesion of the cerebral continuation of the column of Goll,
and, what I regard as its homologue, the comma-shaped area of Burdach, together
with the constant association of marked degeneration of these columns with motor
ataxia, is strong positive proof of its relation to this symptom, there is equally strong
evidence negativing its relation to any other of the prominent symptoms of tabes
dorsalis. Thus Babesin (Virchow's Archiv, lxxvi. p. 74) found degeneration of the
posterior columns limited to the column of Goll, and the patellar reflex was not
destroyed; the root-fields at the upper lumbar levels were intact. That the columns of
Goll have been found profoundly affected without bladder disturbance has been
stated previously, and constitutes a stronger argument against Strümpell's view than
the frequent observation of bladder trouble in spinal diseases, along with which these
columns may be entirely free.

Among the various constituents of the posterior columns which

appear to present a relative immunity to the disease, aside from the
area near the posterior commissure and the laurel-leaf-shaped area
of the posterior septum, Strümpell noted one which is situated at the
periphery of the cord, bordering on the entry-line of the posterior
roots and the inner contour of the posterior roots. In the few cases
where it was found destroyed there was, what is a comparatively
rare thing in moderately severe tabes, complete analgesia. It is not,
however, certain that there is a necessary connection between the
pathological and clinical fact here. The nerve-roots themselves are
involved within the diseased area of the root-zones. The lesion is
one of a kind which, affecting a nerve-trunk, would produce first
irritation of, then impediment to, and ultimately destruction of, its
function. The clinical parallel to this is the occurrence of the lightning-
like pains in the earlier phases of the disease, which are followed by
delayed pain-conduction, and finally by loss of sensation. Of the
rootlets or fibres subservient to the various sensory and reflex
functions mediated by the posterior roots, those which convey the
centripetal impression normally evoking the patellar jerk appear to be
the most vulnerable, or, because of their limited number, the earliest
to be destroyed or compressed, with the result of total functional
paralysis.45
45 It is now conceded that, as Westphal claimed, the patellar jerk is always abolished
when the upper lumbar level of the root-zones (bandelettes externes of Charcot) is
involved. Tshirijew has shown that the translation of the knee-jerk reflex occurs in a
single segment of the rabbit's cord at the homologous level. In transverse sections a
distinct fascicle may be seen coursing from the innermost root-fibres toward the
antero-intermediate cell-group of the anterior horn; it furnishes a pictorial substratum,
if not anatomical proof, for the patho-physiological observation, and harmonizes with
the fact that it is the innermost rootlets which usually suffer first. Perhaps the delicacy
of this tract accounts for the frequent disappearance of the jerk in old people as a
result of senile sclerosis.

With regard to the interpretation of the various tactile sensory

disturbances of tabes and the delayed pain-sense conduction little
positive advance has been made. That the lesion of the root-zones
and gray substance is responsible for them seems to be the general
conclusion of French and German observers. Recent researches
have shown that disease of the peripheral nerves (multiple neuritis)
may produce sensory phenomena which it had previously been
customary to regard as pathognomonic of disturbance of their
intraspinal terminations and continuations. At the same time, we are
confounded by the observation of Erb, that even retardation of
conduction of the pain-sense, which the dictum of Schiff taught us to
regard as a sign of disturbed function of the spinal gray matter, is
also produced by peripheral neuritis.

Immediately adjoining the gelatinous apex of the posterior horn there

is a column of vertical fibres which bear a relation to it resembling
that of the ascending root of the fifth pair in the oblongata to the
tuber cinereum of Rolando. Sclerosis of this column, as well as of
the ascending root in question (Demange), has been found in tabes,
and usually in association with pronounced trophic disturbances. In a
number of cases (Oppenheim, Eisenlohr, and others) where
spasmodic laryngeal crises had been a marked feature during life,
lesion of the floor of the fourth ventricle, or atrophy of the
pneumogastric nerve, or even of its nucleus, was found. In one case
with marked gastric crises I found sclerotic changes of the arteries in
the ala cinerea.

With regard to the involvement of the peripheral nerves proper, aside

from the optic, opinion is somewhat divided. The discovery of
multiple neuritis, and the recognition of the fact that it had been
erroneously confounded with tabes, led Dejerine to claim that tabes
might be of peripheral origin. He even proposed to account for the
oculo-motor trouble on the basis of an affection of the oculo-motor
nerves. This explanation has been repudiated by Westphal and
sound authorities generally. That the peripheral nerves are
occasionally involved in tabes was already known to Friedreich in
1863,46 and later Westphal found the cutaneous branches in an
advanced tabic patient to present similar changes to those
discovered by Friedreich: these findings are confirmed by his
Japanese pupil, Sakaky. The nerve-tubes are atrophied, the axis-
cylinder being often preserved, and the interneural connective
substance is proliferated. But there is no constant relation between
these changes and the symptoms of the disease; in one of Sakaky's
cases the nerves of an extremity which had been the seat of severe
sensory symptoms were entirely normal. The findings in the
peripheral nerves of tabic subjects lose much of their value in view of
their recent discovery in marantic persons47 advanced in life, but who
had no nervous disease whatever.
46 Virchow's Archiv, vol. xxvi. pp. 399-452.

47 Krause, Neurologisches Centralblatt, 1885, p. 53.

It has been attempted to bring the tabic process in relation to a

supposed primary meningitis. Tabes is a rather common nervous
affection, and primary spinal meningitis is one of the very rarest. The
leptomeninges are found considerably thickened in one out of ten
tabic cords, and those who defend the meningitic origin of the
disease base their theory on this inconstant finding, and allege that
in the cases where it is absent the meningitis has disappeared while
the cord lesion progressed. It is a fatal objection to this view that the
part of the posterior columns immediately adjoining the pia is often
quite free from disease. A meningitic affection, either as an
etiological or a complicating factor, can be admitted in those cases
only where there is a marginal sclerosis.

The changes in the optic nerve resemble those of the white columns
of the cord in their naked-eye and minute character as well as in the
controversial nature of the various interpretations made. When
affected, the nerve is found to be firmer than normal, and discolored;
later it becomes quite gray, and may eventually shrink to two-thirds,
and even less, of its normal diameter. It is generally believed that, as
in the cord, the myelin undergoes wasting before the axis-cylinder
disappears, and that the latter may survive a long time, thus
explaining why the patient may retain his visual power for a
considerable period after the ophthalmoscope determines the
existence of atrophy. No satisfactory explanation has as yet been
offered for the optic-nerve affection of tabes. There is no direct
continuity of the spinal and optic sclerosis. Two theoretical
possibilities suggest themselves. The first is that the lesion of the
cord exerts a remote effect upon the physiological, and through this