Professional Documents
Culture Documents
(Download PDF) Risk Based and Factor Investing 1St Edition Jurczenko Online Ebook All Chapter PDF
(Download PDF) Risk Based and Factor Investing 1St Edition Jurczenko Online Ebook All Chapter PDF
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 ...
https://textbookfull.com/product/index-fund-management-a-
practical-guide-to-smart-beta-factor-investing-and-risk-premia-
fadi-zaher/
https://textbookfull.com/product/equity-smart-beta-and-factor-
investing-for-practitioners-khalid-ghayur/
https://textbookfull.com/product/asset-management-a-systematic-
approach-to-factor-investing-1st-edition-andrew-ang/
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/
https://textbookfull.com/product/real-estate-investing-market-
analysis-valuation-techniques-and-risk-management-1st-edition-
benedetto-manganelli-auth/
https://textbookfull.com/product/investing-in-hedge-funds-a-
guide-to-measuring-risk-and-return-characteristics-1st-edition-
turban-bali/
https://textbookfull.com/product/risk-return-analysis-the-theory-
and-practice-of-rational-investing-volume-one-1st-edition-harry-
markowitz/
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:
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.
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
Emmanuel JURCZENKO
Ecole Hotelière De Lausanne, HES-SO
University of Applied Sciences Western Switzerland
September 2015
1
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
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?
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.
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.
Market cap
weight: 53% 29% 14% 4% 100%
Return
statistics:
Correlations:
HY 0.78 0.90
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%
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.
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
(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.
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.
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).
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.
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]
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.
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
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
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
max σ p2 = ∑ ∑ wi w jσ iσ j ρ ij [1.4]
{w} i j
∂σ p ∂σ p
= [1.5]
∂wi ∂w j
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
Eq Tsies IG HY MVP
(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.
The average portfolio return over the risk-free rate, the portfolio risk premium,
follows as:
rp = ∑ i∈ p wi ri [1.15]
σ p2 = ∑ i ∑ j wi w jσ ij [1.16]
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:
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 .
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.
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.
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.
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.
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.
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.
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.
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.
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.
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