-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathQuant.html
289 lines (273 loc) · 40.3 KB
/
Quant.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
<!DOCTYPE html>
<html lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<title>Chapter 5 Quantitative portfolio construction with ESG data and criteria | Perspectives in sustainable equity investing</title>
<meta name="author" content="Guillaume Coqueret">
<meta name="description" content=".container-fluid main { max-width: 55rem; font-size: 1rem; } 5.1 Simple portfolio choice solutions ESG indicators are very well suited for portfolio integration and dozens, if not hundreds of...">
<meta name="generator" content="bookdown 0.24 with bs4_book()">
<meta property="og:title" content="Chapter 5 Quantitative portfolio construction with ESG data and criteria | Perspectives in sustainable equity investing">
<meta property="og:type" content="book">
<meta property="og:description" content=".container-fluid main { max-width: 55rem; font-size: 1rem; } 5.1 Simple portfolio choice solutions ESG indicators are very well suited for portfolio integration and dozens, if not hundreds of...">
<meta name="twitter:card" content="summary">
<meta name="twitter:title" content="Chapter 5 Quantitative portfolio construction with ESG data and criteria | Perspectives in sustainable equity investing">
<meta name="twitter:description" content=".container-fluid main { max-width: 55rem; font-size: 1rem; } 5.1 Simple portfolio choice solutions ESG indicators are very well suited for portfolio integration and dozens, if not hundreds of...">
<!-- JS --><script src="https://cdnjs.cloudflare.com/ajax/libs/clipboard.js/2.0.6/clipboard.min.js" integrity="sha256-inc5kl9MA1hkeYUt+EC3BhlIgyp/2jDIyBLS6k3UxPI=" crossorigin="anonymous"></script><script src="https://cdnjs.cloudflare.com/ajax/libs/fuse.js/6.4.6/fuse.js" integrity="sha512-zv6Ywkjyktsohkbp9bb45V6tEMoWhzFzXis+LrMehmJZZSys19Yxf1dopHx7WzIKxr5tK2dVcYmaCk2uqdjF4A==" crossorigin="anonymous"></script><script src="https://kit.fontawesome.com/6ecbd6c532.js" crossorigin="anonymous"></script><script src="libs/header-attrs-2.11/header-attrs.js"></script><script src="libs/jquery-3.6.0/jquery-3.6.0.min.js"></script><meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<link href="libs/bootstrap-4.6.0/bootstrap.min.css" rel="stylesheet">
<script src="libs/bootstrap-4.6.0/bootstrap.bundle.min.js"></script><script src="libs/bs3compat-0.3.1/transition.js"></script><script src="libs/bs3compat-0.3.1/tabs.js"></script><script src="libs/bs3compat-0.3.1/bs3compat.js"></script><link href="libs/bs4_book-1.0.0/bs4_book.css" rel="stylesheet">
<script src="libs/bs4_book-1.0.0/bs4_book.js"></script><script src="libs/htmlwidgets-1.5.4/htmlwidgets.js"></script><link href="libs/leaflet-1.3.1/leaflet.css" rel="stylesheet">
<script src="libs/leaflet-1.3.1/leaflet.js"></script><link href="libs/leafletfix-1.0.0/leafletfix.css" rel="stylesheet">
<script src="libs/proj4-2.6.2/proj4.min.js"></script><script src="libs/Proj4Leaflet-1.0.1/proj4leaflet.js"></script><link href="libs/rstudio_leaflet-1.3.1/rstudio_leaflet.css" rel="stylesheet">
<script src="libs/leaflet-binding-2.0.4.1/leaflet.js"></script><script src="https://cdnjs.cloudflare.com/ajax/libs/autocomplete.js/0.38.0/autocomplete.jquery.min.js" integrity="sha512-GU9ayf+66Xx2TmpxqJpliWbT5PiGYxpaG8rfnBEk1LL8l1KGkRShhngwdXK1UgqhAzWpZHSiYPc09/NwDQIGyg==" crossorigin="anonymous"></script><script src="https://cdnjs.cloudflare.com/ajax/libs/mark.js/8.11.1/mark.min.js" integrity="sha512-5CYOlHXGh6QpOFA/TeTylKLWfB3ftPsde7AnmhuitiTX4K5SqCLBeKro6sPS8ilsz1Q4NRx3v8Ko2IBiszzdww==" crossorigin="anonymous"></script><!-- CSS -->
</head>
<body data-spy="scroll" data-target="#toc">
<div class="container-fluid">
<div class="row">
<header class="col-sm-12 col-lg-3 sidebar sidebar-book"><a class="sr-only sr-only-focusable" href="#content">Skip to main content</a>
<div class="d-flex align-items-start justify-content-between">
<h1>
<a href="index.html" title="">Perspectives in sustainable equity investing</a>
</h1>
<button class="btn btn-outline-primary d-lg-none ml-2 mt-1" type="button" data-toggle="collapse" data-target="#main-nav" aria-expanded="true" aria-controls="main-nav"><i class="fas fa-bars"></i><span class="sr-only">Show table of contents</span></button>
</div>
<div id="main-nav" class="collapse-lg">
<form role="search">
<input id="search" class="form-control" type="search" placeholder="Search" aria-label="Search">
</form>
<nav aria-label="Table of contents"><h2>Table of contents</h2>
<ul class="book-toc list-unstyled">
<li><a class="" href="index.html"><span class="header-section-number">1</span> Introduction</a></li>
<li><a class="" href="ESGdata.html"><span class="header-section-number">2</span> ESG Data</a></li>
<li><a class="" href="Investors.html"><span class="header-section-number">3</span> Investors and SRI</a></li>
<li><a class="" href="Perf.html"><span class="header-section-number">4</span> ESG investing and financial performance</a></li>
<li><a class="active" href="Quant.html"><span class="header-section-number">5</span> Quantitative portfolio construction with ESG data and criteria</a></li>
<li><a class="" href="Climate.html"><span class="header-section-number">6</span> Climate change risk</a></li>
<li><a class="" href="Equilibrium.html"><span class="header-section-number">7</span> SRI in economic equilibria</a></li>
<li><a class="" href="Conc.html"><span class="header-section-number">8</span> Conclusion</a></li>
<li><a class="" href="Bib.html"><span class="header-section-number">9</span> Bibliography</a></li>
</ul>
<div class="book-extra">
</div>
</nav>
</div>
</header><main class="col-sm-12 col-md-9 col-lg-7" id="content"><div id="Quant" class="section level1" number="5">
<h1>
<span class="header-section-number">5</span> Quantitative portfolio construction with ESG data and criteria<a class="anchor" aria-label="anchor" href="#Quant"><i class="fas fa-link"></i></a>
</h1>
<style>
.container-fluid main {
max-width: 55rem;
font-size: 1rem;
}
</style>
<div id="simple-portfolio-choice-solutions" class="section level2" number="5.1">
<h2>
<span class="header-section-number">5.1</span> Simple portfolio choice solutions<a class="anchor" aria-label="anchor" href="#simple-portfolio-choice-solutions"><i class="fas fa-link"></i></a>
</h2>
<p>ESG indicators are very well suited for portfolio integration and dozens, if not hundreds of articles have been written on this topic. We begin by mentioning a few papers that are very thoroughly written and provide the reader with a detailed account of the portfolio construction process. For instance, <span class="citation"><a href="Bib.html#ref-de2015benefits" role="doc-biblioref">De and Clayman</a> (<a href="Bib.html#ref-de2015benefits" role="doc-biblioref">2015</a>)</span> give a precise description of their methodology and report many results. They find that the best solution is to eliminate stocks from the lower tail of the ESG distribution. <span class="citation"><a href="Bib.html#ref-oikonomou2018socially" role="doc-biblioref">Oikonomou, Platanakis, and Sutcliffe</a> (<a href="Bib.html#ref-oikonomou2018socially" role="doc-biblioref">2018</a>)</span> test several quantitative methods (mean-variance, Black-Litterman, and robust estimation) and find that they work better than simple heuristic approaches (e.g., equally-weighted portfolios). <span class="citation"><a href="Bib.html#ref-jondeau2021building" role="doc-biblioref">Jondeau, Mojon, and Pereira da Silva</a> (<a href="Bib.html#ref-jondeau2021building" role="doc-biblioref">2021</a>)</span> argue that excluding a few heavy polluters is sufficient to drastically reduce the carbon footprint of benchmark portfolios. <span class="citation"><a href="Bib.html#ref-alessandrini2020esg" role="doc-biblioref">Alessandrini and Jondeau</a> (<a href="Bib.html#ref-alessandrini2020esg" role="doc-biblioref">2020</a>)</span> perform robustness checks across regions and sectors, but also compare best-in-class versus exclusion approaches. Lastly, <span class="citation"><a href="Bib.html#ref-bruder2019integration" role="doc-biblioref">Bruder et al.</a> (<a href="Bib.html#ref-bruder2019integration" role="doc-biblioref">2019</a>)</span> also look at the importance of regions and disentangle the three dimensions of ESG and their impact on returns and information ratios.</p>
</div>
<div id="improved-mean-variance-allocation" class="section level2" number="5.2">
<h2>
<span class="header-section-number">5.2</span> Improved mean-variance allocation<a class="anchor" aria-label="anchor" href="#improved-mean-variance-allocation"><i class="fas fa-link"></i></a>
</h2>
<p>One common direction that is pursued by researchers is to integrate ESG scores directly in the utility function that the investor maximizes. This makes sense because, indeed, investors value sustainability and do obtain utility outside financial gains (see <span class="citation"><a href="Bib.html#ref-bollen2007mutual" role="doc-biblioref">Bollen</a> (<a href="Bib.html#ref-bollen2007mutual" role="doc-biblioref">2007</a>)</span>, <span class="citation"><a href="Bib.html#ref-riedl2017investors" role="doc-biblioref">Riedl and Smeets</a> (<a href="Bib.html#ref-riedl2017investors" role="doc-biblioref">2017</a>)</span>, <span class="citation"><a href="Bib.html#ref-barber2018impact" role="doc-biblioref">Barber, Morse, and Yasuda</a> (<a href="Bib.html#ref-barber2018impact" role="doc-biblioref">2021</a>)</span>, <span class="citation"><a href="Bib.html#ref-dyck2019institutional" role="doc-biblioref">Dyck et al.</a> (<a href="Bib.html#ref-dyck2019institutional" role="doc-biblioref">2019</a>)</span>, <span class="citation"><a href="Bib.html#ref-hartzmark2019investors" role="doc-biblioref">Hartzmark and Sussman</a> (<a href="Bib.html#ref-hartzmark2019investors" role="doc-biblioref">2019</a>)</span>). Thus, instead of maximizing expected returns for a given level of risk, agents may seek to maximize a combination between returns and weighted-average ESG scores. This simple idea is presented in <span class="citation"><a href="Bib.html#ref-barracchini2012ethical" role="doc-biblioref">Barracchini and Addessi</a> (<a href="Bib.html#ref-barracchini2012ethical" role="doc-biblioref">2012</a>)</span> and <span class="citation"><a href="Bib.html#ref-gasser2017markowitz" role="doc-biblioref">Gasser, Rammerstorfer, and Weinmayer</a> (<a href="Bib.html#ref-gasser2017markowitz" role="doc-biblioref">2017</a>)</span>. Some variants of these methods are (in chronological order):</p>
<p>
Table: Examples of ESG-driven portfolio optimization techniques</p>
<div class="inline-table"><table class="table table-sm">
<colgroup>
<col width="50%">
<col width="50%">
</colgroup>
<thead><tr class="header">
<th>Reference</th>
<th>Contribution</th>
</tr></thead>
<tbody>
<tr class="odd">
<td><img width="250/"></td>
<td><img width="580/"></td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-drut2010social" role="doc-biblioref">Drut</a> (<a href="Bib.html#ref-drut2010social" role="doc-biblioref">2010</a>)</span></td>
<td>proposes the maximization of a mean-variance utility, subject to a minimum level of aggregate ESG score.</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-dorfleitner2012safety" role="doc-biblioref">Dorfleitner and Utz</a> (<a href="Bib.html#ref-dorfleitner2012safety" role="doc-biblioref">2012</a>)</span></td>
<td>introduce a utility function that is based on financial and sustainability returns of the portfolio.</td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-benedetti2019climate" role="doc-biblioref">Benedetti et al.</a> (<a href="Bib.html#ref-benedetti2019climate" role="doc-biblioref">2021</a>)</span></td>
<td>integrate views about carbon taxes within a Bayesian portfolio optimization problem. </td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-branch2019guide" role="doc-biblioref">M. Branch, Goldberg, and Hand</a> (<a href="Bib.html#ref-branch2019guide" role="doc-biblioref">2019</a>)</span></td>
<td>resort to optimized exclusion: they first shrink the investment set and then optimize on the remaining assets.</td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-fish2019esg" role="doc-biblioref">Fish, Kim, and Venkatraman</a> (<a href="Bib.html#ref-fish2019esg" role="doc-biblioref">2019</a>)</span></td>
<td>adjust returns with ESG metrics before they run optimizations (mean-variance and hierarchical risk parity).</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-geczy2019efficient" role="doc-biblioref">Geczy, Guerard, and Samonov</a> (<a href="Bib.html#ref-geczy2019efficient" role="doc-biblioref">2019</a>)</span></td>
<td>implement the so-called APT Mean-Variance Tracking Error at Risk optimization.</td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-alessandrini2020esg" role="doc-biblioref">Alessandrini and Jondeau</a> (<a href="Bib.html#ref-alessandrini2020esg" role="doc-biblioref">2020</a>)</span></td>
<td>maximize the ESG rating of the portfolio under many constraints (related to benchmark tracking error, turnover, regional and industry exposition, etc.).</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-bender2020reducing" role="doc-biblioref">Bender et al.</a> (<a href="Bib.html#ref-bender2020reducing" role="doc-biblioref">2020</a>)</span></td>
<td>add carbon constraints to their objective functions and show that it is possible to reduce the carbon emissions of portfolios without degrading the volatility of low-volatility portfolios. </td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-chen2020integrated" role="doc-biblioref">M. Chen and Mussalli</a> (<a href="Bib.html#ref-chen2020integrated" role="doc-biblioref">2020</a>)</span></td>
<td>propose to optimize the information ratio of a portfolio while integrating both alpha and ESG considerations.</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-chan2020esg" role="doc-biblioref">Chan et al.</a> (<a href="Bib.html#ref-chan2020esg" role="doc-biblioref">2020</a>)</span></td>
<td>optimize a traditional quadratic utility function, but with constraints on factor exposures and with additional requirements on aggregate ESG and carbon scores.</td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-schmidt2020optimal" role="doc-biblioref">Schmidt</a> (<a href="Bib.html#ref-schmidt2020optimal" role="doc-biblioref">2020</a>)</span></td>
<td>minimizes the volatility minus the ESG score under a constraint of expected return.</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-pedersen2020responsible" role="doc-biblioref">Pedersen, Fitzgibbons, and Pomorski</a> (<a href="Bib.html#ref-pedersen2020responsible" role="doc-biblioref">2021</a>)</span></td>
<td>the agent maximizes the Sharpe ratio of the portfolio for a given level of ESG score.</td>
</tr>
<tr class="even">
<td>
<span class="citation"><a href="Bib.html#ref-geczy2005investing" role="doc-biblioref">Geczy, Stambaugh, and Levin</a> (<a href="Bib.html#ref-geczy2005investing" role="doc-biblioref">2021</a>)</span> (originally written in the early 2000s)</td>
<td>focuses on allocations to mutual funds. The agent postulates a model for future returns and these models are subject to uncertainty, which is modelled in a Bayesian framework. The authors determine the financial cost of screening funds according to whether they are socially responsible or not. They provide some conditions under which this cost is high or low.</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-coqueret2021boosting" role="doc-biblioref">Coqueret et al.</a> (<a href="Bib.html#ref-coqueret2021boosting" role="doc-biblioref">2021</a>)</span></td>
<td>propose to combine ESG attributes to other variables in the asset pricing literature (e.g., market capitalization or valuation ratios) to boost ESG-based optimization.</td>
</tr>
<tr class="even">
<td><span class="citation"><a href="Bib.html#ref-alekseev2021quantity" role="doc-biblioref">Alekseev et al.</a> (<a href="Bib.html#ref-alekseev2021quantity" role="doc-biblioref">2021</a>)</span></td>
<td>hedge climate risk portfolios that mimic climate shocks. They test different definitions for these shocks and compare their mimicking portfolios to traditional sorts (narrative-based approach).</td>
</tr>
<tr class="odd">
<td><span class="citation"><a href="Bib.html#ref-fischer2021green" role="doc-biblioref">Fischer and Lundtofte</a> (<a href="Bib.html#ref-fischer2021green" role="doc-biblioref">2021</a>)</span></td>
<td>combine the DICE model (see Section <a href="Equilibrium.html#dice">7.2</a>) with Merton’s ICAPM to derive a four fund separation theorem.</td>
</tr>
</tbody>
</table></div>
<p>A very favorable feature of most of these frameworks is that they often allow closed-form solutions that are easily implementable. Henceforth, we present the framework laid out in <span class="citation"><a href="Bib.html#ref-pedersen2020responsible" role="doc-biblioref">Pedersen, Fitzgibbons, and Pomorski</a> (<a href="Bib.html#ref-pedersen2020responsible" role="doc-biblioref">2021</a>)</span>. For consistency purposes, we follow the notations of the model of <span class="citation"><a href="Bib.html#ref-pastor2020sustainable" role="doc-biblioref">Pastor, Stambaugh, and Taylor</a> (<a href="Bib.html#ref-pastor2020sustainable" role="doc-biblioref">2021b</a>)</span> (which is outlined below in Section <a href="Equilibrium.html#AP">7.1</a>.</p>
<p>The agent chooses an allocation <span class="math inline">\(w\)</span> over <span class="math inline">\(N\)</span> assets so that future wealth <span class="math inline">\(W_1=W_0(1+r_f+w'r)\)</span>, where <span class="math inline">\(r_f\)</span> is the risk-free rate and <span class="math inline">\(r\)</span> is the vector of returns of the risky assets. In addition to returns, assets are characterized by an observable ESG score g (which here has a vector form). The weighted average ESG score of the portfolio is thus <span class="math inline">\(\bar{g}=w'g/(w'1)\)</span>, where here <span class="math inline">\(1\)</span> denotes a vector of <span class="math inline">\(N\)</span> ones.<a class="footnote-ref" tabindex="0" data-toggle="popover" data-content='<p>ESG score is not equivalent to ESG risk, see, e.g., <span class="citation"><a href="Bib.html#ref-gaussel2020esg" role="doc-biblioref">Gaussel and Le Saint</a> (<a href="Bib.html#ref-gaussel2020esg" role="doc-biblioref">2020</a>)</span>.</p>'><sup>33</sup></a> As is discussed in Section <a href="Investors.html#invpref">3.1</a>, the investor has preferences over both pecuniary and social performance, and seeks to solve the following program:
<span class="math display" id="eq:meanvar1">\[\begin{equation}
\underset{w}{\max} \left\{\underbrace{w'\mu - \frac{\gamma}{2}w'\Sigma w }_{\text{mean-variance}}+\underbrace{f\left(\frac{w'g}{w'1}\right)}_{\text{ESG tilt}}, \text{ subject to } w'1>0 \right\},
\tag{5.1}
\end{equation}\]</span>
where <span class="math inline">\(\mu\)</span> and <span class="math inline">\(\Sigma\)</span> are the mean vector and covariance matrix of the returns <span class="math inline">\(r\)</span>, and <span class="math inline">\(f\)</span> is the ESG preference function. As is customary, <span class="math inline">\(\gamma>0\)</span> is the risk aversion parameter. The constraint <span class="math inline">\(w'1>0\)</span> implies that the portfolio has a long-only bias, i.e., that long positions strictly outweigh short ones. Under additional factor assumptions, <span class="citation"><a href="Bib.html#ref-varmaz2021portfolio" role="doc-biblioref">Varmaz, Fieberg, and Poddig</a> (<a href="Bib.html#ref-varmaz2021portfolio" role="doc-biblioref">2022</a>)</span> show that it is possible to simplify this program to a plain quadratic optimization with linear constraints.</p>
<p>In traditional mean-variance analysis, the agent has the choice between fixing one level of average return and minimizing the volatility, or maximizing the average return of a given level of return dispersion. In the above formulation, there is one additional choice to make, and it relates to the average ESG score <span class="math inline">\(\bar{g}\)</span>. The efficient frontier is defined by maximizing the expected return under constraints
<span class="math display" id="eq:frontier1">\[\begin{equation}
\max \left\{w'\mu, \ \text{subject to } \begin{array}{lll} w'1>0, & \text{leverage constraint}\\ \sqrt{w'\Sigma w} < \sigma_\text{target}, & \text{volatility constraint} \\ w'g > g_\text{target} & \text{ESG constraint} \end{array}\right\},
\tag{5.2}
\end{equation}\]</span>
or minimizing volatility under constraints
<span class="math display" id="eq:frontier2">\[\begin{equation}
\min \left\{ \sqrt{w'\Sigma w}, \ \text{subject to } \begin{array}{lll} w'1>0, & \text{leverage constraint}\\ w'\mu > \mu_\text{target}, & \text{return constraint} \\ w'g > g_\text{target} & \text{ESG constraint} \end{array}\right\}.
\tag{5.3}
\end{equation}\]</span></p>
<p>This implies that the frontier is a function of the target for the ESG score. This is shown in Figure , where we represent three different stylized frontiers. As the investor asks for a higher ESG score, the frontier shifts away from the unconstrained frontier and the investor must accept possibly lower returns and/or higher volatility.</p>
<p></p>
<div class="figure" style="text-align: center">
<span style="display:block;" id="fig:esgfrontier"></span>
<img src="images/ESG_frontier.png" alt="Stylized representation of ESG constrained frontiers. We present a diagram that depicts three efficient frontiers for various levels of targeted ESG score. The big dots at the intersection of lines and frontiers are the tangency portfolios" width="680px"><p class="caption">
FIGURE 5.1: Stylized representation of ESG constrained frontiers. We present a diagram that depicts three efficient frontiers for various levels of targeted ESG score. The big dots at the intersection of lines and frontiers are the tangency portfolios
</p>
</div>
<p>In a typical mean-variance optimization, the best portfolio is the one that maximizes the Sharpe ratio (tangency portfolio). It is thus useful to characterize the maximum Sharpe ratio that can be attained for a given level of aggregate ESG score:
<span class="math display">\[\text{SR}(\bar{g})= \underset{w}{\max} \left\{ \frac{w'\mu}{\sqrt{w'\Sigma w}}, \text{ subject to } \left[ \begin{array}{l} w'1>0 \\ w'g/(w'1)=\bar{g} \end{array} \right. \right\} .\]</span></p>
<p>In order to solve the original program defined in Equation <a href="Quant.html#eq:meanvar1">(5.1)</a>, the agent must choose the following level of average ESG score:
<span class="math display">\[g^*=\underset{\bar{g}}{\max} \left\{\text{SR}(\bar{g})^2 +2\gamma f(\bar{g}) \right\}.\]</span>
The general solution to Equation <a href="Quant.html#eq:meanvar1">(5.1)</a> is a fund separation that reads:</p>
<p><span class="math display" id="eq:wopt">\[\begin{equation}
w^*=\gamma^{-1} \left( \underbrace{\Sigma^{-1}\mu}_{\text{TP}} + c_{\text{MV}} \underbrace{\Sigma^{-1} 1}_{\text{MV}} + c_{\text{ESG}} \underbrace{\Sigma^{-1}g}_{\text{ESG}}\right).
\tag{5.4}
\end{equation}\]</span>
The risky portion of the portfolio thus consists of three layers: the tangency portfolio (TP) which maximizes the Sharpe ratio, the minimum variance (MV) portfolio, and a so-called ESG tangency portfolio. The scaling constants <span class="math inline">\(c_{\text{MV}}\)</span> and <span class="math inline">\(c_{\text{ESG}}\)</span> can be found in the original paper.</p>
</div>
<div id="other-quantitative-techniques" class="section level2" number="5.3">
<h2>
<span class="header-section-number">5.3</span> Other quantitative techniques<a class="anchor" aria-label="anchor" href="#other-quantitative-techniques"><i class="fas fa-link"></i></a>
</h2>
<p>Beyond simple extensions of mean-variance preferences, several papers have sought to resort to more complex optimization schemes and often solve them by means of fuzzy systems or genetic algorithms. We refer to the following list (chronologically sorted): <span class="citation"><a href="Bib.html#ref-bilbao2012fuzzy" role="doc-biblioref">Bilbao-Terol, Arenas-Parra, and Cañal-Fernández</a> (<a href="Bib.html#ref-bilbao2012fuzzy" role="doc-biblioref">2012a</a>)</span>, <span class="citation"><a href="Bib.html#ref-bilbao2012selection" role="doc-biblioref">Bilbao-Terol, Arenas-Parra, and Cañal-Fernández</a> (<a href="Bib.html#ref-bilbao2012selection" role="doc-biblioref">2012b</a>)</span>, <span class="citation"><a href="Bib.html#ref-calvo2015finding" role="doc-biblioref">Calvo, Ivorra, and Liern</a> (<a href="Bib.html#ref-calvo2015finding" role="doc-biblioref">2015</a>)</span>, <span class="citation"><a href="Bib.html#ref-bilbao2016multi" role="doc-biblioref">Bilbao-Terol et al.</a> (<a href="Bib.html#ref-bilbao2016multi" role="doc-biblioref">2016</a>)</span> and <span class="citation"><a href="Bib.html#ref-hilario2020tri" role="doc-biblioref">Hilario-Caballero et al.</a> (<a href="Bib.html#ref-hilario2020tri" role="doc-biblioref">2020</a>)</span>. In <span class="citation"><a href="Bib.html#ref-chen2021social" role="doc-biblioref">Li Chen et al.</a> (<a href="Bib.html#ref-chen2021social" role="doc-biblioref">2021</a>)</span>, the authors resort to screening methods combined to ESG score and portfolio optimization. Likewise, <span class="citation"><a href="Bib.html#ref-alessandrini2021optimal" role="doc-biblioref">Alessandrini and Jondeau</a> (<a href="Bib.html#ref-alessandrini2021optimal" role="doc-biblioref">2021</a>)</span> propose a very exhaustive optimization scheme in which the ESG score is maximized, subject to a large palette of constraints: tracking error, turnover, regional and industry weights, factor exposure, and box constraints. The only drawback is the fundamentally black-box nature of the outcome. </p>
<p>Other contributions go even further and rely on systematic approaches, often linked to machine learning. For instance, <span class="citation"><a href="Bib.html#ref-lanza2020mind" role="doc-biblioref">Lanza, Bernardini, and Faiella</a> (<a href="Bib.html#ref-lanza2020mind" role="doc-biblioref">2020</a>)</span> compute brute force trees (test all combinations of ESG indicators) to derive the best possible portfolios. Likewise, <span class="citation"><a href="Bib.html#ref-assael2021esg" role="doc-biblioref">Assael and Challet</a> (<a href="Bib.html#ref-assael2021esg" role="doc-biblioref">2021</a>)</span> resort to boosted trees. In a similar fashion, <span class="citation"><a href="Bib.html#ref-de2020esg2" role="doc-biblioref">Margot et al.</a> (<a href="Bib.html#ref-de2020esg2" role="doc-biblioref">2021</a>)</span> derives an ad-hoc ML algorithm that tries to link ESG ratings to returns in a non-linear fashion. <span class="citation"><a href="Bib.html#ref-goldberg2019sustainable" role="doc-biblioref">Goldberg and Mouti</a> (<a href="Bib.html#ref-goldberg2019sustainable" role="doc-biblioref">2019</a>)</span> resort to supervised learning to predict maximum drawdown and find that ESG scores help improve the forecasting accuracy of the algorithms. <span class="citation"><a href="Bib.html#ref-dash2021behavioral" role="doc-biblioref">Dash and Kajiji</a> (<a href="Bib.html#ref-dash2021behavioral" role="doc-biblioref">2021</a>)</span> use ML to extract pervasive ESG factors and predict expected returns, which are then fed to a hierarchical multi-objective portfolio optimization model. <span class="citation"><a href="Bib.html#ref-sokolov2021weak" role="doc-biblioref">Sokolov, Caverly, et al.</a> (<a href="Bib.html#ref-sokolov2021weak" role="doc-biblioref">2021</a>)</span> mix natural language processing with Black-Litterman views to build ESG-tilted portfolios. <span class="citation"><a href="Bib.html#ref-zhang2021socially" role="doc-biblioref">J. Zhang and Chen</a> (<a href="Bib.html#ref-zhang2021socially" role="doc-biblioref">2021</a>)</span> combine screening, ML and portfolio optimization techniques to build efficient (green) equity portfolios in the Chinese stock market.</p>
<p>Lastly, though it does not refer to portfolio construction, but performance attribution, we mention the work of <span class="citation"><a href="Bib.html#ref-bolliger2021active" role="doc-biblioref">Bolliger and Cornilly</a> (<a href="Bib.html#ref-bolliger2021active" role="doc-biblioref">2021</a>)</span>. The authors detail a methodology that aims to evaluate the carbon footprint of a portfolio. The authors contend that this can largely be driven by sectorial biases and that this dimension must be taken into account. Relatedly, <span class="citation"><a href="Bib.html#ref-lo2021quantifying" role="doc-biblioref">Lo and Zhang</a> (<a href="Bib.html#ref-lo2021quantifying" role="doc-biblioref">2021</a>)</span> outline a technique dedicated to evaluate the financial benefits (or harm) induced by sustainable investing. In their framework, the key quantity is the correlation between the <strong>impact factor</strong> (e.g., ESG score) of each stocks and their <strong>alpha</strong> (excess return in a factor model).</p>
</div>
<div id="miscellaneous-tips-methods-and-other-integration-techniques" class="section level2" number="5.4">
<h2>
<span class="header-section-number">5.4</span> Miscellaneous tips, methods and other integration techniques<a class="anchor" aria-label="anchor" href="#miscellaneous-tips-methods-and-other-integration-techniques"><i class="fas fa-link"></i></a>
</h2>
<p><span class="citation"><a href="Bib.html#ref-statman2016classifying" role="doc-biblioref">Statman and Glushkov</a> (<a href="Bib.html#ref-statman2016classifying" role="doc-biblioref">2016</a>)</span> find that the two main portfolio construction methods (ESG score screening and industry filtering) do not deliver the same results. They find value for the former but not for the latter. <span class="citation"><a href="Bib.html#ref-mohanty2021alpha" role="doc-biblioref">Mohanty, Mohanty, and Ivanof</a> (<a href="Bib.html#ref-mohanty2021alpha" role="doc-biblioref">2021</a>)</span> introduce ESG target. They construct overlays by over weighting stocks with higher ESG metrics until the portfolio improves its ESG score by 20% compared to a given benchmark. <span class="citation"><a href="Bib.html#ref-gurvich2021carbon" role="doc-biblioref">Gurvich and Creamer</a> (<a href="Bib.html#ref-gurvich2021carbon" role="doc-biblioref">2021</a>)</span> show that several carbon emission based sorts lead to various outcomes. Depending on whether portfolios are built on raw emissions (which introduces a size bias), emissions divided by market capitalization, or emissions divided by sales, the Sharpe ratios shift from good to outstanding.</p>
<p><span class="citation"><a href="Bib.html#ref-branch2012socially" role="doc-biblioref">B. Branch and Cai</a> (<a href="Bib.html#ref-branch2012socially" role="doc-biblioref">2012</a>)</span> propose an original idea: build ESG portfolios that mimick the S&P 500. They show it is possible to deliver market performance with portfolios relying only on socially responsible stocks. <span class="citation"><a href="Bib.html#ref-fan2020sustainable" role="doc-biblioref">Fan and Michalski</a> (<a href="Bib.html#ref-fan2020sustainable" role="doc-biblioref">2020</a>)</span> improve momentum and quality factors with ESG screens and sorting procedures. <span class="citation"><a href="Bib.html#ref-henriksson2019integrating" role="doc-biblioref">Henriksson et al.</a> (<a href="Bib.html#ref-henriksson2019integrating" role="doc-biblioref">2019</a>)</span> introduce an original solution to overcome issues when firms do not have ESG ratings. They advocate the creation of an ESG factor (Good minus Bad) and the evaluation of each stock’s exposure to this factor. A loading significantly above zero should mean that the firm’s returns are positively driven by the ESG factor. At the aggregate level, the aim is to tilt portfolios toward <em>good</em> ESG proxies. <span class="citation"><a href="Bib.html#ref-khan2019corporate" role="doc-biblioref">Khan</a> (<a href="Bib.html#ref-khan2019corporate" role="doc-biblioref">2019</a>)</span> proposes a new methodology to construct robust governance and ESG scores that seem to do a good job at explaining returns. <span class="citation"><a href="Bib.html#ref-palazzolo2020car" role="doc-biblioref">Palazzolo, Pomorski, and Zhao</a> (<a href="Bib.html#ref-palazzolo2020car" role="doc-biblioref">2020</a>)</span> discuss the challenges in the crafting of carbon neutral portfolios. </p>
<p>In <span class="citation"><a href="Bib.html#ref-widyawati2020systematic" role="doc-biblioref">Widyawati</a> (<a href="Bib.html#ref-widyawati2020systematic" role="doc-biblioref">2020</a>)</span> (p. 633), the author makes the case that “<em>there is little empirical knowledge about the most effective way to apply SRI as a quantitative financial model and how this application affects market equilibrium</em>.” It seems more reasonable to conclude that there is a large body of empirical work, but the problem is that conclusions of some studies often contradict the findings of other analyses. There is no consensus on how precisely to efficiently integrate ESG ingredients in a portfolio strategy that performs well financially out-of-sample.<a class="footnote-ref" tabindex="0" data-toggle="popover" data-content="<p>In fact, delivering relative outperformance is hard, even without any ESG constraints at all.</p>"><sup>34</sup></a> There is probably less uncertainty in the social good that comes from SRI, compared to its pecuniary benefits, although <span class="citation"><a href="Bib.html#ref-cappucci2018esg" role="doc-biblioref">Cappucci</a> (<a href="Bib.html#ref-cappucci2018esg" role="doc-biblioref">2018</a>)</span> warns against the perils of <strong>half measures</strong> that are only implemented for marketing purposes (see also <span class="citation"><a href="Bib.html#ref-statman2020esg" role="doc-biblioref">Statman</a> (<a href="Bib.html#ref-statman2020esg" role="doc-biblioref">2020</a>)</span> on this matter).</p>
<p>At an aggregate level, <span class="citation"><a href="Bib.html#ref-harper2020one" role="doc-biblioref">Harper</a> (<a href="Bib.html#ref-harper2020one" role="doc-biblioref">2020</a>)</span> provides some tips on how to integrate ESG criteria when searching and selecting an investment manager. <span class="citation"><a href="Bib.html#ref-schoenmaker2019investing" role="doc-biblioref">Schoenmaker and Schramade</a> (<a href="Bib.html#ref-schoenmaker2019investing" role="doc-biblioref">2019</a>)</span> draw the contours of an alternative investment paradigm for investors who seek long-term value creation. <span class="citation"><a href="Bib.html#ref-cosemans2021impact" role="doc-biblioref">Cosemans, Hut, and Dijk</a> (<a href="Bib.html#ref-cosemans2021impact" role="doc-biblioref">2021</a>)</span> specify a VAR model that includes temperature change as a predictor for (aggregate) equity returns. They estimate their model with a Bayesian method and subsequently maximize a constant relative risk aversion (CRRA) utility function based on these beliefs.
<span class="citation"><a href="Bib.html#ref-umar2020static" role="doc-biblioref">Umar, Kenourgios, and Papathanasiou</a> (<a href="Bib.html#ref-umar2020static" role="doc-biblioref">2020</a>)</span> document that ESG indices are interconnected at the global level, thereby challenging the benefits of geographical diversification. Finally, <span class="citation"><a href="Bib.html#ref-parker2021achieving" role="doc-biblioref">Parker</a> (<a href="Bib.html#ref-parker2021achieving" role="doc-biblioref">2021</a>)</span> wraps the notion of impact investing in a goal-based allocation framework to help balance financial and sustainability targets.</p>
<p>One very interesting contribution is the work of <span class="citation"><a href="Bib.html#ref-raynaud2020portfolio" role="doc-biblioref">Raynaud, Tankov, and Voisin</a> (<a href="Bib.html#ref-raynaud2020portfolio" role="doc-biblioref">2020</a>)</span>. The goal of the article is to articulate a methodology for portfolio managers who want to craft their allocation so as to participate to the global initiative toward the 2°C alignment (fixed by the 2015 Paris accord). The paper links climate metrics, macro-economics scenarios, and portfolio engineering in a non-technical and insightful fashion. For more details, the interested reader can have a peek at the <strong>alignment cookbook</strong> written by the authors. The assessment of the carbon footprint of portfolios, a material topic, is discussed in <span class="citation"><a href="Bib.html#ref-erlandsson2021carbon" role="doc-biblioref">Erlandsson</a> (<a href="Bib.html#ref-erlandsson2021carbon" role="doc-biblioref">2021</a>)</span>. </p>
<p>Finally, the first issues of the <em>Journal of Impact and ESG Investing</em> are filled with various pieces of advice. For instance, <span class="citation"><a href="Bib.html#ref-grim2020esg" role="doc-biblioref">Grim and Berkowitz</a> (<a href="Bib.html#ref-grim2020esg" role="doc-biblioref">2020</a>)</span> provide general guidance when including ESG criteria in the investment process and mention a few use cases. <span class="citation"><a href="Bib.html#ref-chan2020esg" role="doc-biblioref">Chan et al.</a> (<a href="Bib.html#ref-chan2020esg" role="doc-biblioref">2020</a>)</span> show how to use intangible value and corporate culture proxies to improve the performance of more traditional factors (value and quality).</p>
</div>
</div>
<div class="chapter-nav">
<div class="prev"><a href="Perf.html"><span class="header-section-number">4</span> ESG investing and financial performance</a></div>
<div class="next"><a href="Climate.html"><span class="header-section-number">6</span> Climate change risk</a></div>
</div></main><div class="col-md-3 col-lg-2 d-none d-md-block sidebar sidebar-chapter">
<nav id="toc" data-toggle="toc" aria-label="On this page"><h2>On this page</h2>
<ul class="nav navbar-nav">
<li><a class="nav-link" href="#Quant"><span class="header-section-number">5</span> Quantitative portfolio construction with ESG data and criteria</a></li>
<li><a class="nav-link" href="#simple-portfolio-choice-solutions"><span class="header-section-number">5.1</span> Simple portfolio choice solutions</a></li>
<li><a class="nav-link" href="#improved-mean-variance-allocation"><span class="header-section-number">5.2</span> Improved mean-variance allocation</a></li>
<li><a class="nav-link" href="#other-quantitative-techniques"><span class="header-section-number">5.3</span> Other quantitative techniques</a></li>
<li><a class="nav-link" href="#miscellaneous-tips-methods-and-other-integration-techniques"><span class="header-section-number">5.4</span> Miscellaneous tips, methods and other integration techniques</a></li>
</ul>
<div class="book-extra">
<ul class="list-unstyled">
</ul>
</div>
</nav>
</div>
</div>
</div> <!-- .container -->
<footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
<div class="col-12 col-md-6 mt-3">
<p>"<strong>Perspectives in sustainable equity investing</strong>" was written by Guillaume Coqueret. It was last built on 2022-07-18.</p>
</div>
<div class="col-12 col-md-6 mt-3">
<p>This book was built by the <a class="text-light" href="https://bookdown.org">bookdown</a> R package.</p>
</div>
</div></div>
</footer><!-- dynamically load mathjax for compatibility with self-contained --><script>
(function () {
var script = document.createElement("script");
script.type = "text/javascript";
var src = "true";
if (src === "" || src === "true") src = "https://mathjax.rstudio.com/latest/MathJax.js?config=TeX-MML-AM_CHTML";
if (location.protocol !== "file:")
if (/^https?:/.test(src))
src = src.replace(/^https?:/, '');
script.src = src;
document.getElementsByTagName("head")[0].appendChild(script);
})();
</script><script type="text/x-mathjax-config">const popovers = document.querySelectorAll('a.footnote-ref[data-toggle="popover"]');
for (let popover of popovers) {
const div = document.createElement('div');
div.setAttribute('style', 'position: absolute; top: 0, left:0; width:0, height:0, overflow: hidden; visibility: hidden;');
div.innerHTML = popover.getAttribute('data-content');
var has_math = div.querySelector("span.math");
if (has_math) {
document.body.appendChild(div);
MathJax.Hub.Queue(["Typeset", MathJax.Hub, div]);
MathJax.Hub.Queue(function() {
popover.setAttribute('data-content', div.innerHTML);
document.body.removeChild(div);
})
}
}
</script>
</body>
</html>