P2 Chp 5 Analyzing the Spread of Industry Performance

Question 1

Why should you analyze the spread of industry performance? (this is also the Learning Objective covered in the chapter

Answer 1

Because assessing whether information about the dispersion of an industry’s sustainability performance influences interpretations of a company’s performance

Question 2

When it comes to the evolution of the public understanding of ESG data, compare and contrast GHG accounting from human capital performance data

Answer 2

GHG accounting is widely understood and practiced by companies around the world, supplying generally robust data across the market.
Human capital performance continues to evolve as corporate and investor focus on human capital continues to increase.

Question 3

Users of ESG data must be able to interpret performance across a ______ of data types

Answer 3

range

Question 4

As the volume of sustainability disclosed by companies around the globe continues to _______, asset managers and other investment professionals are challenged with the task of managing the ________ of sustainability data to ensure quality inputs into the investment process

Answer 4

increase

proliferation of sustainability data

Question 5

In an information environment often characterized by fragmentation, the ability to effectively and efficiently aggregate quality ESG data of different types from different sources may present a _______ _______.

Answer 5

competitive advantage

Question 6

What is the dispersion of ESG performance data and how can it be used?

Answer 6

the natural spread or “higher and lower” results of performance of ESG to know where companies stand relative to one another or to an industry benchmark

It represents an important source of information that can be used in comparative and fundamental analysis

Question 7

What can the dispersion of company performance in an industry be shaped by? (three answers)

Answer 7

-regulatory environment
-industry-level competitive drivers
-data quality (in some cases)

Question 8

Descriptive statistics help summarize a given data set. They can generally be grouped into one of two categories. Name the two categories and the types of measures each category includes

Answer 8

Measures of central tendancy, such as:
-mean / average
-median
-mode

Measures of variability, such as:
-standard deviation
-mean absolute deviation (MAD) (average distance between each data value and the mean)
-Minimum and Maximum variables (difference between the two)

Question 9

When is normalization likely needed related to dispersion? And which descriptive statistics can provide a less subjective means of assessing dispersion?

Answer 9

When the spread of industry performance is too large to effectively interpret

Mesures of variability can provide a less subjective means to assess the spread

Question 10

Consider and interpret the following example from the Investment Banking & Brokerage industry: ISSUE CATEGORY: BUSINESS ETHICS
Professional Integrity
FN-IB-510b.3:
Total amount of monetary losses as a result of legal proceedings associated with professional integrity, including duty of care.

Companies in the industry reported the following information on this metric
COMPANY

AMOUNT OF MONETARY
LOSSES (IN MILLIONS)

REVENUE (MILLIONS)

FINES AND
SETTLEMENTS AS
A PERCENTAGE OF REVENUE
Company A $4,006 $33,565 11.94%
Company B $1,337 $23,220 5.76%
Company C $563 $34,537 1.63%
Company D $2,455 $92,342 2.66%
Company E $1,200 $83,270 1.44%
Company F $1,534 $76,100 2.02%
Company G $2,362 $61,897 3.82%
Company H $6,705 $26,787 25.03%
Company I $11 $2,228 0.50%
Company J $188 $1,171 16.02%

Mean $2,036 7.08%
MAD $1,477 6.35%

Answer 10

The average amount of monetary losses in this industry is $2.03 billion (rounded). Yet the spread of industry performance is quite wide, ranging from a minimum value of $11 million to a maximum value of $6.7 billion. The mean absolute deviation of the un-normalized
data is nearly $1.5 billion. This further indicates that this data is widely dispersed and would benefit from normalization. Normalizing the data by company revenue can provide an effective way to understand the financial impact of fines and settlements on a company.
Within the normalized data, the majority of the results are clustered in a much narrower range—between 0.5 percent and 5.76 percent—with a few outliers. Notice also that one standard deviation in the normalized dataset is much smaller than that in the un-normalized dataset. Perhaps most interestingly, Company J has gone from one performance extreme to the other. While the firm paid out a relatively small amount in terms of absolute data ($188 million in fines and settlements), this turns out to be a comparatively high percentage of its revenue when the results are normalized (the total represents more than 16 percent of its revenues). Company D, on the other hand, which paid out a relatively high absolute total (almost $2.5 billion), appears to be an above-average performer when normalized data is looked at, as its results are lower than the mean (7.08 percent) for the dataset.

Question 11

Which sustainability dimension often have fines as a metric? And what is a typical normalization activity metric for fines / why is it good to normalize with it?

Answer 11

Recall that fines, a metric that often appears in the Leadership & Governance sustainability dimension, can be normalized by revenue to better assess the cash flow implications of fine-related losses.

Question 12

Why is it true that the direct effect of fines or settlements may not be the only material impacts from managing or mismanaging the issue for invesment banking / professional integrity / total monetary losses as a result of legal proceedings associated with professional integrity including duty of care?

Answer 12

Because companies could also face material impacts or indirect costs associated with:
-potential reputational damage
-diminished customer trust
-loss of market share

Question 13

What is a quantile and how is it used?

Answer 13

A quantile divides a frequency distribution into equally sized groups, each containing the same number of observations.

Quantiles are another way to simply compare ESG performance and can be helpful in determining thresholds for good, average, and poor performance in an industry

Question 14

What are the two ways a quantile is used for comparing ESG performance?

Answer 14

1) to indicate the data value at a given quartile (e.g. “the 90th percentile is 15 million cubic meters)

2) to indicate the ranking of a given data value (e.g. Company A’s performance, 18 million cubic meters, was in the top quartile).

Question 15

What are the four most commonly used quantiles / quantile measurements?

Answer 15

-quartiles (four)
-quintiles (five)
-deciles (ten)
-percentiles (100)

Question 16

What might an analyst conclude from this quantile data?

COMPANY
EMPLOYEE
ENGAGEMENT AS A
PERCENTAGE
QUARTILE

1
A 77%
B 75%
C 68%
D 60%

2
E 55%
F 55%
G 51%
H 50%

3
I 46%
J 45%
K 41%
L 32%

4
M 31%
N 23%
O 23%
P 14%

Answer 16

using quartiles, one could divide companies that report on the metric
“employee engagement as a percentage” so that the top quartile represents the best performers. An analyst looking at this dispersion might conclude that companies in the top quartile are best at managing human capital, while those in the middle two quartiles are average, and those in the bottom are the laggards

Question 17

What is important to note about the “top” quantile?

Answer 17

The “top” quantile is not always an indicator of superior performance. E.g. companies falling into the bottom quantiles based upon GHG emissions, energy consumption or regulatory fines would be industry leaders

Question 18

What is a normal distribution / what does it look like?

Answer 18

A normal distribution is in
essence a smooth, bell-shaped graph representing the most common distribution of data—one with many datapoints concentrated in the middle (mean) of the range of performance and the remaining values trailing off symmetrically on each side. In other words, normal distribution of data represents a common range of
probabilities that one point in a dataset will take on a specific value or set of values. Most data values cluster around the average value. The farther a value is from the
mean, the less likely it is to occur.

Question 19

On a normal distribution bell curve, what is the significance of 68 percent, 95 percent and 99.7 percent? What is the significance related to these percentages for evaluating ESG performance?

Answer 19

Within a normal distribution curve, the percent (or probability) of values occurring within a range are always the same: 68 percent of values will fall within one standard deviation of the mean, 95 percent will fall within two standard deviations, and 99.7 percent will fall within three standard deviations. Interpreting data through standard deviation offers another effective way to understand companies’ relative performance on ESG issues.

Question 20

Interpret the data and standard deviations of this distribution for the Meat, Poultry & Dairy industry, Water Management disclosure topic for metric: (1) Total water withdrawn, and
(2) total water consumed percentage of each in regions with High or Extremely High baseline water stress

Ten companies in the industry report the following data:
COMPANY
PERCENTAGE OF TOTAL WATER WITHDRAWN IN REGIONS WITH HIGH OR
EXTREMELY HIGH BASELINE WATER STRESS
Company A 9%
Company B 12%
Company C 15%
Company D 17%
Company E 17%
Company F 18%
Company G 19%
Company H 22%
Company I 23%
Company J 29%

Mean 18.1%
Standard deviation 5.69

Answer 20

The mean for this dataset is 18.1 percent. One standard deviation is 5.69. Assuming normal distribution, this means that 68 percent of companies in this dataset reported a percentage of total water withdrawn in water-stressed regions that falls within the range of 12.41 percent (18.1 percent - 5.69) to 23.79 percent (18.1 + 5.69). Consider a user who is particularly focused on evaluating Company J, which has the highest rate of water withdrawn from water-stressed regions, at 29 percent. Company J seems
to be an outlier. To evaluate its performance relative to its peers, a user can apply standard deviation. This company is nearly two standard deviations above the mean:
29% - 18.1% = 10.9
10.9 ÷ 5.69 = 1.92 standard deviations
Still assuming normal distribution, this company has a greater ratio of water withdrawn from water-stressed regions than approximately 97.5 percent of its peers, accounting for a percent of the population in a normal distribution that falls below 2 (1.92 rounded) standard deviations above the mean.

Question 21

Consider the following dataset as it relates to outliers:

AUTOMOBILES INDUSTRY
TRANSPORTATION SECTOR
Disclosure Topic Sustainability Accounting Metric(s)
Product Safety TR-AU-250a.3:
Number of vehicles recalled
Ten companies in the industry might report the following data:
COMPANY
NUMBER OF VEHICLES RECALLED
Company A 593,314
Company B 3,438,476
Company C 2,258,248
Company D 6,718,698
Company E 4,300,264
Company F 943,028
Company G 1,060,617
Company H 6,179
Company I 5,357,455
Company J 339,275

Mean 2,501,555
Standard deviation 2,337,184

Answer 21

The dataset is widely dispersed, with a range from 6,179 on the low end to
6,718,698 on the high end. However, all but a few of the datapoints fall within one standard deviation of the mean. Perhaps most notably, Company H has reported a significantly smaller number of recalls than its competitors, which may lead a user of the data to consider this company to be an outlier and conclude that Company H produces extraordinarily safe automobiles.
However, such a conclusion cannot be verifiably drawn without additional
information.

Question 22

Once outliers have been identified in a given dataset, a user may ask two key questions to explain them. What are the two questions?

Answer 22

1. Is the company an outlier because it faces specific, significant risks or opportunities?
2. Does the presence of an outlier indicate that the comparison of absolute
   values reported by companies may be improper for drawing conclusions about
   their relative performance?

Question 23

When a company is an outlier because it faces specific, significant risks or opportunities, what does company-specific context often find?

Answer 23

Information that explains the extraordinary performance, whether due to internal operating factors, external operating environment, or management and governance decisions.

Question 24

If a company has outlier performance on loan default rates for companies in the Mortgage Finance industry from weather-related natural catastrophes for flooding, what would be an example of company-specific context?

Answer 24

if a company makes many loans for properties within 100-year flood zones

Question 25

When a company is an outlier because the comparison for using absolute values reported may be improper for drawing conclusions on relative performance, what must you do to improve comparability of the data?

Answer 25

Normalization

Question 26

Consider the following dataset as it relates to outliers:

AUTOMOBILES INDUSTRY
TRANSPORTATION SECTOR
Disclosure Topic Sustainability Accounting Metric(s)
Product Safety TR-AU-250a.3:
Number of vehicles recalled
Ten companies in the industry might report the following data:
COMPANY RECALLS-TO-SALES RATIO
Company A 2.02
Company B 1.94
Company C 1.00
Company D 2.40
Company E 3.08
Company F 1.34
Company G 0.96
Company H 1.20
Company I 2.68
Company J 1.03

Mean 1.77
Standard deviation 0.77

Answer 26

In this case, normalizing the data seems to have made the companies more
comparable. In addition to clustering the data more closely around the mean (i.e.,
narrowing the range of performance), it has reduced the number of outliers. Indeed, it has arguably changed our perception of which companies might be outliers.
Notice that in the normalized data, Company H’s performance does not differ dramatically from that of its peers or from the industry average.

Question 27

In many cases, ______will exist for perfectly valid reasons and may continue to exist even after normalization. For that reason, it is important to investigate why performance data falls outside the typical range of performance
when _______ are present.

Answer 27

outliers…
outliers

Question 28

When data sets are not normally distributed and are clustered around certain points in the distribution range, what are they called?

Answer 28

Asymmetrical or skewed distributions

Question 29

What is skewness a measure of?

Answer 29

Skewness is a measure of the symmetry or asymmetry of distribution in a data set. In other words, how often values fall above or below the average.

Question 30

When plotted on a distribution curve, how can you visually spot “skewness”?

Answer 30

There is a longer “tail” on one side of the curve because more results are concentrated on one end of the range.

Question 31

Describe which side of an asymmetrical distribution is negatively skewed versus positively skewed

Answer 31

If the left tail is longer along the x-axis of a distribution curve, distribution is negatively skewed. If the right tail on the x-axis is longer, it is positively skewed.

Question 32

What are three causes of skewness at the industry level?

Answer 32

• underlying financial or operational characteristics of companies;
• industry-wide regulatory conditions; and/or
• data characteristics such as performance floors and ceilings (e.g., total weight of packaging in absolute terms cannot be negative—a natural floor; and percentages cannot exceed 100 percent—a natural ceiling).

Question 33

For instance, a company’s performance output depends significantly on its ______, _______ ________, ________ ________, and other factors. (3 phrases)

Answer 33

-size
-capital intensity
-production volume

Question 34

In the example where 80 percent of companies in an industry have sales ranging from $100M to $10B, while only a small handful of companies have revenues of $100B or more…is this a negatively skewed or positively skewed distribution?

Answer 34

Positively skewed

Question 35

Regulations that impose hard and soft caps on a company’s allowable performance can explain what type of performance or data distribution?

Answer 35

It can expain asymmetrical distribution of sustainability performance in some instances

Question 36

In the following example and associated regulatory mechanisms, which metric would be the “floor” and which metric would be the “ceiling”? And which metrics would be positively versus negatively skewed?

AUTOMOBILES INDUSTRY GENERAL ISSUE CATEGORY: PRODUCT DESIGN & LIFECYCLE MANAGEMENT

Fuel Economy & Use-Phase Emissions
TR-AU-410a.1:
Sales-weighted average passenger-fleet fuel economy, by region

For this metric, companies may report using four different units of measurement based on regulatory requirements in the region(s) where sales occur: miles per gallon (mpg), kilometers per liter (km/L), grams carbon dioxide per kilometer (gCO2/km), and liters per kilometer (L/km). As a result, there may also be four different distributions for this data. Fuel efficiency and emissions regulations for automakers aim to reduce emissions through two mechanisms:
1. They place a cap on the maximum allowable emissions per distance traveled.
2. They mandate the minimum allowable distance traveled per unit of fuel used.

Answer 36

In other words, companies that report this metric in mpg or km/L must perform
above a certain “floor” for fuel efficiency. On the other hand, companies that report emissions from vehicle engines in L/km or gCO2/km face a regulatory cap, or ceiling, on maximum allowable amounts.
In the first scenario, a user finds the distribution of reported data to be heavily concentrated on the left side of the figure (just above the minimum allowable mpg or km/L values), indicating a positively skewed distribution. In the second scenario (shown at right in the figure below), the data shows the opposite picture. Companies are clustered on the right side (just below
the maximum allowable L/km or gCO2/km values), and the distribution has a long left tail or negative skew

Question 37

Describe the relative risk within the automobile industry when a firm’s sales-weighted fuel efficiency is relatively close to the cluster just under or above a mandated limit compared to being in the tail of the distribution curve.

Answer 37

When an automobile company is part of a cluster just under or just above a mandated limit—e.g., its automobiles’ sales-weighted fuel efficiency is relatively close to regulatory standards—the firm could be exposed to risks of non-compliance in the event these standards become stricter. On the other hand, when a company’s performance falls within the tail of the distribution curve, it is likely to be less exposed to risks associated with increased regulatory stringency and therefore has a competitive advantage, which could allow it to obtain pricing power and capture a larger share of the market.

Question 38

Generally speaking, in industries characterized by high levels of regulation, such a skewed distribution suggests that a significant share of companies in an industry will be exposed to regulatory risks, which can be interpreted as industry-wide or ______ risk. Companies may be required to invest significant amounts of _______ and ______ _____ to improve performance, which would result in lower margins and slower growth across the industry.

Answer 38

…or systemic risk.
…R&D and capital expenditures (CapEx)

Question 39

When considering data quality, the ability to apply tools for statistical analysis is only possible when the ESG data has both of two aspects. What are they?

Answer 39

when the ESG data:
–covers many or most companies in an industry, and
–is standardized so that it can be compared

Question 40

What is the following study indicative of?

For example, in a study researchers compiled the metrics used by 50 large, publicly listed companies when reporting on the issue of employee health and safety. They found that companies reported this information in more than 20 different ways, using different units of measure and different terminology
to communicate performance.

Answer 40

Considerations for data quality and ESG data consistency / comparability

Question 41

What are three reasons that companies produce a wide range of ESG metrics? These reasons can lead to investor-focused communications having inconsistency posing a challenge when comparing company performance

Answer 41

–support internal management
–report to regulators
–otherwise communicate to a broad range of stakeholders

Question 42

When there is inconsistency in ESG performance data and it cannot be converted to the same unit of measure, what must you do? Why does this result in different conclusions being drawn from the same data?

Answer 42

You must rely on your judgment to determine which metrics best capture performance on a topic or, if all are relevant, how to aggregate the data to capture performance. As a result, different judgments can lead to different conclusions using the same data

Question 43

When reported data is not comparable or complete, what do data aggregators such as ESG ratings and rankings agencies do?

Answer 43

When reported ESG data is not comparable or complete, however, data aggregators must develop techniques to assess the relative performance of companies. For such techniques, data aggregators will collect data structured according to a proprietary set of indicators, which may be industry agnostic or industry specific.

Question 44

When it comes to ESG ratings agencies and data aggregators, what is the labor risk management score indicative of?

Answer 44

these indicators represent providers’ best attempt to find common ground across metrics when sustainability information is not reported in a standardized way. For example, they may develop indicators such as a labor risk management score using the different qualitative and quantitative information provided by issuers to assign scores to individual companies

Question 45

Why is the scope of information used by data aggregators to rate companies being variable both within and across different sustainability topics important to understand when looking at the weightage / scoring?

Answer 45

In effect, sustainability topics may be unevenly weighted from data aggregator
to data aggregator as a result of differences in the quantity of information used to evaluate performance on a topic.

For example, one data aggregator that includes a range of human capital topics in its scoring process may base most of its analysis of human capital management on compensation and benefits performance, while another data aggregator may put more relative emphasis on labor relations

Data aggregators may additionally explicitly assign different weights to factors within their scoring models. For example, one data aggregator’s ESG scoring model may heavily weight environmental topics, while a different data aggregator may heavily weight human capital issues based on that aggregator’s view of the materiality of each issue.

Question 46

What are the three core discrepancies or differences across the ESG data landscape that combined lead to you getting a very different picture of performance for the same company when relying on different data aggregators?

Answer 46

1) different measurements and metrics
2) different scope of information
3) different weightings of ESG factors

Question 47

For companies seeking to improve its ESG scores or ratings, lack of clear
provider methodologies can create disparate data demands, which can ultimately _______ the burden of disclosure for information that is not necessarily ________ to companies’ position, performance, and outlook.

Answer 47

increase information that is not necessarily relevant

Question 48

What are two questions for analysis related to data quality of ESG performance data?

Answer 48

–Are differences in reported datapoints representative of differences in sustainability performance, or are those differences more attributable to data quality issues?
–Do peer companies report information on the same topic using different metrics?

Question 49

[CHECK FOR UNDERSTANDING] What can ESG data distribution, including normal distribution, tell a user about the overall, or typical, performance of an industry?

Answer 49

The dispersion of performance in an industry lets a user know:
* how wide or “spread out” the values in a dataset are; and
* the extent to which they can rely on assumptions for analysis related to normal distribution.
Based on these two factors, users gain a summary-level understanding of how companies across an industry generally perform, which is highly useful for interpreting relative performance.
Data distribution can be interpreted using several statistical measures; standard deviation tends to be particularly useful. Low standard deviation indicates that the values within a dataset tend to be close to the mean (i.e. companies tend to perform similarly). High standard deviation indicates that values are spread out over a wider range (i.e. companies do not tend to perform similarly)

Question 50

[CHECK FOR UNDERSTANDING]
What can ESG data distribution, including non-normal distribution, tell a user about the industry’s opportunities and risks?

Answer 50

Data distribution can provide helpful insight into the risk profile of an entire industry related to a specific ESG issue. When considering normal distribution, represented by the symmetrical bell curve, a user can summarize industry-level risk exposure related to an issue using standard deviation. Consider the example in Section 5.2., where 68 percent of companies reported a percentage of total water withdrawn
in water-stressed regions that falls within the range of 12.41 to 23.79 percent.
This means that the majority of companies are exposed at least some level of risk related to this issue, which can be interpreted as an industry-wide risk.

By identifying the source of skewed (i.e., non-normal) distribution, users can similarly glean industry-level insights related to risk exposure. Skewness of non-normalized ESG data can occur because of underlying financial or operational characteristics of companies, industry-wide regulatory conditions, and/or characteristics of the data such as natural performance floors and ceilings.
The first two of these factors are useful in understanding industry-level risk and opportunity.

Consider the example in Section 5.3.2, where energy consumption is positively correlated with company revenues. Revenue data for the industry is positively skewed, resulting in a very similar, positively skewed distribution of non-normalized energy performance data. Since the skewness of energy data is clearly correlated with total revenues, users can understand through the data that energy is a significant input for value creation on an industry level. In other words, risks associated with energy consumptions are relevant to company
performance on an industry-wide basis.

Further consider the example in Section 5.3.2., where regulatory caps and floors explain the skewness of data for companies in the Automobiles
industry. Where performance is clustered around a specific cap or floor in heavily regulated industries, skewed distribution suggests that a significant share of companies in an industry will be exposed to regulatory risks, which can be
interpreted as industry-wide or systemic risk (see Section 5.3.2.).

Question 51

[CHECK FOR UNDERSTANDING] How does a lack of comparability inhibit an analysis of ESG data distribution and dispersion?

Answer 51

Lack of comparability in ESG data can inhibit analysis through two main mechanisms: lack of data coverage and lack of standardization. Where a limited
number of companies across an industry disclose performance data on an
ESG issue, an overall lack in the availability of information prevents users from understanding and comparing performance on that issue. Where ESG data lacks standardization, companies report data on the same ESG issues using different metrics and different units of measure. Where data does not adhere to the same metrics and units of measure, it cannot be easily (or feasibly) compared