Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Deutsche Bank Markets Research
Global
Quantitative Strategy
Signal Processing
Date
19 May 2015
Macro and Micro JobEnomics
Gleaning alpha and macro insights from job postings
________________________________________________________________________________________________________________
Deutsche Bank Securities Inc.
Note to U.S. investors: US regulators have not approved most foreign listed stock index futures and options for US investors. Eligible investors may be able to get exposure through over-the-counter products. Deutsche Bank does and seeks to do business with companies covered in its research reports. Thus, investors should be aware that the firm may have a conflict of interest that could affect the objectivity of this report. Investors should consider this report as only a single factor in making their investment decision. DISCLOSURES AND ANALYST CERTIFICATIONS ARE LOCATED IN APPENDIX 1.MCI (P) 124/04/2015.
Javed Jussa
George Zhao
Yin Luo, CFA
Miguel-A Alvarez
Sheng Wang
Gaurav Rohal, CFA
Allen Wang
David Elledge
Kevin Webster
North America: +1 212 250 8983
Europe: +44 20 754 71684
Asia: +852 2203 6990
What would you do if? What would you do if you had access to over 44m unique job postings representing approximately 28,000 distinct private and public companies with over 32,000 new jobs posted daily as well as 16bn words captured in job posting descriptions?
Job postings and equity alpha In this novel research piece, we utilize the LinkUp job posting dataset in search of new sources of alpha. We find that accounting and quant factors based on the job posting data set provide incremental and uncorrelated alpha. We also find that the textual or word components in job posting descriptions are indicative of current industry trends and fads.
Job postings and macro indicators We further find that job postings are correlated to various job-centric macroeconomic indicators such as non-farm payrolls and the unemployment rate. Job postings can even assist portfolio managers with sector rotation strategies.
A rare and insightful dataset As investment managers are continually in search of new and distinct sources of alpha, we think it’s worthwhile to investigate the job posting dataset as a differentiated source of stock specific and macro alpha. We hope you enjoy the remainder of the report.
Source: gettyimages.com
19 May 2015
Signal Processing
Page 2 Deutsche Bank Securities Inc.
Table Of Contents
A letter to our readers .................................................................... 3 The monster of job postings ............................................................................................. 3
The Jobs dataset ............................................................................ 4 Introducing job postings ................................................................................................... 4 Job postings have strong coverage .................................................................................. 4 Sectors are well represented ............................................................................................ 7 Job duration is increasing ................................................................................................. 9
Job trends and fads ...................................................................... 10 Geographical trends ....................................................................................................... 10 Textual trends ................................................................................................................. 11
Job micro alpha ............................................................................ 15 Stock selection strategies and factors ............................................................................ 15 Individual strategy results ............................................................................................... 17 Overall backtesting results ............................................................................................. 19
Job macro alpha ........................................................................... 23 Macroeconomic indicators ............................................................................................. 23 Are job postings correlated to macro indices? ............................................................... 23
Forecasting non-farm payrolls ...................................................... 26 Devising systematic forecasts for non-farm payrolls ...................................................... 26 Forecasting NFP using NFP ............................................................................................ 28 The optimal NFP forecast – Bridging job postings .......................................................... 32
Sector rotation using jobs ............................................................ 35 Forecasting industry groups ........................................................................................... 35 Labor intensive industries ............................................................................................... 37
References .................................................................................... 39
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 3
A letter to our readers
The monster of job postings
Recently, job and employment numbers have become a critical and highly scrutinized
economic data point. This is because employment-related data is being closely
monitored by the FED as they continually battle with the decision of when to raise the
Federal Funds Target Rate. As such, investors are highly anticipatory and anxious of any
changes in employment-related numbers and forecasts. Investor’s attention has rotated
to closely watching any employment-related indicators.
One issue with the traditional macroeconomic employment indicators is that they tend
to be backward looking. They typically only measure whether individuals are currently
employed or not (i.e. non-farm payrolls, initial jobless claims, unemployment rate, etc).
In our opinion, what would be more applicable to investors is the outlook on future
employment prospects.
Systematically gauging future job prospects is a challenging feat which involves
gathering new datasets that are potentially predictive of future job growth. Luckily, we
have taken this challenge upon ourselves. We have obtained a new data source called
LinkUp, one of the largest job engines in the market. LinkUp crawls thousands of
company websites for job postings and other data. What better metric to gauge future
job prospects than job postings?
We start our analysis by introducing and examining the LinkUp dataset. The novel
feature about this dataset is that it has job postings for individual companies. And, as
such, we can form and test various stock-specific strategies based on this dataset. Do
company job postings contain any predictive ability in forecasting their future stock
prices?
The LinkUp dataset also provides the textual job description for each posting. We utilize
this textual component along with a natural language processing algorithms to
ascertain the current trends and fads in the job marketplace. Maybe current job trends
can shed some light on the sector or market trends?
Lastly, we aggregate company level job metrics to attempt to forecast various
macroeconomic indices including employment-based economic indicators. In summary,
this research piece introduces a new and differentiated dataset and shows how
investors can potentially utilize job postings for stock selection and global macro.
In the coming months, we will be publishing a series of reports and studies on rare yet
insightful datasets. So please stay tuned. We hope you enjoy the remainder of our
report.
Regards,
Yin, Javed, George and the quant team
Deutsche Bank Quantitative Strategy
19 May 2015
Signal Processing
Page 4 Deutsche Bank Securities Inc.
The Jobs dataset
Introducing job postings
LinkUp is one of the largest and fastest growing job search engines. The company
uniquely sources job postings data from thousands of company websites. LinkUp is
essentially the backend service provider for various job search engines. Currently, the
dataset is more focused on US companies; however, LinkUp is in the process of
expanding its coverage globally.
LinkUp’s search programs crawl the web to identify and archive job postings in real-
time from companies’ websites. LinkUp also provides a multitude of detailed data
analytics and predictive indicators. Figure 1 shows a sample of the job Meta data that is
captured by LinkUp.
LinkUp provides the job posted/removed date along with the job title, category,
location, website, as well as a unique job ID and employer/company code. LinkUp also
provides the textual job description for each job posting (see Figure 2).
Figure 1: Sample job posting
Job Posting Job Data
Date posted: October 25, 2013
Date removed: May 22, 2014
Company Name DRW Trading Group
Job Title: Software Engineer
Job Category: Software Development
City, State, Zip Code, County: Chicago, IL, 60602
Employer code: 8719
Job Category code: 152
Company URL: http://www.drwtrading.com/ Source: LinkUp
Figure 2: Sample job textual description
Job Textual Description
DRW Trading Group is a principal trading organization, meaning all of our trading is for our own account, and all of our methods, systems and applications are solely for our own use. Unlike hedge funds, brokerage firms and banks, DRW has no customers, clients or investors. Using internally developed methods, models and technology, we trade across a wide range of asset classes both domestically and internationally. Founded in 1992, our mission is to empower a team of exceptional individuals to identify and capture trading opportunities in the global markets by leveraging and integrating technology, risk management and quantitative research. With that spirit, DRW has embraced the integration of trading and technology and has devoted extensive time, capital and resources to develop fast, precise and reliable infrastructure and applications. Our technology, along with our commitment and creativity, has greatly enhanced our ability to improve and expand our operations, solve complex problems and capture new opportunities.DRW is headquartered in Chicago, and has offices in New York and London, and employs over 450 people worldwide from many different disciplines and backgrounds.DRW is looking for exceptional individuals to become part of our dynamic organization. We are seeking undergraduate and graduate students for our full-time Software Engineer positions. Prior knowledge of financial markets is not required. Please note that you must apply for positions with DRW Trading Group through your University career services website and also directly on our company website at www.drw.com.Responsibilities:In-house proprietary software design, development, and testing focusing on data processing and analysis, research systems, and post-trade analysis. Work within a large-scale grid computing environment to minimize computation times and maximize throughput of research into algorithmic trading opportunities. Design and simulate prototype algorithmic trading strategies. Implement statistical and machine learning algorithms designed to perform efficiently on large data sets Work directly with Traders, ...
Source: LinkUp
Job postings have strong coverage
Prior to integrating and utilizing a new dataset, we need to understand the basic
features of the dataset such as: coverage, sector, country or other effects that need to
be accounted for when devising alpha strategies.
LinkUp’s search programs
crawl the web to identify and
archive job postings in real-
time from companies’
websites.
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 5
Using the LinkUp dataset, job postings can be classified into two broad categories:
Jobs Created and Jobs Active. Jobs Created is simply the jobs created by a company
and found by LinkUp on a daily basis. Active jobs are the job postings that are currently
active and open (i.e. not filled or deleted).
Figure 3 shows the time series of the total number of active jobs in the LinkUp
database. Note that this includes active jobs from public as well as private companies.
The dataset commences August 1, 2007 and jobs postings are updated on a daily basis.
The active job coverage is expanding. As of the end of 2014, LinkUp had approximately
2m active job postings.
Figure 3: Total number of active jobs for all companies in LinkUp
0.0
0.5
1.0
1.5
2.0
2.5
Nu
mb
er
of
Act
ive
Jo
bs M
illio
ns
The total number of job postings within LinkUp is
evolving
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
One company can of course post multiple jobs. Figure 4 shows the time series of the
number of unique public and private companies within the dataset. Again, the company
coverage is expanding. At the end of 2014, LinkUp had approximately 14,000 unique
companies in their database.
Figure 4: Total number of unique companies within LinkUp
0
2
4
6
8
10
12
14
16
Nu
mb
er
of
Co
mp
anie
s Tho
usa
nd
s
The total number of companies within LinkUp
is expanding
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
As of the end of 2014, LinkUp
had approximately 2m active
job postings
At the end of 2014, LinkUp
had approximately fourteen
thousand unique companies
in their database
19 May 2015
Signal Processing
Page 6 Deutsche Bank Securities Inc.
The expansion in job postings and company coverage could be a result of an expanding
dataset or continued economic job growth. To account for this duality, we also
calculate the average active jobs per company. This is simply the total active jobs
divided by the total number of companies (see Figure 5). This may be more
representative of underlying economic job growth prospects. At a high level, we see
that the average active jobs per company somewhat follow or mimic the equity market.
In fact, during the financial crisis, the average active jobs per company were near an all-
time low.
Figure 5: Average number of active jobs per company in LinkUp
0
200
400
600
800
1000
1200
1400
0
20
40
60
80
100
120
140
160
Ru
sse
ll 3
00
0 I
nd
ex
Ave
rage
job
s p
er
com
pan
y
Average Active Jobs per Company Russell 3000 Index
The number of active jobs per company declined
during the Financial Crisis
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Although the LinkUp dataset contains both public and private job postings, investment
manager may be more focused on job postings associated with public companies. After
integrating the LinkUp dataset within our investment universe (i.e. the Russell 3000),
we find the company coverage to be fairly strong.
Figure 6 shows the time series coverage of companies with at least one job posting
within the LinkUp dataset. As of the end of 2014, LinkUp covered over 50% of the
companies within the Russell 3000. The company coverage continues to evolve within
the US and globally.
At a high level, we see that
the average active jobs per
company somewhat follows
or mimics the market.
After integrating the LinkUp
dataset within our investment
universe (i.e. the Russell
3000), we find the company
coverage to be fairly strong
LinkUp covers over 50% of
the companies within the
Russell 3000
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 7
Figure 6: Comparison of company coverage
0
500
1,000
1,500
2,000
2,500
3,000
3,500
Nu
mb
er
of
Co
mp
anie
s
Russell 3000 Russell 3000 companies that have >=1 active job in LinkUp
Linkup covers more than 50% of the
companies within the Russell 3000
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Sectors are well represented
Next, we explore the sector coverage of all the companies within LinkUp that are a part
of the Russell 3000 (Figure 8 and Figure 9). We find that the sector breakdown is fairly
consistent with the Russell 3000 constituents. It is important to ensure that alpha is
driven primarily by stock selection and not dominated by sector effects.
Figure 7: Number of companies by sector in LinkUp Figure 8: Number of companies by sector in Russell 3000
0
200
400
600
800
1,000
1,200
1,400
1,600
1,800
Nu
mb
er
of
Co
mp
anie
s
Telecommunication Services Utilities
Materials Energy
Consumer Staples Health Care
Industrials Financials
Consumer Discretionary Information Technology
0
500
1,000
1,500
2,000
2,500
3,000
3,500
Nu
mb
er
of
Co
mp
anie
s
Telecommunication Services Utilities MaterialsEnergy Consumer Staples Health CareIndustrials Financials Consumer DiscretionaryInformation Technology
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Figure 9 and Figure 10 show the same results but analyze the data as a percentage of
sector breakdowns.
We find that the sector
breakdown is fairly similar to
the Russell 3000 constituents
19 May 2015
Signal Processing
Page 8 Deutsche Bank Securities Inc.
Figure 9: Percentage of companies by sector in LinkUp
Figure 10: Percentage of companies by sector in Russell
3000
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Pe
r ce
nta
ge o
f C
om
pan
ies
Telecommunication Services
Utilities
Materials
Energy
Consumer Staples
Health Care
Industrials
Financials
Consumer Discretionary
Information Technology
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Pe
r ce
nta
ge o
f C
om
pan
ies
Telecommunication Services
Utilities
Materials
Energy
Consumer Staples
Health Care
Industrials
Financials
Consumer Discretionary
Information Technology
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
It’s also equally important to analyze the sector breakdown of active jobs. Figure 11
shows the sector breakdown of active jobs. Interestingly, most of the active jobs are
dominated within the consumer discretionary sector. This is expected since service
sectors are more labor intensive (Figure 12). This is an important finding to highlight.
When we design factors or strategies based on the jobs data, we must keep in mind
that depending upon the strategy; certain sectors may be favored. We discuss this in
more detail later in the report.
Figure 11: Active jobs by sector in LinkUp Figure 12: Percentage of employees by sector
0
100
200
300
400
500
600
700
800
900
1,000
Nu
mb
er
of
Co
mp
anie
s
Tho
usa
nd
s
Telecommunication Services Utilities
Materials Energy
Consumer Staples Health Care
Industrials Financials
Consumer Discretionary Information Technology
The Consumer Discretionary sector
has the most number
0%
20%
40%
60%
80%
100%
Pe
rce
nta
ge o
f Em
plo
yee
s
Information Technology Consumer Discretionary Financials
Industrials Health Care Consumer Staples
Energy Materials Utilities
Telecommunication Services
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Interestingly, most of the
active jobs are dominated
within the consumer
discretionary sector
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 9
Job duration is increasing
Another interesting metric that can be derived from the jobs dataset is duration. Job
duration is essentially the time (in days) required to fill a position.1 The average job
duration for companies within the Russell 3000 is approximately 40 days (see Figure
13). However, it seems that job duration is increasing.
Interestingly, jobs within the consumer staples sector, (a relatively stable and low job
turnover industry), take much longer to get filled whereas jobs in the financial sector (a
more cyclical and high job turnover industry) get filled much quicker (see Figure 14).
Figure 13: Overall average job duration in Russell 3000 Figure 14: Average job duration by sector in Russell 3000
0
20
40
60
80
100
120
140
160
180
Ave
rage
Jo
b D
ura
tio
n (
Cal
en
dar
Day
s)
Average Job Duration is
43 Days
0
10
20
30
40
50
60
70
80
90
100
Ove
rall
aver
age
job
dura
tion
(cal
enda
r da
ys)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
It’s also interesting to analyze the average job duration by month. Interestingly, jobs get
filled sooner during the first half of the year. Understandably, jobs take longer to get
filled during the summer months and towards the end of the year. So, if you are looking
for a new job, its best to kick start your search at the beginning of the year.
Figure 15: Percentage of active jobs by sector in DBEQS universe
0
10
20
30
40
50
60
Ave
rage
job
du
rati
on
(ca
len
dar
day
s)
Jobs take longer to get filled during holiday
and budgeting months
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
1 Note that our calculation of job duration does not distinguish between a job being filled and a job being deleted
because it could not be filled.
It takes on an average forty
days to fill a job
Jobs within the consumer
staples’ sector take much
longer to get filled whereas
jobs in the financial sector get
filled much quicker
Understandably, jobs take
longer to get filled during the
summer months and at the
end of the year
19 May 2015
Signal Processing
Page 10 Deutsche Bank Securities Inc.
Job trends and fads
Geographical trends
As discussed earlier, the LinkUp jobs dataset also contains the location of the job (i.e. the
state). We can utilize this information to glean some geographical information on job
prospects. To start, we can simply chart the number of active jobs by state (Figure 16)2.
This should be somewhat correlated to the population of each state. In fact, we find
that the most populous states such as California, Texas, Florida, New York and Illinois
have the most number of active jobs. The next logical question to ask is which states
show the most significant job growth prospects?
Figure 16: Average active jobs per state from 2010 to 2014
AL
AZ
AR
CA
CO
CT
DE
FL
GA
ID
IL IN
IA
KS
KY
LA
ME
MD
MAMI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
2e+05
4e+05
6e+05
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Figure 17 shows the active job growth from 2010 to 2014 by state. Essentially, this
shows the number of jobs that have increased or decreased on a percentage basis over
the past five years.
Interestingly, we see that job growth during the past five years has been primarily
isolated in oil and agricultural rich states such as Wyoming, North Dakota, Montana,
Oklahoma, Kansas, and Utah.
2 Note that these are public company jobs that are a part of the Russell 3000 index.
In fact, we find that the most
populous states such as
California, Texas, Florida,
New York and Illinois have
the most number of active
jobs
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 11
Figure 17: Active job growth per state from 2010 to 2014
AL
AZ
AR
CA
CO
CT
DE
FL
GA
ID
IL IN
IA
KS
KY
LA
ME
MD
MAMI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
300
400
500
600
700
800
Percentage
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Textual trends
As discussed earlier, the LinkUp dataset also contains the textual or word descriptions.
This is intriguing for us as we can mine the textual job descriptions using natural
language processing algorithms. This can potentially unravel key industry themes,
trends and fads. The coverage of job descriptions is evolving. The vast majority of job
descriptions contained in the LinkUp dataset commence from 2014 onwards (see
Figure 18). Again, majority of the job descriptions are coming from the consumer
discretionary sector.
We can mine the textual job
descriptions using natural
language processing
algorithms
19 May 2015
Signal Processing
Page 12 Deutsche Bank Securities Inc.
Figure 18: Coverage of textual job description by sector
0
50
100
150
200
250
300
350
400
January-11 January-12 January-13 January-14
Nu
mb
er
of
Job
De
scri
pto
ns
Tho
usa
nd
s
Utilities Materials
Energy Telecommunication Services
Consumer Staples Financials
Information Technology Health Care
Industrials Consumer Discretionary
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
After running the job descriptions through various language processing algorithms, we
uncover some interesting themes. First, focusing on the information technology sector,
some of the common technology themes we find are: cloud computing, big data, CRM,
SAAS (software as a service), data analysis, agile development and social media (see
Figure 19 and Figure 20). Interestingly, these are the current common trends occurring
within the tech sector.
It is also worthwhile to point out the increased focused on employee diversity within
the technology sector. Words such as females, EEO affirmative (equal opportunity and
affirmative), and sponsorship visa also occur frequently in job postings.
Figure 19: Key words within the tech sector by
occurrences
Figure 20: Visualize key words within the tech sector
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
Pe
rce
nta
ge o
f O
curr
en
ces
Information Technology Sector
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Next, we focus on the financial sector. Interestingly, words like mortgage, risk
management, regulation, minorities, proven track record, females, and EEO statement
are fairly popular.
Common words include:
cloud computing, big data,
CRM, SAAS, females, EEO
affirmative…
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 13
Figure 21: Key words within the financial sector by
occurrences
Figure 22: Visualize key words within the financial sector
0.0%
5.0%
10.0%
15.0%
20.0%
Pe
rce
nta
ge o
f O
ccu
rnce
s
Financial Sector
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Transitioning to the energy sector, we find that the more popular words include: ESG
(environmental, social, and governance), hydrocarbons, shale, shale-oil, and regulations
(see Figure 23 and Figure 24). Again, this is in-line with the current industry trends
within the energy sector.
Figure 23: Key words within the energy sector by
occurrences
Figure 24: Visualize key words within the energy sector
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
Pe
rce
nta
ge o
f O
ccu
rnce
s
Energy Sector
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Analyzing the job descriptions for companies within the entire Russell 3000 universe
unveils that many company are seeking employees with extensive job experience, team
oriented, educated (with a bachelors degree) and innovated (see Figure 25 and Figure
26). The job posting data also reveals that most companies are equal opportunity
employers.
19 May 2015
Signal Processing
Page 14 Deutsche Bank Securities Inc.
Figure 25: Key words within all sectors by occurrences Figure 26: Visualize key words within all sector
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
Pe
rce
nta
ge o
f O
ccu
rnce
s
All Sectors
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Undoubtedly, more time series textual data would enable us to test whether word
trends in job posting have some alpha selection ability. For example, sector rotation
strategies may be correlated to job posting word trends. As the various job related
datasets evolve and expand, we hope to do more of this type of research in the future.
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 15
Job micro alpha
Stock selection strategies and factors
Aside from the trends and fads in jobs, investment managers are concerned as to
whether job posting related data contains any stock selection and return prediction
ability. In this section, we explore in detail whether jobs posting data contains any
untapped alpha.
To accomplish this, we form equity strategies or quant factors based on the jobs
posting data. There are a numerous potential strategies that we can create based on the
job posting data. We bucket these strategies into the following categories:
Job Creation Factors: We develop various factors based on the number of jobs
created per company within a month or year time frame. Since a larger
company will likely create more jobs, we scale our job created factor by market
cap, number of employees, total sales, total assets, and total earnings,
respectively.3
Growth in Job Creation: We develop various factors based on job creation
growth. Essentially, we calculate the quarterly and annual growth in job
creation per company. Additionally, we compare job creation growth versus
earnings growth. The rationale is that if a company is hiring faster than its
earnings growth, this could be a warning sign. We also compare job creation
growth to growth in sales, total assets, and market cap.
Job Active Factors: We develop various factors based on the number of active
jobs per company within a month or year time frame. Since a larger company
will likely have more active jobs, we again scale our job created factor by
market cap, number of employees, total sales, total assets, and total earnings,
respectively.
Growth in Active Jobs: We develop various factors based on active job growth.
Essentially, we calculate the quarterly and annual growth in active jobs per
company. Additionally, we compare active job growth versus earnings growth.
Again, the rationale is that if a company has faster growth in active jobs than
earnings or sales growth, this could be a red flag. We also compare active job
growth to growth in sales, total assets, and market cap.
Employment Growth: We develop various factors based on the number of total
employees per company. 4 Since a larger company will likely have more
employees, we again scale the number of employees by market cap, total
sales, total assets, and total earnings, respectively. We also look at the year-
over-year growth in the number of employees.
Figure 27 provides a summary of all the factors we created. Next, we form equally
weighted long/short portfolios based on these factors. We backtest these portfolios
over a six-year period using monthly rebalancing. All the factors are backtested over the
same period so that portfolio returns across all strategies are comparable.
3 Note we also used an OLS regression to neutralize for size effects. These factors are denoted with a *.
4 Note that the employee data is obtained from Compustat.
We explore in detail whether
jobs posting data contains any
untapped alpha
We develop various factors
based on job creation growth.
We develop various factors
based on the number of
active jobs per company
within a month or year time
frame
We develop various factors
based on the number of total
employees per company
19 May 2015
Signal Processing
Page 16 Deutsche Bank Securities Inc.
Figure 27: Potential job alpha factors
Description Factor Groups Backtesting Months Description
Employee Number/Maket Cap 72 EMP_to_MKTCAP
Employee Number/Total Assets 72 EMP_to_ATQ
Employee Number/Total Income 72 EMP_to_INCOME
Monthly Created Jobs /Number of Employees 72 JOBS_CREATED_MON_TO_EMP
Annual Employee Growth 72 EMP_YR_GROWTH
Annually Created Jobs/Number of Employees 72 JOBS_CREATED_YR_TO_EMP
Monthly Active Jobs /Number of Employees 72 JOBS_ACTIVE_MON_TO_EMP
Employee Number/Total Sales 72 EMP_to_SALES
Annually Active Jobs/Number of Employees 72 JOB_ACTIVE_YR_TO_EMP
Monthly Active Jobs/Total Asset 72 Monthly Active Jobs/Total Asset
Monthly Active Jobs /Number of Employees 72 Monthly Active Jobs /Number of Employees
Monthly Active Jobs/Net Income 72 Monthly Active Jobs/Net Income
Monthly Active Jobs/Market Cap 72 Monthly Active Jobs/Market Cap
Monthly Active Jobs/Total Sales 72 Monthly Active Jobs/Total Sales
Annually Active Jobs/Total Assets 72 Annually Active Jobs/Total Assets
Annually Active Jobs/Number of Employees 72 Annually Active Jobs/Number of Employees
Annually Active Jobs/Net Income 72 Annually Active Jobs/Net Income
Annually Active Jobs/Market Cap 72 Annually Active Jobs/Market Cap
Annually Active Jobs/Total Sales 72 Annually Active Jobs/Total Sales
Quaterly Active Job Number Growth 72 Quaterly Active Job Number Growth
Annually Active Job Number Growth 72 Annually Active Job Number Growth
Active Job Growth Quaterly - Total Quaterly Asset Growth 72 Active Job Growth Quaterly - Total Quaterly Asset Growth
Active Job Growth Annually - Total Annually Asset Growth 72 Active Job Growth Annually - Total Annually Asset Growth
Active Job Growth Quaterly - Net Income Quaterly Growth 72 Active Job Growth Quaterly - Net Income Quaterly Growth
Active Job Growth Annually - Net Income Growth Annually 72 Active Job Growth Annually - Net Income Growth Annually
Active Job Growth Quaterly - Market Cap Quaterly Growth 72 Active Job Growth Quaterly - Market Cap Quaterly Growth
Active Job Growth Annually - Market Cap Annually Growth 72 Active Job Growth Annually - Market Cap Annually Growth
Active Job Growth Quaterly - Total Sales Quaterly Growth 72 Active Job Growth Quaterly - Total Sales Quaterly Growth
Active Job Growth Annually - Total Sales Annually Growth 72 Active Job Growth Annually - Total Sales Annually Growth
Active Job Growth Annually - Total Asset Quaterly Growth 72 Active Job Growth Annually - Total Asset Quaterly Growth
Active Job Growth Annually - Net Income Quaterly Growth 72 Active Job Growth Annually - Net Income Quaterly Growth
Active Job Growth Annually - Market Cap Quaterly Growth 72 Active Job Growth Annually - Market Cap Quaterly Growth
Active Job Growth Annually - Total Sales Quaterly Growth 72 Active Job Growth Annually - Total Sales Quaterly Growth
Monthly Created Jobs/Total Asset 72 Monthly Created Jobs/Total Asset
Monthly Created Jobs /Number of Employees 72 Monthly Created Jobs /Number of Employees
Monthly Created Jobs/Net Income 72 Monthly Created Jobs/Net Income
Monthly Created Jobs/Market Cap 72 Monthly Created Jobs/Market Cap
Monthly Created Jobs/Total Sales 72 Monthly Created Jobs/Total Sales
Annually Created Jobs/Total Assets 72 Annually Created Jobs/Total Assets
Annually Created Jobs/Number of Employees 72 Annually Created Jobs/Number of Employees
Annually Created Jobs/Net Income 72 Annually Created Jobs/Net Income
Annually Created Jobs/Market Cap 72 Annually Created Jobs/Market Cap
Annually Created Jobs/Total Sales 72 Annually Created Jobs/Total Sales
Quaterly Created Job Number Growth 72 Quaterly Created Job Number Growth
Annually Created Job Number Growth 72 Annually Created Job Number Growth
Created Job Growth Quaterly - Total Quaterly Asset Growth 72 Created Job Growth Quaterly - Total Quaterly Asset Growth
Created Job Growth Annually - Total Annually Asset Growth 72 Created Job Growth Annually - Total Annually Asset Growth
Created Job Growth Quaterly - Net Income Quaterly Growth 72 Created Job Growth Quaterly - Net Income Quaterly Growth
Created Job Growth Annually - Net Income Growth Annually 72 Created Job Growth Annually - Net Income Growth Annually
Created Job Growth Quaterly - Market Cap Quaterly Growth 72 Created Job Growth Quaterly - Market Cap Quaterly Growth
Created Job Growth Annually - Market Cap Annually Growth 72 Created Job Growth Annually - Market Cap Annually Growth
Created Job Growth Quaterly - Total Sales Quaterly Growth 72 Created Job Growth Quaterly - Total Sales Quaterly Growth
Created Job Growth Annually - Total Sales Annually Growth 72 Created Job Growth Annually - Total Sales Annually Growth
Created Job Growth Annually - Total Asset Quaterly Growth 72 Created Job Growth Annually - Total Asset Quaterly Growth
Created Job Growth Annually - Net Income Quaterly Growth 72 Created Job Growth Annually - Net Income Quaterly Growth
Created Job Growth Annually - Market Cap Quaterly Growth 72 Created Job Growth Annually - Market Cap Quaterly Growth
Created Job Growth Annually - Total Sales Quaterly Growth 72 Created Job Growth Annually - Total Sales Quaterly Growth
Employee Factors
Active Job Factors
Active Job Growth Factor
Job Creation Factor
Job Creation Growth Factor
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Before comparing the performance of all the factors, we examine the backtesting
results of a select few job factors to better understand the return structure.
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 17
Individual strategy results
We start by analyzing the performance results for a specific factor: active jobs within
the month scaled by market cap. Figure 28 shows the factor coverage which is
expanding. More than half the stocks in the Russell 3000 are included within the
LinkUp dataset. Figure 29 shows the factor quantile values. It shows that the long leg of
the portfolio typically includes companies that have between 4% -10%percent of active
jobs as a function of market cap.
Figure 28: Coverage - jobs created per year/market cap Figure 29: Quantiles - jobs created per year/market cap
2009 2010 2011 2012 2013 2014 2015
0
500
1000
1500
2000
Coverage of JOBS_ACTIVE_MON_TO_MKTCAP
# o
f sto
cks
# of stocks12-month moving average
Quantiles
Qu
an
tile
Va
lue
2009 2010 2011 2012 2013 2014 2015
0.00
0.02
0.04
0.06
0.08
0.10
0.1220%40%
60%80%
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Figure 30 shows the time series long/short monthly return of the job strategy. The
performance is fairly strong with a monthly return of approximately 50 bps; albeit the
strategy showed strong performance in 2009. Figure 31 shows the returns of each
quintile portfolio. The return pattern is fairly monotonic meaning that companies with
more hiring (adjusted for market cap) tend to show better performance.
Figure 30: Long/short spread - jobs created per
year/market cap
Figure 31: Quintile returns - jobs created per year/market
cap
2009 2010 2011 2012 2013 2014 2015
-5
0
5
10
Spread
Sp
rea
d (
%)
2009 2010 2011 2012 2013 2014 2015
-5
0
5
10
Spread (%), Ascending order12-month moving average
Avg = 0.51 %
Std.Dev = 2.59 %Min = -6.62 %Max = 9.93 %
Avg/Std.Dev = 0.2 %
1 2 3 4 5 L/S
Fractile Portfolio Annualized Returns (%)
Fra
ctile
Po
rtfo
lio
An
nu
alize
d R
etu
rns (
%)
05
10
15
20
25
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
More than half the stocks in
the Russell 3000 are included
within the LinkUp dataset
The performance is fairly
strong with a monthly return
of approximately 50 bps
19 May 2015
Signal Processing
Page 18 Deutsche Bank Securities Inc.
Figure 32 shows the strategy’s cumulative performance. We also find that the strong
performance mostly comes from two periods: 2009 and 2012 to 2014. Lastly, the monthly
two-way turnover of the strategy is modest at approximately 50%5 (see Figure 33). This is
a good finding as it implies that trading costs should not be too impactful on the strategy.
Figure 32: Wealth - jobs created per year / market cap Figure 33: Turnover - jobs created per year / market cap
2009 2010 2011 2012 2013 2014 2015
1.0
1.1
1.2
1.3
1.4
1.5
1.6
Long-Short Fractile Portfolio
L/S
Fra
ctile
Po
rtfo
lio
2009 2010 2011 2012 2013 2014 2015
1.0
1.1
1.2
1.3
1.4
1.5
1.6 Long-Short Fractile Portfolio
2009 2010 2011 2012 2013 2014 2015
0
20
40
60
80
100
120
140
Turnover L/S
Tu
rno
ve
r (%
)
2009 2010 2011 2012 2013 2014 2015
0
20
40
60
80
100
120
140 Turnover L/S12-month moving average
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
We also briefly analyze the performance results for active job growth factors. To form
this factor, we calculate the annual growth in active jobs per company. Next, we
compare active job growth versus market cap growth. Again, the rationale is that if a
company has faster growth in active jobs than its market cap, this could be a warning
sign. The empirical results confirm our intuition. Figure 34 and Figure 35 show that
companies that grow their active jobs faster than their market cap tend to
underperform. Note that a portfolio strategy that utilizes this factor would essentially
underweight or short companies where job growth outpaces market cap growth.
Figure 34: Quintile returns – active annual jobs growth
minus market cap growth
Figure 35: Wealth – active annual jobs growth minus
market cap growth
1 2 3 4 5 L/S
Fractile Portfolio Annualized Returns (%)
Fra
ctile
Po
rtfo
lio
An
nu
alize
d R
etu
rns (
%)
05
10
15
20
25
2009 2010 2011 2012 2013 2014 2015
0.8
0.9
1.0
1.1
Long-Short Fractile Portfolio
L/S
Fra
ctile
Po
rtfo
lio
2009 2010 2011 2012 2013 2014 2015
0.8
0.9
1.0
1.1
Long-Short Fractile Portfolio
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
5 Note that the maximum turnover is 400%.
The turnover of the strategy is
fairly low at around 50%.
The rationale is that if a
company has faster growth in
active jobs than its market
cap, this could be a red flag
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 19
Now that we have briefly glimpsed the performance results for two job factors, next we
analyze the performance of all the job factors and compare their performance to
traditional quantitative factors.
Overall backtesting results
Figure 36 shows the monthly return of the job factors alongside the traditional quant
factors. The results are promising. The job-based factors stack up fairly well compared
to traditional quant factors backtested over the same period. In fact, most of the
employee, job creation, and active job-based factors showed good relative performance.
Job growth factors showed moderate performance as well.
Figure 36: Long/short monthly quintile return spread of traditional and job factors
0.00%
0.20%
0.40%
0.60%
0.80%
1.00%
Pri
ce/S
ales
, [D
]Sa
les/
EV, [
A]
Emp
loye
e N
um
ber
/Mak
et C
ap, [
A]
Ass
et T
urn
ove
r, [
A]
Targ
et p
rice
imp
lied
ret
urn
, [A
]To
tal r
etu
rn, 1
D, [
D]
Emp
loye
e N
um
ber
/To
tal A
sset
s, [
A]
Pri
ce/E
arn
ings
, [D
]P
rice
-to
-FY0
EP
S, [
D]
Earn
ings
yie
ld, F
Y0, [
A]
Emp
loye
e N
um
ber
/To
tal I
nco
me,
[A
]C
ash
flo
w y
ield
, FY0
, [A
]To
tal r
etu
rn, 1
26
0D
(6
0M
), [
D]
Cap
ex t
o D
ep, [
D]
Mo
nth
ly A
ctiv
e Jo
bs/
Mar
ket
Cap
, [A
]Ea
rnin
gs y
ield
, fo
reca
st F
Y2 m
ean
, [A
]Ea
rnin
gs y
ield
, fo
reca
st F
Y1 m
ean
, [A
]M
on
thly
Cre
ated
Jo
bs/
Mar
ket
Cap
, [A
]To
tal r
etu
rn, 2
52
D (
12
M),
[D
]R
etu
rn o
n E
qu
ity,
[A
]A
nn
ual
ly A
ctiv
e Jo
bs/
Mar
ket
Cap
, [A
]A
nn
ual
ly C
reat
ed J
ob
s/M
arke
t C
ap, [
A]
Pri
ce/B
oo
k, [
D]
Wee
kly
Tota
l Ret
urn
, [D
]EB
ITD
A t
o E
V, [
A]
Cre
ated
Jo
b G
row
th A
nn
ual
ly -
Mar
ket
Cap
An
nu
ally
Gro
wth
, [D
]A
nn
ual
ly C
reat
ed J
ob
s/M
arke
t C
ap*,
[A
]R
eco
mm
end
atio
n, m
ean
, [D
]A
ctiv
e Jo
b G
row
th A
nn
ual
ly -
Mar
ket
Cap
An
nu
ally
Gro
wth
, [D
]R
etu
rn o
n In
vest
ed C
apit
al, [
A]
An
nu
ally
Act
ive
Job
s/M
arke
t C
ap*
, [A
]R
etu
rn o
n A
sset
s, [
A]
Mo
nth
ly C
reat
ed J
ob
s/M
arke
t C
ap*,
[A
]N
orm
aliz
ed a
bn
orm
al v
olu
me,
[A
]M
on
thly
Act
ive
Job
s/M
arke
t C
ap*,
[A
]A
ctiv
e Jo
b G
row
th A
nn
ual
ly -
Net
Inco
me
Gro
wth
An
nu
ally
, [D
]1
2M
-1M
to
tal r
etu
rn, [
D]
Act
ive
Job
Gro
wth
qu
arte
rly
-M
arke
t C
ap q
uar
terl
y G
row
th, [
D]
IBES
LTG
Mea
n E
PS
Rev
isio
n,
1M
, [A
]IB
ES F
Y1 E
PS
dis
per
sio
n, [
A]
EPS
Gro
wth
, [A
]A
nn
ual
ly C
reat
ed J
ob
Nu
mb
er G
row
th, [
D]
IBES
LTG
EP
S m
ean
, [A
]D
ivid
end
yie
ld, F
Y2, [
D]
Mo
nth
ly C
reat
ed J
ob
s /N
um
ber
of
Emp
loye
es, [
D]
Div
iden
d y
ield
, FY1
, [D
]Lo
ng-
term
deb
t/eq
uit
y, [
A]
Mo
nth
ly A
ctiv
e Jo
bs/
Tota
l Sal
es, [
D]
Mo
nth
ly C
reat
ed J
ob
s/N
et In
com
e, [
A]
Est
Bo
ok-
to-p
rice
, med
ian
, [A
]A
nn
ual
Em
plo
yee
Gro
wth
, [D
]A
nn
ual
ly A
ctiv
e Jo
bs/
Tota
l Ass
ets,
[A
]M
on
thly
Act
ive
Job
s/To
tal A
sset
, [A
]IB
ES F
Y1 M
ean
RO
E R
evis
ion
, 1
M, [
D]
Mo
vin
g av
erag
e cr
oss
ove
r, 1
5W
-36
W,
[D]
Mer
ton
's d
ista
nce
to
def
ault
, [D
]C
reat
ed J
ob
Gro
wth
An
nu
ally
-N
et In
com
e G
row
th A
nn
ual
ly ,
[D]
Tota
l ret
urn
, 21
D (
1M
), [
D]
Act
ive
Job
Gro
wth
qu
arte
rly
-To
tal q
uar
terl
y A
sset
Gro
wth
, [D
]IB
ES F
Y1 M
ean
SA
L R
evis
ion
, 3M
, [A
]R
ealiz
ed v
ol,
1Y
dai
ly, [
A]
IBES
5Y
EPS
gro
wth
/sta
bili
ty, [
A]
An
nu
ally
Cre
ated
Jo
bs/
Nu
mb
er o
f Em
plo
yees
, [D
]A
ctiv
e Jo
b G
row
th A
nn
ual
ly -
Tota
l An
nu
ally
Ass
et G
row
th, [
D]
YoY
chan
ge in
deb
t o
uts
tan
din
g, [
D]
IBES
FY2
mea
n D
PS
gro
wth
, [D
]C
reat
ed J
ob
Gro
wth
qu
arte
rly
-M
arke
t C
ap q
uar
terl
y G
row
th, [
D]
Act
ive
Job
Gro
wth
An
nu
ally
-To
tal S
ales
qu
arte
rly
Gro
wth
, [A
]A
nn
ual
ly A
ctiv
e Jo
b N
um
ber
Gro
wth
, [D
]q
uar
terl
y C
reat
ed J
ob
Nu
mb
er G
row
th, [
A]
qu
arte
rly
Act
ive
Job
Nu
mb
er G
row
th, [
A]
IBES
LTG
Mea
n E
PS
Rev
isio
n,
3M
, [A
]A
ctiv
e Jo
b G
row
th A
nn
ual
ly -
Mar
ket
Cap
qu
arte
rly
Gro
wth
, [D
]M
ean
rec
om
men
dat
ion
rev
isio
n, 3
M, [
D]
Mo
nth
ly A
ctiv
e Jo
bs
/Nu
mb
er o
f Em
plo
yees
, [D
]M
on
thly
Cre
ated
Jo
bs/
Tota
l Ass
et, [
A]
An
nu
ally
Cre
ated
Jo
bs/
Tota
l Ass
ets,
[A
]A
nn
ual
ly A
ctiv
e Jo
bs/
Tota
l Sal
es, [
D]
Cre
ated
Jo
b G
row
th q
uar
terl
y -
Tota
l qu
arte
rly
Ass
et G
row
th, [
D]
Act
ive
Job
Gro
wth
qu
arte
rly
-N
et In
com
e q
uar
terl
y G
row
th, [
D]
Act
ive
Job
Gro
wth
An
nu
ally
-N
et In
com
e q
uar
terl
y G
row
th, [
D]
IBES
5Y
EPS
gro
wth
, [A
]M
on
thly
Cre
ated
Jo
bs/
Tota
l Sal
es, [
D]
Cre
ated
Jo
b G
row
th A
nn
ual
ly -
Tota
l Sal
es A
nn
ual
ly G
row
th, [
A]
IBES
FY1
EP
S u
p/d
ow
n r
atio
, 3M
, [A
]A
nn
ual
ly A
ctiv
e Jo
bs/
Net
Inco
me,
[A
]C
reat
ed J
ob
Gro
wth
qu
arte
rly
-To
tal S
ales
qu
arte
rly
Gro
wth
, [D
]B
erry
Rat
io, [
D]
IBES
FY1
Mea
n S
AL
Rev
isio
n, 1
M, [
D]
Emp
loye
e N
um
ber
/To
tal S
ales
, [A
]IB
ES F
Y1 M
ean
RO
E R
evis
ion
, 3
M, [
D]
Mo
nth
ly A
ctiv
e Jo
bs/
Net
Inco
me,
[A
]C
reat
ed J
ob
Gro
wth
An
nu
ally
-To
tal S
ales
qu
arte
rly
Gro
wth
, [A
]A
ctiv
e Jo
b G
row
th q
uar
terl
y -
Tota
l Sal
es q
uar
terl
y G
row
th, [
D]
Act
ive
Job
Gro
wth
An
nu
ally
-To
tal S
ales
An
nu
ally
Gro
wth
, [D
]C
reat
ed J
ob
Gro
wth
An
nu
ally
-N
et In
com
e q
uar
terl
y G
row
th, [
D]
Cre
ated
Jo
b G
row
th q
uar
terl
y -
Net
Inco
me
qu
arte
rly
Gro
wth
, [D
]IB
ES F
Y1 M
ean
EP
S R
evis
ion
, 1
M, [
D]
Act
ive
Job
Gro
wth
An
nu
ally
-To
tal A
sset
qu
arte
rly
Gro
wth
, [A
]C
reat
ed J
ob
Gro
wth
An
nu
ally
-To
tal A
sset
qu
arte
rly
Gro
wth
, [A
]C
reat
ed J
ob
Gro
wth
An
nu
ally
-M
arke
t C
ap q
uar
terl
y G
row
th, [
D]
An
nu
ally
Cre
ated
Jo
bs/
Tota
l Sal
es, [
D]
Skew
nes
s, 1
Y d
aily
, [D
]IB
ES F
Y1 E
PS
up
/do
wn
rat
io, 1
M, [
D]
An
nu
ally
Act
ive
Job
s/N
um
ber
of
Emp
loye
es, [
D]
Gro
ss m
argi
n, [
A]
IBES
FY1
mea
n E
PS
gro
wth
, [D
]D
ivid
end
yie
ld, t
raili
ng
12
M, [
D]
Cu
rren
t ra
tio
, [A
]IB
ES F
Y1 M
ean
EP
S R
evis
ion
, 3
M, [
D]
An
nu
ally
Cre
ated
Jo
bs/
Net
Inco
me,
[D
]C
reat
ed J
ob
Gro
wth
An
nu
ally
-To
tal A
nn
ual
ly A
sset
Gro
wth
, [D
]
Ave
rage
Mo
nth
ly L
on
g/Sh
ort
Qu
inti
le P
ort
folio
Re
turn
(%
)
Many job factors stack up fairly well
compared to traditional factors
Job Factors
Traditional Quant Factors
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Thus far we have backtested our portfolios using a one-month rebalance frequency.
However, we would hypothesize that companies may reap the benefits of added
innovation, productivity, revenue etc., that coincide with employee growth, several
months after jobs are posted and new employees commence. As such, we re-test our
job growth factors using longer rebalance frequencies. In particular, we test using a
one-month, six-month, and 12-month rebalance frequency.
Figure 37 compares the performance of all the job factors using a one-month, six-
month, and 12-month rebalance frequency. The results are interesting. The one-month
rebalance frequency still beats most other rebalance frequencies. However, we still see
strong performance with less frequent portfolio rebalances. In fact, most job growth
factors show even better returns using longer rebalance frequencies (i.e. six or 12
month).
The job-based factors stack
up fairly well compared to
traditional quant factors
backtested over the same
period
We would hypothesize that
companies may reap the
benefits off added innovation,
productivity, revenue etc…
that coincide with employee
growth, several months after
jobs are posted and new
employees commence
19 May 2015
Signal Processing
Page 20 Deutsche Bank Securities Inc.
Figure 37 Long/short monthly quintile return spread of job factors using various rebalance frequencies
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
Active Job Factors Active Job Growth Factor Job Creation Factor Job Creation Growth Factor
Ave
rage
Lo
ng/
Sho
rt A
nn
ual
ize
d R
etu
rn (
%)
1month rebalance 6month rebalance 12month rebalance
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Figure 38 shows the correlation between the active jobs per month factor and typical
quantitative factor portfolios.6 Interestingly, the job factor is negatively correlated to
typically quantitative strategies. This is a promising finding as it suggests that the
inclusion of the job factor into a multi-strategy portfolio is likely to improve
diversification. It is also important to note that the job factor has a small cap tilt.
Figure 38: Overall correlation between job and traditional quant factors
-35.00%
-30.00%
-25.00%
-20.00%
-15.00%
-10.00%
-5.00%
0.00%
Size Quality Value Growth Momentum Sentiment
Co
rre
lati
on
wit
h t
rad
itio
anl f
acto
rs
Job factors are tilted towards smaller cap
stocks
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
6 Note that the job factor for this analysis is active jobs within the month scaled by market cap. The market cap scaling
is done by using an OLS regression.
Interestingly, the job factor is
negatively correlated to
typically quantitative
strategies. This is a promising
finding as it suggests that the
inclusion of the job factor into
a multi-strategy portfolio will
improve diversification
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 21
The next logical question to ask is whether these job factors perform well in a larger cap
universe. This should provide some insight into whether the small cap size tilt is the main
driver and source of alpha. Recall that LinkUp currently covers more than 50% of the
stocks within the Russell 3000 universe and approximately 90% of the stocks in the S&P
500 (see Figure 39).
Figure 39: Job factor coverage in different universes
0
500
1,000
1,500
2,000
2,500
Nu
mb
er
of
Sto
cks
S&P 500 Russell 3000
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Next, we backtest the performance of the jobs factors within the S&P 500 universe and
analyze the portfolio quintile return performance (Figure 40). Interestingly, we find that
the job-related factors perform better in a large-cap universe.
Figure 40 Long/short monthly quintile return spread of job factors using different investment universes
0.00%
0.20%
0.40%
0.60%
0.80%
1.00%
1.20%
1.40%
Ave
rage
Lo
ng/
Sho
rt Q
uin
tile
Mo
nth
ly P
ort
folio
Re
turn
(%
)
Russell 3000 Universe S&P 500 Universe
Job factors tend to perform better in a larger cap universe
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Simply comparing the long/short quintile portfolio return performance may be
somewhat inappropriate since the portfolios have different number of stocks. A better
measure may be to analyze the Sharpe ratio of each factor (Figure 41). Interestingly, we
find similar results. Job-related factors perform slightly better in a large-cap universe on
a risk adjusted basis. This is a promising finding, since most traditional quant factors
struggle to add alpha within a large cap universe.
Is the small cap size tilt (or
other anti-quant tilts) the main
drivers and sources of alpha?
Or is there incremental alpha
in job-related postings?
Interestingly, we find that the
job related factors perform
slightly better in a large-cap
universe
Job-related factors perform
slightly better in a large-cap
universe on a risk adjusted
basis
19 May 2015
Signal Processing
Page 22 Deutsche Bank Securities Inc.
Figure 41 Information ratio of job factors using different investment universes
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
Shar
pe
Rat
io
Russell 3000 S&P 500Job factors tend to perform
better on a risk adjusted basis in a larger cap universe
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Figure 42 and Figure 43 summarize our findings and reiterate that job related factors
tend to perform better within a large cap universe.
Figure 42: Compare average long/short spread of all jobs
related factors in different universes
Figure 43: Compare average Sharpe of all jobs related
factors in different universes
Russell 3000
S&P 500
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
Ave
rage
Ab
solu
te L
on
g/Sh
ort
Re
turn
(%
)
Russell 3000
S&P 500
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
Shar
pe
Rat
io
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 23
Job macro alpha
In the previous section, we showed that the job posting dataset has some underlying
stock selection ability. Motivated by Choi and Varian [2011], who use Google Trends
data to predict the unemployment rate, in this section, we explore whether the job
posting dataset can be ”employed“ to aid in the forecasting of macroeconomic
indicators respectively, including employment indicators.
Macroeconomic indicators
We start by compiling a small set of macroeconomic indicators that we intuitively think
can be forecasted using the jobs posting dataset. The unemployment rate, jobless
claims, and non-farm payrolls are the obvious choices. However, undoubtedly, job
growth has a multiplier effect and can hence benefit other industries.
As such, we attempt to forecast the Cash-Shiller national home price index, the
purchasing managers index (PMI-an indicator of the overall economic cycle), retail
sales, and consumer sentiment. Figure 44 shows all the macroeconomic indicators in
our study.
Figure 44: Various macro-economic indicators tested
Macro-Economic Indicators Description
Case-Shiller U.S. National Home Price Index A leading measures of U.S. residential real estate prices, tracking changes in the value of residential real estate both nationally
Total Nonfarm Private Payroll Employment A statistic researched, recorded and reported by the U.S. Bureau of Labor Statistics intended to represent the total number of paid U.S. workers of any non-farm business
Civilian Unemployment Rate The unemployment rate represents the number of unemployed as a percentage of the labor force. Labor force data are restricted to people 16 years of age and older, who currently reside in 1 of the 50 states or the District of Columbia, who do not reside in institutions (e.g., penal and mental facilities, homes for the aged), and who are not on active duty in the Armed Forces.
PMI Composite Index An indicator of the economic health of the manufacturing sector. The PMI index is based on five major indicators: new orders, inventory levels, production, supplier deliveries and the employment environment. A PMI reading above 50 percent indicates that the manufacturing economy is generally expanding; below 50 percent that it is generally declining.
Initial Claims A measure of the number of jobless claims filed by individuals seeking to receive state jobless benefits. This number is watched closely by financial analysts because it provides insight into the direction of the economy.
Consumer Price Index A measure of the average monthly change in the price for goods and services paid by urban consumers between any two time periods. It can also represent the buying habits of urban consumers. This particular index includes roughly 88 percent of the total population, accounting for wage earners, clerical workers, technical workers, self-employed, short-term workers, unemployed, retirees, and those not in the labor force.
Real Retail and Food Services Sales An aggregated measure of the sales of retail goods over a stated time period, typically based on a data sampling that is extrapolated to model an entire country
University of Michigan: Consumer Sentiment A consumer confidence index published monthly by the University of Michigan and Thomson Reuters. The index is normalized to have a value of 100
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, FRED, Deutsche Bank Quantitative Strategy
Are job postings correlated to macro indices?
To test the predictive ability of utilizing job postings to forecast macroeconomic
indicators, we first calculate the number of active jobs for a basket of companies. Next,
for that same basket of companies, we calculate the number of active jobs in the
19 May 2015
Signal Processing
Page 24 Deutsche Bank Securities Inc.
following month. Lastly, we compute the percentage change in the number of active
jobs. Figure 45 shows percentage change in active job postings.7
Figure 45: Month-over-month percentage change in same company active
jobs
-15
-10
-5
0
5
10
15
Tota
l Mo
nth
to
Mo
nth
Ch
ange
in A
ctiv
e J
ob
s (%
)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
We then analyze the relationship between the percentage change in active jobs and the
percentage change in a select group of macro-economic indicators to determine if the
two metrics are correlated. Please note that most economic data series have reporting
lags. For example, NFP data for March 2015 was released during the first week of April
2015. For all predictive analysis, we use the month end job data to predict the
subsequent month’s economic data which may be released in later months. For
correlation analysis, we analyze month end jobs data with economic data that is
released after the month end.8
Figure 46 shows the correlation between changes in active jobs and changes in non-
farm payrolls from one to 12 months ahead9. The results are promising. We can clearly
see that changes in active jobs are highly positively correlated to changes in non-farm
payrolls. And, changes in active jobs can potentially forecast changes in non-farm
payrolls several months in to the future. The changes in active jobs factor has a positive
serial correlation.10 Additionally, job postings take time to fill and as such, the impact to
payrolls may be seen a few months after jobs are posted.
Figure 47 shows the correlation between changes in active jobs and unemployment
rate. Interestingly, the correlation is predominantly negative. This is expected, since an
increase in active jobs should intuitively lead to a lower unemployment rate.
7 Note this is in effect the month-over-month, percentage change in same company active jobs.
8 For example, for correlation analysis, we compute the correlation between end of March’s jobs data from Linkup and
March’s macroeconomic data which is released in early April Note that Linkup’s job data is released at the end of
March. 9 Note that all the analysis with macroeconomic variables is based on the initial first reported data instead of the
revised numbers. In Luo, et al [2010], we discuss in detail about the restatement bias in macroeconomic data in asset
return prediction. The NFP number is seasonally adjusted. 10
Note that the correlation is formed using a limited number of data points, specifically from end of January 2009
onwards using a monthly frequency.
We can clearly see that
changes in active jobs are
highly positively correlated to
changes in non-farm payrolls
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 25
Figure 46: Correlation of non-farm payrolls and average
active jobs per company
Figure 47: Correlation of unemployment rate and average
active jobs per company
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
1 2 3 4 5 6 7 8 9 10 11 12Co
rre
lati
on
be
twe
en
NFP
an
d C
han
ge in
A
ctiv
e J
ob
s
Month Ahead
-30.0%
-25.0%
-20.0%
-15.0%
-10.0%
-5.0%
0.0%
5.0%
10.0%
15.0%
1 2 3 4 5 6 7 8 9 10 11 12
Co
rre
lati
on
be
twe
en
Un
em
plo
yme
nt
Rat
e a
nd
Ch
ange
in A
ctiv
e J
ob
s
Month Ahead
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, FRED, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, FRED, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Figure 48 shows the one-month ahead correlation between changes in active jobs and
various economic indicators. We find that changes in non-farm payrolls are highly
correlated to changes in active jobs whereas the unemployment rate has the most
negative correlation.
Figure 48: Correlation between changes in active jobs and one-month ahead
macro indices
-20.0%
-15.0%
-10.0%
-5.0%
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
Co
rre
lati
on
be
twe
en
Ch
ange
s in
Act
ive
Jo
bs
and
On
e M
on
th
Ah
ead
Mac
roe
con
om
ic In
dic
ato
rs (
%)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, FRED, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
In summary, we found that the job posting dataset is most strongly correlated to
changes in NFP. Next, we explore whether we can utilize the job posting dataset to
develop a better and more accurate forecast of NFP.
We find that changes in non-
farm payrolls are highly
correlated to changes in
active jobs whereas the
unemployment rate has the
most negative correlation
19 May 2015
Signal Processing
Page 26 Deutsche Bank Securities Inc.
Forecasting non-farm payrolls
Devising systematic forecasts for non-farm payrolls
In this section, we investigate whether we can devise a better or more accurate
prediction of NFP using jobs data. Figure 49 shows the total NFP values from 1987
onwards. NFP has been steadily increasing since the late 1980s, albeit there are some
bouts where NFP drops sharply. Investors are generally focused on the monthly
changes in NFP (Figure 50)
Figure 49: Time series of total NFP (monthly) Figure 50: Month-over-month changes in NFP
100
105
110
115
120
125
130
135
140
145
Tota
l No
nFa
rm P
ayro
ll (i
n m
illio
ns)
-800
-600
-400
-200
0
200
400
600
800
Mo
nth
to
Mo
nth
Ch
ange
in N
on
Farm
P
ayro
ll (i
n t
ho
usa
nd
s)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
A deeper analysis of the changes in NFP (Figure 51) shows the average change in NFP
is approximately 100,000 jobs. The month-over-month range in changes in NFP is
approximately between -650,000 and 700,000. We find that changes in NFP are
negatively skewed. This means that changes in NFP are predominantly positive;
however, there are a few instances where month-over-month changes in NFP are
significantly negative.
NFP has been steadily
increasing since the late
1980s albeit there are some
bouts where NFP drops
sharply
We find that changes in NFP
are negatively skewed. This
means that changes in NFP
are predominantly positive;
however, there are a few
instances where month over
month changes in NFP are
significantly negative
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 27
Figure 51: Histogram of month over month changes in NFP (in thousands)
0
10
20
30
40
50
60
70
80
90
100
-800 -700 -600 -500 -400 -300 -200 -100 0 100 200 300 400 500 600 700 800
Fre
qu
en
cy
Month Over Month Changes in NFP (in thousands)
Average = 103Standard Deviation = 189Minimum = (-663)Maximum = 705Skew = (-0.95)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
We can further analyze changes in NFP by performing a decomposition. A signal
decomposition simply isolates certain characteristics of a dataset to better understand
its nature and structure. A signal can be decomposed into a trend, seasonal, and
residual component. Figure 52 shows the signal decomposition of month over month
changes in NFP.
The analysis shows the changes in NFP has a significant trend component. Changes in
NFP also have a seasonal pattern. Interestingly, the “residual” or remainder component
(i.e. the part of the signal that is not explained by trend or seasonality) is fairly
significant. This may imply that changes in NFP can have a significant surprise factor.
This may imply that changes
in NFP can have a significant
surprise factor. This will likely
make it trickier to forecast
changes to NFP
19 May 2015
Signal Processing
Page 28 Deutsche Bank Securities Inc.
Figure 52: Decomposition of NFP into trend, seasonality, and remainder
Month over Month Change in NFP
Seasonal component of changes in NFP
Trend component of changes in NFP
Remaining component in changes in NFP
Source: : Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Forecasting NFP using NFP
There are many methods investors can employ to forecast time-series data. One
common method is to utilize an ARIMA model.11 The inherent assumption underlying
ARIMA model forecasting is that there is autocorrelation in the signal. Broadly speaking,
an ARIMA model utilizes a combination of past signal values to forecast the future
signal value. Deciding which combination of previous signal values to employ in the
forecast requires some deeper analysis. A partial correlation chart aids with this
analysis.
Figure 53 shows the partial correlation plots for changes in NFP. The partial correlation
chart shows the autocorrelation in the signal after remove the effects of other time lags.
The chart shows that month over month changes in NFP are highly correlated or best
forecasted using the previous 3 month values for NFP changes.
11 ARIMA stands for Autoregressive Integrated Moving Average models
Broadly speaking, an ARIMA
model utilizes a combination
of past signal values to
forecast the future signal
value
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 29
Figure 53: Autocorrelation and partial autocorrelation analysis
NFP is best predicted using the previous 3 month
values
Source: : Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
After further analysis and testing various models, we determined that the ARIMA(3,1,1)
model is most accurate utilizing a 60 month training window.12 Figure 54 compares our
ARIMA forecast versus the actual NFP changes. On the surface, the ARIMA model
seems to be performing moderately well.
12 The ARIMA (3,1,1) model had the lowest Akaike information criterion (AIC).
On the surface, the ARIMA
model seems to be
performing moderately well
19 May 2015
Signal Processing
Page 30 Deutsche Bank Securities Inc.
Figure 54: Actual NFP change versus ARIMA forecast
-800
-600
-400
-200
0
200
400
600
800
Actual NFP ARIMA Model Predicted NFP
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Figure 55 shows a scatter plot of our ARIMA forecasted NFP versus the actual NFP
numbers. A perfectly accurate model would fit a line with a slope and R2 of one. The
ARIMA model performs reasonably well. Figure 56 shows a scatter plot of the
Bloomberg consensus NFP forecasts versus the actually NFP numbers. The consensus
Bloomberg forecast appears to be more accurate than our ARIMA model based on the
slope and R2 values.13 However, comparing the accuracy of our ARIMA model against
Bloomberg consensus forecasts is not entirely fair.
This is because the ARIMA model determines the NFP forecast one month prior to
when the actual NFP number is published. However, the Bloomberg consensus NFP
forecast is a more current estimate. Bloomberg consensus NFP forecasts can be
updated days prior to when the actual NFP numbers are released. This of course gives a
significant advantage to the consensus Bloomberg estimates over our ARIMA model.
13 Note that for both of the scatter plots, we have forced the intercept term to zero.
The consensus Bloomberg
forecast appears to be more
accurate than our ARIMA
model. However, comparing
the accuracy of our ARIMA
model against Bloomberg
consensus forecasts is not
entirely fair
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 31
Figure 55: Actual NFP versus ARIMA forecast NFP
Figure 56: Actual NFP versus Bloomberg consensus
forecast NFP
y = 0.926xR² = 0.5589
-800
-600
-400
-200
0
200
400
600
-800 -600 -400 -200 0 200 400
Act
ual
NFP
Ch
ange
ARIMA Forecast NFP change
y = 0.9583xR² = 0.7948
-800
-600
-400
-200
0
200
400
600
-800 -600 -400 -200 0 200 400 600
Act
ual
NFP
Ch
ange
Bloomberg Consensus Forecast NFP change
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Another method to gauge the accuracy of our AIRMA model versus consensus
Bloomberg estimates is to compute the mean squared estimation error. We find that
the Bloomberg consensus estimates have smaller estimation error than our ARIMA
model. Again, this is expected since Bloomberg consensus estimates are more timely
than our ARIMA model.14
Figure 57: Average estimate error for Bloomberg consensus and ARIMA
Bloomberg Consensus
ARIMA
-
20
40
60
80
100
120
140
Ave
rage
Est
imat
ion
Err
or
(in
NFP
th
ou
san
ds)
Source: : Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
14 Note that Bloomberg estimates could also be more accurate because the underlying analysts’ estimates are more
accurate than our systematic ARIMA model.
19 May 2015
Signal Processing
Page 32 Deutsche Bank Securities Inc.
Unfortunately, we do not have a point-in-time daily running estimate of Bloomberg
consensus NFP forecasts so that we can make a more fair comparison. However, we
do have LinkUp jobs database which does provide current data on active jobs. Perhaps
we can utilize changes in active jobs to provide more timeliness to our NFP estimates.
We explore this next.
The optimal NFP forecast – Bridging job postings
We compare the performance of various models to see if adding the job postings data
can add incremental value and improve the accuracy of the forecasts. We compare the
results of various models:
Bloomberg Consensus: Our baseline benchmark is the Bloomberg consensus
forecast of NFP. All subsequent models will be assessed against our baseline
benchmark model.
ARIMA: This is the time series model discussed in the previous section
ARIMA + Change Active Job Postings: This model combines the ARIMA model
with the change in active job postings, approximately a month prior to the
economic release date.15 We utilize the month prior job change so that posted
jobs have an opportunity to be filled and impact payroll. We use the forecast
from the ARIMA model and the change in active jobs, a month prior, to predict
NFP using a bivariate OLS regression. The inclusion of the active jobs
dependant variable in the regression adds more timeliness to our ARIMA
forecast.16
ARIMA + Change in Active Job Postings + Bloomberg Consensus: Here we
also add the Bloomberg consensus NFP forecasts. The combination of all of our
models may show some additional forecasting ability.17
Figure 58 shows the timeline of each data series used in our various predictive models.
15 To predict March’s NFP which is released in early April, we utilize end of February’s job numbers from Linkup.
16 The job postings data utilized in the model begins in January 2009. Note that this model uses a 48 month training
window. Therefore our out of sample results are over a two year period. The change in active job postings is a total
absolute change not a percentage change. 17
Note for this model, the Bloomberg forecast employed is the day prior to the actually NFP release.
Unfortunately, we do not
have a point-in-time daily
running estimate of
Bloomberg consensus NFP
forecasts so that we can
make a more fair comparison.
However, we do have LinkUp
jobs database which does
provide current data on active
jobs
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 33
Figure 58: Timeline example of data sets used in our forecasting models
January 30th
February 27th
March 31st
April 30th
May 8, 2015
Initial NFP Released for April
May 7, 2015
Bloomberg Final Consensus on April's NFP
May 29th
March 31, 2015
Job Posting for Forecasting April's NFP
April 3, 2015
ARIMA data for forecasting April's NFP
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Figure 59 shows the estimation error of each model. This is used to assess the accuracy
of the model. Adding the job posting dataset to our ARIMA model reduces the
estimation error thereby improving the accuracy of the model and more importantly, it
outperforms the Bloomberg estimates based on the estimation error. In fact, all models
utilizing the job posting dataset improve the accuracy of the NFP forecast and
outperform the Bloomberg consensus forecast. This outperformance is achieved by
using data as of the previous month end whereas Bloomberg estimates are updated a
few days prior to the release of NFP.
Figure 59: Average estimation error of all our backtested models
ARIMA + Job Postings
ARIMA + Job Postings + Bloomberg Consensus
Bloomberg Consensus
ARIMA
40
45
50
55
60
65
70
Ave
rage
Est
imat
ion
Err
or
(in
NFP
th
ou
snd
s)
Adding job posting changes to forecast NFP
improves accuracy
Source: : Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Lastly, to better gauge the accuracy of the models, we form various scatter plots of our
forecasted models versus actual NFP numbers (Figure 60 to Figure 63). A perfectly
accurate model would fit a line with a slope and R2 of one. Again, we see that inclusion
of the job posting dataset beats the Bloomberg consensus estimates. We note that
there are few data points for this analysis. However, as more jobs posting related data
becomes available, we can extend the training dataset and show performance results
over a longer time period in the future.
This outperformance is
achieved by using data as of
the previous month end
whereas Bloomberg
estimates are updated a few
days prior to the release of
NFP
19 May 2015
Signal Processing
Page 34 Deutsche Bank Securities Inc.
Figure 60: Actual NFP & ARIMA forecast NFP Figure 61: Actual NFP & ARIMA + Jobs
y = 0.4891x + 108.95R² = 0.0803
0
50
100
150
200
250
300
350
0 50 100 150 200 250 300
Act
ual
NFP
Ch
ange
(n
i th
ou
sdan
ds)
ARIMA Model Forecast (in thousands)
y = 0.9095x + 29.428R² = 0.3013
0
50
100
150
200
250
300
350
0 50 100 150 200 250 300
Act
ual
NFP
Ch
ange
(in
th
ou
san
ds)
ARIMA Model + Active Job Change Forecast (in thousands)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Figure 62: Actual NFP & ARIMA + Jobs + Bloomberg Figure 63: Actual NFP & Bloomberg
y = 0.9123x + 29.992R² = 0.2939
0
50
100
150
200
250
300
350
0 50 100 150 200 250 300
Act
ual
NFP
Ch
ange
(in
th
ou
san
ds)
ARIMA Model + Active Job Change + Bloomberg Consensus Forecast (in thousands)
y = 0.6245x + 73.257R² = 0.1026
0
50
100
150
200
250
300
350
0 50 100 150 200 250 300
Act
ual
NFP
Ch
ange
(in
th
ou
san
ds)
Bloomerg Consensus Forecast (in thousands)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Haver, Deutsche Bank Quantitative Strategy
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 35
Sector rotation using jobs
Forecasting industry groups
In this brief section, we analyze whether the job posting dataset has the ability to
predict industry returns. Figure 65 shows the coverage of companies within the job
posting dataset by industry group. The coverage within each of the 24 industry groups
is reasonable and as such we can attempt to devise a sector rotation strategy based on
industry groups.
Figure 64: Time series coverage of GICS industry group within jobs dataset
0
200
400
600
800
1,000
1,200
1,400
1,600
1,800
2,000
Ind
ust
ry G
rou
p M
on
thly
Co
vera
ge
Software & Services
Capital Goods
Health Care Equipment & Services
Technology Hardware & Equipment
Banks
Pharmaceuticals, Biotechnology & Life Sciences
Retailing
Energy
Semiconductors & Semiconductor Equipment
Materials
Commercial & Professional Services
Consumer Durables & Apparel
Consumer Services
Insurance
Utilities
Real Estate
Diversified Financials
Media
Food, Beverage & Tobacco
Transportation
Telecommunication Services
Food & Staples Retailing
Household & Personal Products
Automobiles & Components
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Our strategy involves aggregating the company level job factors discussed in the
previous sections to an industry group level. We aggregate the company level job
factors by simply computing the average factor value within each industry group. Next,
we form an industry portfolio by longing the top and shorting the bottom industries
based on the average factor value. We track the performance of the portfolio
overtime.18
18 Note that the performance of each industry is based on the equally weighted average return of companies within
that industry.
The coverage within each of
the 24 industry groups is fairly
good
We aggregate the company
level job factors by simply
computing the average factor
value within each industry
group
19 May 2015
Signal Processing
Page 36 Deutsche Bank Securities Inc.
Figure 65 shows the performance of our industry rotation strategy for each job factor.
Note that the results show the factor rank information coefficient (rank IC). An industry
rotation strategy based on the job factor performs reasonably well. Some of the top
performing factors have a rank IC of between 7% and 8%.
Figure 65: Average monthly performance (absolute Rank IC) of industry rotation strategies using jobs factors
0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 9.0%
Monthly Created Jobs to Market Cap*
Created Job Growth Annually - Total Sales Annually Growth
Created Job Growth Annually - Net Income Growth Annually
Annually Created Jobs/Total Sales
Active Job Growth Annually - Net Income Growth Annually
Created Job Growth Annually - Total Annually Asset Growth
Created Job Growth Annually - Net Income Quaterly Growth
Monthly Created Jobs/Total Sales
Annually Created Jobs to Market Cap*
Annually Created Job Number Growth
Created Job Growth Annually - Total Asset Quaterly Growth
Monthly Created Jobs/Market Cap
Created Job Growth Annually - Market Cap Quaterly Growth
Monthly Active Jobs/Total Sales
Created Job Growth Annually - Total Sales Quaterly Growth
Active Job Growth Quaterly - Total Sales Quaterly Growth
Active Job Growth Quaterly - Total Quaterly Asset Growth
Created Job Growth Annually - Market Cap Annually Growth
Annually Active Jobs to Market Cap*
Active Job Growth Annually - Market Cap Quaterly Growth
Annually Created Jobs/Market Cap
Monthly Active Jobs /Number of Employees
Quaterly Active Job Number Growth
Active Job Growth Quaterly - Net Income Quaterly Growth
Monthly Created Jobs /Number of Employees
Monthly Created Jobs/Net Income
Annually Active Jobs/Total Sales
Annually Active Job Number Growth
Active Job Growth Annually - Total Sales Quaterly Growth
Active Job Growth Annually - Net Income Quaterly Growth
Active Job Growth Annually - Total Asset Quaterly Growth
Monthly Active Jobs/Market Cap
Annually Active Jobs/Number of Employees
Active Job Growth Annually - Market Cap Annually Growth
Monthly Active Jobs to Market Cap*
Active Job Growth Annually - Total Annually Asset Growth
Monthly Active Jobs/Net Income
Active Job Growth Annually - Total Sales Annually Growth
Annually Active Jobs/Market Cap
Annually Active Jobs/Net Income
Annually Created Jobs/Net Income
Active Job Growth Quaterly - Market Cap Quaterly Growth
Monthly Created Jobs/Total Asset
Annually Created Jobs/Number of Employees
Monthly Active Jobs/Total Asset
Annually Active Jobs/Total Assets
Annually Created Jobs/Total Assets
Annual Employee Growth
Created Job Growth Quaterly - Net Income Quaterly Growth
Created Job Growth Quaterly - Total Sales Quaterly Growth
Created Job Growth Quaterly - Total Quaterly Asset Growth
Quaterly Created Job Number Growth
Created Job Growth Quaterly - Market Cap Quaterly Growth
Average Rank IC (%)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Taking a closer look at a particular job factor (quarterly job growth minus market cap
growth) we find that the time-series rank IC is fairly strong (see Figure 66). In addition, the
long/short portfolio has a Sharpe ratio of approximately 1.0x (see Figure 67). More
interestingly, at the company level, firms with high job growth, exceeding market cap
growth, tend to produce lower subsequent returns (i.e. the so-called asset growth anomaly
– see Cooper, et al [2009]). However, at the industry level, industry groups with strong job
growth (above market cap growth) actually lead to higher returns in the following month.
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 37
Figure 66: Rank IC of quarterly jobs created growth
minus market cap growth
Figure 67: Sharpe of quarterly jobs created growth minus
market cap growth
Spearman Rank IC (%)
IC (
%)
2009 2010 2011 2012 2013 2014 2015
-40
-20
0
20
40
60
80
Spearman rank IC (%), Ascending order12-month moving average
Avg = 7.63%Std. Dev. = 22.95%Min = -35.39%
Max = 66.17%Avg/Std. Dev. = 0.33
1 2 3 4 L/S
Fractile Portfolio IRs
Fra
ctile
Po
rtfo
lio
IR
s
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Labor intensive industries
Lastly, we test whether our industry rotation strategy performs better in labor force
intensive sectors like retailing. To do this, we re-run our industry rotation strategy only
within labor intensive industries. We define labor intensive industries as the top 12
industries ranked by the highest median number of employees per company in that
industry group. Figure 68 highlights (in green) the top 12 industry groups with the highest
number of employees.
Figure 68: Median number of employees per industry group
0
5
10
15
20
25
Me
dia
n E
mp
loye
e N
um
be
r p
er
Ind
ust
ry G
rou
p (
in t
ho
usa
nd
s)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
Lastly, we test whether our
industry rotation strategy
performs well in labor force
intensive sectors like retailing
19 May 2015
Signal Processing
Page 38 Deutsche Bank Securities Inc.
We backtest the industry rotation strategy discussed previously but just within these 12
industries. Again the results are reasonable (see Figure 69). Some of the top performing
factors have a rank IC of between 7% and 9%.
Figure 69: Average monthly performance (absolute Rank IC) of sector rotation strategies in labor intensive industries
0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 9.0% 10.0%
Annually Active Jobs/Total Sales
Annually Created Jobs/Net Income
Annually Active Jobs/Number of Employees
Monthly Created Jobs/Total Asset
Monthly Created Jobs /Number of Employees
Monthly Active Jobs/Net Income
Annually Active Jobs/Net Income
Monthly Active Jobs/Total Sales
Monthly Created Jobs/Total Sales
Active Job Growth Quaterly - Market Cap Quaterly Growth
Annually Created Jobs/Number of Employees
Monthly Active Jobs/Total Asset
Annually Active Jobs/Total Assets
Active Job Growth Annually - Market Cap Quaterly Growth
Monthly Active Jobs /Number of Employees
Monthly Created Jobs/Net Income
Annually Created Job Number Growth
Created Job Growth Annually - Total Sales Quaterly Growth
Annually Created Jobs/Total Assets
Created Job Growth Quaterly - Net Income Quaterly Growth
Created Job Growth Annually - Total Asset Quaterly Growth
Created Job Growth Annually - Market Cap Quaterly Growth
Created Job Growth Annually - Market Cap Annually Growth
Annually Created Jobs to Market Cap*
Monthly Created Jobs/Market Cap
Annually Created Jobs/Market Cap
Annually Active Jobs to Market Cap*
Active Job Growth Annually - Total Sales Quaterly Growth
Monthly Created Jobs to Market Cap*
Annually Active Job Number Growth
Created Job Growth Annually - Total Sales Annually Growth
Active Job Growth Annually - Total Asset Quaterly Growth
Created Job Growth Annually - Total Annually Asset Growth
Monthly Active Jobs to Market Cap*
Active Job Growth Quaterly - Total Quaterly Asset Growth
Active Job Growth Quaterly - Total Sales Quaterly Growth
Quaterly Active Job Number Growth
Created Job Growth Annually - Net Income Quaterly Growth
Active Job Growth Annually - Total Annually Asset Growth
Active Job Growth Annually - Market Cap Annually Growth
Created Job Growth Quaterly - Total Sales Quaterly Growth
Annually Created Jobs/Total Sales
Annually Active Jobs/Market Cap
Active Job Growth Annually - Total Sales Annually Growth
Created Job Growth Quaterly - Total Quaterly Asset Growth
Monthly Active Jobs/Market Cap
Active Job Growth Quaterly - Net Income Quaterly Growth
Quaterly Created Job Number Growth
Created Job Growth Quaterly - Market Cap Quaterly Growth
Active Job Growth Annually - Net Income Quaterly Growth
Created Job Growth Annually - Net Income Growth Annually
Active Job Growth Annually - Net Income Growth Annually
Annual Employee Growth
Average Rank IC (%)
Source: Bloomberg Finance LP, Compustat, IBES, Russell, S&P, Thomson Reuters, Worldscope, LinkUp, Deutsche Bank Quantitative Strategy
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 39
References
Dayton, T. [2012], “Leveraging The LinkUp Index To Predict Future U.S. Job Growth”,
LinkUp job search engine, June 2012
Dayton, T. [2015], “A Bullish, Hands-Free NFP Forecast For Friday’s Jobs Report”,
LinkUp job search engine, May 2015
Luo, Y., Cahan, R., Jussa, J. and Alvarez, M. [2010]. ”Signal Processing: Style rotation“,
Deutsche Bank Quantitative Strategy, 7 September, 2010
Cooper, M.J., Gulen, H., and Schill, M.J. [2009]. ”The Asset Growth Effect in Stock
Returns“, SSRN Working Paper, http://ssrn.com/abstract=1335524, 31 January, 2009
Choi, H., and Varian, H. [2011], "Predicting the Present with Google Trends",
http://people.ischool.berkeley.edu/~hal/Papers/2011/ptp.pdf, 18 December, 2011
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 40
Appendix 1
Important Disclosures
Additional information available upon request
*Prices are current as of the end of the previous trading session unless otherwise indicated and are sourced from local exchanges via Reuters, Bloomberg and other vendors . Other information is sourced from Deutsche Bank, subject companies, and other sources. For disclosures pertaining to recommendations or estimates made on securities other than the primary subject of this research, please see the most recently published company report or visit our global disclosure look-up page on our website at http://gm.db.com/ger/disclosure/DisclosureDirectory.eqsr
Analyst Certification
The views expressed in this report accurately reflect the personal views of the undersigned lead analyst(s). In addition, the undersigned lead analyst(s) has not and will not receive any compensation for providing a specific recommendation or view in this report. Javed Jussa/George Zhao/Yin Luo/Miguel-A Alvarez/Sheng Wang/Gaurav Rohal/Allen Wang/David Elledge/Kevin Webster
Hypothetical Disclaimer
Backtested, hypothetical or simulated performance results have inherent limitations. Unlike an actual performance record
based on trading actual client portfolios, simulated results are achieved by means of the retroactive application of a backtested
model itself designed with the benefit of hindsight. Taking into account historical events the backtesting of performance also
differs from actual account performance because an actual investment strategy may be adjusted any time, for any reason,
including a response to material, economic or market factors. The backtested performance includes hypothetical results that
do not reflect the reinvestment of dividends and other earnings or the deduction of advisory fees, brokerage or other
commissions, and any other expenses that a client would have paid or actually paid. No representation is made that any
trading strategy or account will or is likely to achieve profits or losses similar to those shown. Alternative modeling techniques
or assumptions might produce significantly different results and prove to be more appropriate. Past hypothetical backtest
results are neither an indicator nor guarantee of future returns. Actual results will vary, perhaps materially, from the analysis.
Regulatory Disclosures
1.Important Additional Conflict Disclosures
Aside from within this report, important conflict disclosures can also be found at https://gm.db.com/equities under the
"Disclosures Lookup" and "Legal" tabs. Investors are strongly encouraged to review this information before investing.
2.Short-Term Trade Ideas
Deutsche Bank equity research analysts sometimes have shorter-term trade ideas (known as SOLAR ideas) that are consistent
or inconsistent with Deutsche Bank's existing longer term ratings. These trade ideas can be found at the SOLAR link at
http://gm.db.com.
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 41
Additional Information
The information and opinions in this report were prepared by Deutsche Bank AG or one of its affiliates (collectively "Deutsche
Bank"). Though the information herein is believed to be reliable and has been obtained from public sources believed to be
reliable, Deutsche Bank makes no representation as to its accuracy or completeness.
Deutsche Bank may consider this report in deciding to trade as principal. It may also engage in transactions, for its own
account or with customers, in a manner inconsistent with the views taken in this research report. Others within Deutsche
Bank, including strategists, sales staff and other analysts, may take views that are inconsistent with those taken in this
research report. Others within Deutsche Bank, including strategists, sales staff and other analysts, may take views that are
inconsistent with those taken in this report. Deutsche Bank issues a variety of research products, including fundamental
analysis, equity-linked analysis, quantitative analysis and trade ideas. Recommendations contained in one type of
communication may differ from recommendations contained in others, whether as a result of differing time horizons,
methodologies or otherwise.
Analysts are paid in part based on the profitability of Deutsche Bank AG and its affiliates, which includes investment banking
revenues.
Opinions, estimates and projections constitute the current judgment of the author as of the date of this report. They do not
necessarily reflect the opinions of Deutsche Bank and are subject to change without notice. Deutsche Bank has no obligation
to update, modify or amend this report or to otherwise notify a recipient thereof if any opinion, forecast or estimate contained
herein changes or subsequently becomes inaccurate. This report is provided for informational purposes only. It is not an offer
or a solicitation of an offer to buy or sell any financial instruments or to participate in any particular trading strategy. Target
prices are inherently imprecise and a product of the analyst’s judgment. The financial instruments discussed in this report may
not be suitable for all investors and investors must make their own informed investment decisions. Prices and availability of
financial instruments are subject to change without notice and investment transactions can lead to losses as a result of price
fluctuations and other factors. If a financial instrument is denominated in a currency other than an investor's currency, a
change in exchange rates may adversely affect the investment. Past performance is not necessarily indicative of future results.
Unless otherwise indicated, prices are current as of the end of the previous trading session, and are sourced from local
exchanges via Reuters, Bloomberg and other vendors. Data is sourced from Deutsche Bank, subject companies, and in some
cases, other parties.
Macroeconomic fluctuations often account for most of the risks associated with exposures to instruments that promise to pay
fixed or variable interest rates. For an investor who is long fixed rate instruments (thus receiving these cash flows), increases in
interest rates naturally lift the discount factors applied to the expected cash flows and thus cause a loss. The longer the
maturity of a certain cash flow and the higher the move in the discount factor, the higher will be the loss. Upside surprises in
inflation, fiscal funding needs, and FX depreciation rates are among the most common adverse macroeconomic shocks to
receivers. But counterparty exposure, issuer creditworthiness, client segmentation, regulation (including changes in assets
holding limits for different types of investors), changes in tax policies, currency convertibility (which may constrain currency
conversion, repatriation of profits and/or the liquidation of positions), and settlement issues related to local clearing houses are
also important risk factors to be considered. The sensitivity of fixed income instruments to macroeconomic shocks may be
mitigated by indexing the contracted cash flows to inflation, to FX depreciation, or to specified interest rates – these are
common in emerging markets. It is important to note that the index fixings may -- by construction -- lag or mis-measure the
actual move in the underlying variables they are intended to track. The choice of the proper fixing (or metric) is particularly
important in swaps markets, where floating coupon rates (i.e., coupons indexed to a typically short-dated interest rate
reference index) are exchanged for fixed coupons. It is also important to acknowledge that funding in a currency that differs
from the currency in which coupons are denominated carries FX risk. Naturally, options on swaps (swaptions) also bear the
risks typical to options in addition to the risks related to rates movements.
Derivative transactions involve numerous risks including, among others, market, counterparty default and illiquidity risk. The
appropriateness or otherwise of these products for use by investors is dependent on the investors' own circumstances
including their tax position, their regulatory environment and the nature of their other assets and liabilities, and as such,
investors should take expert legal and financial advice before entering into any transaction similar to or inspired by the
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 42
contents of this publication. The risk of loss in futures trading and options, foreign or domestic, can be substantial. As a result
of the high degree of leverage obtainable in futures and options trading, losses may be incurred that are greater than the
amount of funds initially deposited. Trading in options involves risk and is not suitable for all investors. Prior to buying or
selling an option investors must review the "Characteristics and Risks of Standardized Options”, at
http://www.optionsclearing.com/about/publications/character-risks.jsp. If you are unable to access the website please contact
your Deutsche Bank representative for a copy of this important document.
Participants in foreign exchange transactions may incur risks arising from several factors, including the following: ( i) exchange
rates can be volatile and are subject to large fluctuations; ( ii) the value of currencies may be affected by numerous market
factors, including world and national economic, political and regulatory events, events in equity and debt markets and
changes in interest rates; and (iii) currencies may be subject to devaluation or government imposed exchange controls which
could affect the value of the currency. Investors in securities such as ADRs, whose values are affected by the currency of an
underlying security, effectively assume currency risk.
Unless governing law provides otherwise, all transactions should be executed through the Deutsche Bank entity in the
investor's home jurisdiction.
United States: Approved and/or distributed by Deutsche Bank Securities Incorporated, a member of FINRA, NFA and SIPC.
Non-U.S. analysts may not be associated persons of Deutsche Bank Securities Incorporated and therefore may not be subject
to FINRA regulations concerning communications with subject company, public appearances and securities held by the
analysts.
Germany: Approved and/or distributed by Deutsche Bank AG, a joint stock corporation with limited liability incorporated in the
Federal Republic of Germany with its principal office in Frankfurt am Main. Deutsche Bank AG is authorized under German
Banking Law (competent authority: European Central Bank) and is subject to supervision by the European Central Bank and by
BaFin, Germany’s Federal Financial Supervisory Authority.
United Kingdom: Approved and/or distributed by Deutsche Bank AG acting through its London Branch at Winchester House, 1
Great Winchester Street, London EC2N 2DB. Deutsche Bank AG in the United Kingdom is authorised by the Prudential
Regulation Authority and is subject to limited regulation by the Prudential Regulation Authority and Financial Conduct
Authority. Details about the extent of our authorisation and regulation are available on request.
Hong Kong: Distributed by Deutsche Bank AG, Hong Kong Branch.
Korea: Distributed by Deutsche Securities Korea Co.
South Africa: Deutsche Bank AG Johannesburg is incorporated in the Federal Republic of Germany (Branch Register Number
in South Africa: 1998/003298/10).
Singapore: by Deutsche Bank AG, Singapore Branch or Deutsche Securities Asia Limited, Singapore Branch (One Raffles Quay
#18-00 South Tower Singapore 048583, +65 6423 8001), which may be contacted in respect of any matters arising from, or in
connection with, this report. Where this report is issued or promulgated in Singapore to a person who is not an accredited
investor, expert investor or institutional investor (as defined in the applicable Singapore laws and regulations), they accept
legal responsibility to such person for its contents.
Japan: Approved and/or distributed by Deutsche Securities Inc.(DSI). Registration number - Registered as a financial
instruments dealer by the Head of the Kanto Local Finance Bureau (Kinsho) No. 117. Member of associations: JSDA, Type II
Financial Instruments Firms Association, The Financial Futures Association of Japan, and Japan Investment Advisers
Association. Commissions and risks involved in stock transactions - for stock transactions, we charge stock commissions and
consumption tax by multiplying the transaction amount by the commission rate agreed with each customer. Stock
transactions can lead to losses as a result of share price fluctuations and other factors. Transactions in foreign stocks can lead
to additional losses stemming from foreign exchange fluctuations. We may also charge commissions and fees for certain
categories of investment advice, products and services. Recommended investment strategies, products and services carry the
risk of losses to principal and other losses as a result of changes in market and/or economic trends, and/or fluctuations in
market value. Before deciding on the purchase of financial products and/or services, customers should carefully read the
relevant disclosures, prospectuses and other documentation. "Moody's", "Standard & Poor's", and "Fitch" mentioned in this
19 May 2015
Signal Processing
Deutsche Bank Securities Inc. Page 43
report are not registered credit rating agencies in Japan unless Japan or "Nippon" is specifically designated in the name of the
entity. Reports on Japanese listed companies not written by analysts of DSI are written by Deutsche Bank Group's analysts
with the coverage companies specified by DSI. Some of the foreign securities stated on this report are not disclosed according
to the Financial Instruments and Exchange Law of Japan.
Malaysia: Deutsche Bank AG and/or its affiliate(s) may maintain positions in the securities referred to herein and may from
time to time offer those securities for purchase or may have an interest to purchase such securities. Deutsche Bank may
engage in transactions in a manner inconsistent with the views discussed herein.
Qatar: Deutsche Bank AG in the Qatar Financial Centre (registered no. 00032) is regulated by the Qatar Financial Centre
Regulatory Authority. Deutsche Bank AG - QFC Branch may only undertake the financial services activities that fall within the
scope of its existing QFCRA license. Principal place of business in the QFC: Qatar Financial Centre, Tower, West Bay, Level 5,
PO Box 14928, Doha, Qatar. This information has been distributed by Deutsche Bank AG. Related financial products or
services are only available to Business Customers, as defined by the Qatar Financial Centre Regulatory Authority.
Russia: This information, interpretation and opinions submitted herein are not in the context of, and do not constitute, any
appraisal or evaluation activity requiring a license in the Russian Federation.
Kingdom of Saudi Arabia: Deutsche Securities Saudi Arabia LLC Company, (registered no. 07073-37) is regulated by the
Capital Market Authority. Deutsche Securities Saudi Arabia may only undertake the financial services activities that fall within
the scope of its existing CMA license. Principal place of business in Saudi Arabia: King Fahad Road, Al Olaya District, P.O. Box
301809, Faisaliah Tower - 17th Floor, 11372 Riyadh, Saudi Arabia.
United Arab Emirates: Deutsche Bank AG in the Dubai International Financial Centre (registered no. 00045) is regulated by the
Dubai Financial Services Authority. Deutsche Bank AG - DIFC Branch may only undertake the financial services activities that
fall within the scope of its existing DFSA license. Principal place of business in the DIFC: Dubai International Financial Centre,
The Gate Village, Building 5, PO Box 504902, Dubai, U.A.E. This information has been distributed by Deutsche Bank AG.
Related financial products or services are only available to Professional Clients, as defined by the Dubai Financial Services
Authority.
Australia: Retail clients should obtain a copy of a Product Disclosure Statement (PDS) relating to any financial product referred
to in this report and consider the PDS before making any decision about whether to acquire the product. Please refer to
Australian specific research disclosures and related information at https://australia.db.com/australia/content/research-
information.html
Australia and New Zealand: This research, and any access to it, is intended only for "wholesale clients" within the meaning of
the Australian Corporations Act and New Zealand Financial Advisors Act respectively.
Additional information relative to securities, other financial products or issuers discussed in this report is available upon
request. This report may not be reproduced, distributed or published by any person for any purpose without Deutsche Bank's
prior written consent. Please cite source when quoting.
Copyright © 2015 Deutsche Bank AG
GRCM2015PROD034049
David Folkerts-Landau Group Chief Economist
Member of the Group Executive Committee
Raj Hindocha Global Chief Operating Officer
Research
Marcel Cassard Global Head
FICC Research & Global Macro Economics
Richard Smith and Steve Pollard Co-Global Heads Equity Research
Michael Spencer Regional Head
Asia Pacific Research
Ralf Hoffmann Regional Head
Deutsche Bank Research, Germany
Andreas Neubauer Regional Head
Equity Research, Germany
Steve Pollard Regional Head
Americas Research
International Locations
Deutsche Bank AG
Deutsche Bank Place
Level 16
Corner of Hunter & Phillip Streets
Sydney, NSW 2000
Australia
Tel: (61) 2 8258 1234
Deutsche Bank AG
Große Gallusstraße 10-14
60272 Frankfurt am Main
Germany
Tel: (49) 69 910 00
Deutsche Bank AG
Filiale Hongkong
International Commerce Centre,
1 Austin Road West,Kowloon,
Hong Kong
Tel: (852) 2203 8888
Deutsche Securities Inc.
2-11-1 Nagatacho
Sanno Park Tower
Chiyoda-ku, Tokyo 100-6171
Japan
Tel: (81) 3 5156 6770
Deutsche Bank AG London
1 Great Winchester Street
London EC2N 2EQ
United Kingdom
Tel: (44) 20 7545 8000
Deutsche Bank Securities Inc.
60 Wall Street
New York, NY 10005
United States of America
Tel: (1) 212 250 2500