Javed Jussa Macro and Micro JobEnomicspages.linkup.com/rs/458-RJT-465/images/DeutscheBankReport.pdf · Deutsche Bank Markets Research Global Quantitative Strategy Signal Processing

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

[email protected]

George Zhao

[email protected]

Yin Luo, CFA

[email protected]

Miguel-A Alvarez

[email protected]

Sheng Wang

[email protected]

Gaurav Rohal, CFA

[email protected]

Allen Wang

[email protected]

David Elledge

[email protected]

Kevin Webster

[email protected]

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


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


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


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


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


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


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


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



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


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


Utilities

Materials

Energy

Consumer Staples

Health Care

Industrials

Financials

Consumer Discretionary

Information Technology



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




Interestingly, most of the

active jobs are dominated

within the consumer

discretionary sector

19 May 2015

Signal Processing


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)



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


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


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


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


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


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


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


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



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


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



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



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


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



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


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.


based on the number of

active jobs per company

within a month or year time

frame


based on the number of total

employees per company

19 May 2015

Signal Processing


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


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


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%



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



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


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



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


L/S

Fra

ctile

Po

rtfo

lio

2009 2010 2011 2012 2013 2014 2015

0.8

0.9

1.0

1.1




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


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


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


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


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


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


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


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


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


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


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



19 May 2015

Signal Processing


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


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 (%

)


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


Figure 46: Correlation of non-farm payrolls and average


Figure 47: Correlation of unemployment rate and average


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


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 (

%)


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


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


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


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)


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


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


Figure 53: Autocorrelation and partial autocorrelation analysis

NFP is best predicted using the previous 3 month

values


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


Figure 54: Actual NFP change versus ARIMA forecast

-800

-600

-400

-200

0

200

400

600

800

Actual NFP ARIMA Model Predicted NFP


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


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



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)


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


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


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


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


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


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)



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)



19 May 2015

Signal Processing


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


Food & Staples Retailing

Household & Personal Products

Automobiles & Components


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


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 (%)


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


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



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)


Lastly, we test whether our

industry rotation strategy

performs well in labor force

intensive sectors like retailing

19 May 2015

Signal Processing


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 (%)


19 May 2015

Signal Processing


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


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


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


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


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

Documents

Javed Jussa Macro and Micro JobEnomicspages.linkup.com/rs/458-RJT-465/images/DeutscheBankReport.pdf · Deutsche Bank Markets Research Global Quantitative Strategy Signal Processing