Economic Downturns, Inventor Mobility, and Technology Trajectories

Erica R. H. Fuchs

Department of Engineering and Public PolicyCarnegie Mellon University

erhf@andrew.cmu.edu

(from papers co-authored with Akinsanmi, Nugent, Reagans, Ventura, and Yang)

Main take-aways

Superstars Non-stars

2

Economic Downturns, Inventor Mobility, and Technology Trajectories

How do sector-specific business cycles (during bubble and post-burst) affect(1) quantity (patents) and (2) direction (emerging GPT-enabler versus rest of the field)

of innovation?Are inventor innovation outcomes in part explained by the inventor's

mobility into and out of telecom? Do star inventors respond in a different way from non-stars?

3

(Akinsanmi, Reagans, Fuchs (2015) Seeing Rainbows While Others Flee: How innovation in the most advanced technologies grew after the burst of the telecommunications bubble. Carnegie Mellon University Working Paper. )

Data: US Optoelectronics Inventors Pre-Burst USPTO Population

– >70,000 US OE Inventors and >175,000 OE Patents total– Patent Data missing (1) inventors who move if they don’t patent

afterwards (2) Career beyond active patenting (3) Inventor background

CV Subsample– 729 US OE Inventors and 12,400 Patents– Contacts provided by SPIE, OSA and IEEE’s Photonics Society

Oral Histories

INVENTORS TARGET SAMPLE CV SAMPLE RESPONSE RATE

Top 1.5% by Total Patents 760 237(78 overlap-Superstars)

30%; 73% of those reached

Top 1.5% by Patents/Year 680 233 34%; 82% of those reached

All Inventors in Emerging GPT Enabler 900 182 20%, 54% of those reached

RANDOM SAMPLE of Non-Emerging GPT Enabler Inventors in OE

1250 180 15%; 83% of those reached

Difference in Difference, Two-stage model

4(Akinsanmi, Reagans, Fuchs (2015) Seeing Rainbows While Others Flee. CMU Working Paper. )

consulting

Hand Coded >2000 firms into 20 market applications in each year

divers. w/ telec

academia

all telecom

defense

divers. w/o telec

1956

con

sent

dec

ree

limiti

ng B

ell L

abs

oper

atio

n

1962

: Exp

ansi

on 1

beg

ins

1969

: Rec

essi

on 1

beg

ins

1971

: Exp

ansi

on 2

beg

ins

1973

: Rec

essi

on 2

beg

ins

1976

: Exp

ansi

on 3

beg

ins

1980

: Rec

essi

on 3

beg

ins

1983

: Exp

ansi

on 4

beg

ins

1984

: Bre

ak u

p of

Bel

l Lab

s

1990

: Rec

essi

on 4

beg

ins

1992

: Exp

ansi

on 5

beg

ins

1996

: Bel

l Lab

s di

vide

d ag

ain

2000

: Rec

essi

on 5

beg

ins

2002

: Exp

ansi

on 6

beg

ins

2008

: Rec

essi

on 6

beg

ins

PRE- TELECOM BUBBLE

1940 1950 1960 1970 1980 1990 2000 20100

20

40

60

80

100

120

140

160

180

Year

Nu

mb

er o

f in

ven

tors

in

eac

h m

arke

t ap

pli

cati

on

BUBBLE

POST-BURST

defense

5(Akinsanmi, Reagans, Fuchs (2015) Seeing Rainbows While Others Flee. CMU Working Paper. )

Bubble burst disproportionately reduced innovation in rest of field compared to emerging GPT-enabler (e.g. Field, 2011)

Super-stars advance the emerging GPT enabler during resource-constrained parts of the business cycle

Their efforts build on the efforts of non-stars during less constrained times

Bubble: Herd mentality by non-stars into emerging GPT enabler

Post-burst: Super-stars see tech. opportunity despite downturn; Majority of emerging GPT-enabler inventors leave field,

dislocated from their IP, stop innovating (Yang et al 2015, Yang & Fuchs WP)

Preliminary Findings

6 (Akinsanmi, Reagans, Fuchs (2015) Seeing Rainbows While Others Flee. CMU Working Paper. )

The Super-stars: In their own words

Quote

“By late 2000 we knew the bubble had burst. I was at Intel Capital and I knew it had burst because we had ratcheted down all the terms on our term sheets. And I went and started a firm the next year. It was like going into the eye of the storm. I was either being an entrepreneur or being very stupid. But I had a novel technology I believed in, I had access to some capital, and I could assemble a world class team” – M. L.

Outcome

Acquired in 2003

“It was a natural evolution to start a company… I didn’t think about bubble or burst. I had a niche technology for short-distance data-com and that market continued to grow…year by year ” – Anon.

Acquired in 2004

“You’re starting a new company … in the start of the worst crash … unaware of whether you’re going to come out of it okay or not. But we decided it was a good challenge. It was a very exciting time to try and do that. The competition at that time, all they were doing was trying to protect against a downside of their current business as opposed to [investing] in the future of where the business needed to go because their revenues were dropping. They were cost-cutting. They were trying to save programs. And they weren’t able to invest in new technology” – D. W.

Initial Public Offering in 2007

7(Akinsanmi, Reagans, Fuchs (2015) Seeing Rainbows While Others Flee. CMU Working Paper. )

The Non-Superstars: Majority leave field, stop innovating

“These days… you cannot find a research-type job. There are very very very few. It’s not like the old days that companies spend a lot of money on research. It’s more… development engineering.”

“It was a tough time in the job market and so I was happy to just find a job. I… [joined] the yield group [in a computing company]… it was my first non-optical-electronics job.”

“They offer[ed] some positions in the headquarters but nobody took it…. families are here, right,… and the positions they were offering were not related to what we did before.”

(Fuchs, Nugent, Yang (2015) Gains from Others Losses: Technology Trajectories and the Global Division of Firms. Research Policy. Accepted. )

The Non-Superstars: Dislocation of Inventors from their IP

“So… every company I’ve ever been with, you sign your rights away to any inventions you make. That’s straight up, you sign away for your salary.”

“…they shut down the foundry, 'cause they can purchase the same function although it's bigger or more bulky optics, discrete optics…we always think about whether we can bring the same technology we developed into some real use, because it has value…. but it’s protected by the patents.”

“[At first after being acquired] I kept filing lots of patents… other companies wanted to license the patents but they made it very difficult to license…. so I would write patents, but then nobody could use the inventions because [Firm] didn’t make that type of product and other people couldn’t license them.”

9 (Fuchs and Yang (2015) The Dislocation of Inventors from their IP. CMU Working Paper. )

The Challenges Measuring Mobility

CVs: disentangles relationship between mobility &

patenting; inventor characteristics; limited by sample; HARD! (Ge et al 2015, Akinsanmi et al WP)

Past disambiguation methods: 10-22% error on available sample closest to full USPTO (splitting), systemic biases (context)

Supervised learning method: maintains under 3% errors on all available

samples (Ventura et al 2015)

Magnitude of different sources of error?Disambiguation? Endogeneity?

10

Smal

l Sam

ple

Smal

ler

and

Lar

ger

Sam

ples

Thank You

11

Main Take-Aways

Theory: Economic downturnsBubble burst disproportionately reduced innovation in rest of

field compared to emerging, general purpose technology (GPT) - enabler (e.g. Field, 2011)

Super-stars advance the emerging GPT during resource-constrained parts of the business cycle (Akinsanmi et al WP)

Majority of emerging GPT inventors leave field, dislocated from their IP, stop innovating (Yang et al 2015, Yang and Fuchs WP)

Methods: Measuring mobilityPast disambiguation methods: 10-22% errors on available

sample closest to full USPTO, systemic biasesSupervised learning method: maintains under 3% errors on

all available samples (Ventura et al 2015)

CVs: disentangles relationship between mobility & patenting; limited by sample (Ge et al 2015, Akinsanmi et al WP)

Tbd: disambiguation or endogeneity greater source of error?12

A Sector-Specific Economic Downturn

1998 1999 2000 2001 2002 2003$0

$5

$10

$15

$20

$25

$30

$35

$40

$45

$50

Year

Actual vs. Forecast U.S. Fiber-Optic Market Sizes

Actual Demand

Forecast Demand

(Cahners Business Information 2000, Turbini & Stafford 2003)

Telecommunications Bubble Burst(March 2000)

Year

US

D (

Bill

ion

s)

1976 1981 1986 1991 1996 2001 20060

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Pat

ents

sub

clas

s/P

aten

ts s

ubcl

ass

in 2

002

Emerging GPT Enabler Takes Off while Rest of Field Plateaus

Rest of Field(N = 184274)

Emerging GPT Enabler(N = 3237)

Data up to 2010 Telecommunications Bubble Burst

Patenting in the Rest of Field rises steadily, reaches its peak in 2002, and then declines Patenting in the emerging enabling technology rises quickly from the mid 1990s, peaks in

2002 and subsequently plateaus

Patenting in Optoelectronics

14

Economic downturns have disparate effects on different firms, people and technologies

Firms

Great Depression most technologically progressive decade of the century, Field ‘03

People

Innovation Trajectories

Economic DTs in general Great Depression Post Telecom Bubble Burst

• Start-ups more likely to fail than larger firms (Geroski & Gregg, 1997)

• New firms less likely to get VC funding (Paik et al, 2013)

• Firm survival dependent on pre-downturn productivity (Bresnahan and Raff, 1991)

• Firm survival dependent on pre-downturn growth strategy (Goldfarb et al, 2006)

• Jobs created low-paying and temporary (Bowlus, 1993, Davis et al 1996)

• Firms more likely to train incumbents while reducing recruitment of new employees (Brunello, 2009)

• Employment of research scientists grew (Mowery and Rosenberg, 1989)

• Technology centers had the highest unemployment rates (Gittel and Sohl, 2005)

• Newest process and product innovations continue to be created (Caballero and Hammour, 1994, Shu, 2012)

• Firms invest in product innovation rather than process innovation (Brechicci et al, 2013)

• 1930s experienced very high rates of innovation (Field, 2011)

• Timing of early stage R&D changed, changing technology trajectories (Nicholas and Nabar, 2009)

?15

Stars may be better able to benefit from mobility

16

Mobility • Mobile inventors more productive than non-mobile (Hoisl 2007).

• Are moves to similar or different contexts most advantageous? - Heterogeneously networked team more productive than homogenous

(Reagans and Zuckerman, 2001);

- Knowledge transfer increases with technological distance (Rosenkopf & Almeida, 2003, Song, Almeida & Wu, 2003)

- For acquisitions, high routine overlap and moderate skill overlap (Kapoor &

Lim, 2007) and prior communication (Agarwal et al, 2012) leads to higher productivity

Stars Star involvement matters: close ties between academics and firm scientists needed for commercialization (Zucker & Darby, 1998)

Star arrival leads to 38% increase in department productivity (Agarwal et al). Star death leads to lasting decline in collaborators’ quality-adjusted publication rates (Azoulay et al, 2010, Oettl, 2012).

Stars less likely to leave their firms (Campbell et al, 2011, Carnahan et al, 2012),Stars who move draw level with or overtake non-movers in productivity

(Hoisl, 2009).

Case: Optoelectronics (OE) Industry

Intersection between electronics and photonics; conversion of electric signals to light signals and vice-versa

Photons:– higher information carrying capacity – lower power consumption

Innovation in OE first driven by telecom but central to advances in computing, biomedical, energy and military

General Purpose Technology “has the potential to be extremely pervasive and used as inputs by a wide range of sectors in the economy” (Helpman and Trajtenberg, 1998)

17

Emerging GPT Enabler: OE Integration

OE Integration: incorporates multiple devices onto a single chip, enabling reduced size and allowing OE access to broader set of markets (Eng, 2010)

OE integration inventors: opportunity to switch market applications while leveraging same technical competency

Traditional Architecture Integration

Ferry, 2010

Source: LuteraSource: U. Wash Source: Stanford U

18

Hypotheses

H1. During the bubble: Inventors who move into telecom increase their patenting in both the

rest of field and the emerging GPT enabler.

H2. Post-burst: Inventors who move out of telecom increase their patenting in the

emerging GPT enabler but not in the rest of field (e.g. Field 2011)

Star inventors who move out of telecom disproportionately increase patenting in emerging GPT enabler but not in rest of field (e.g. Hoisl 2009)

19

Execution: Dr. Eugene Arthurs, Krisinda Plenkovich and the SPIE organizationThe Optical Society of America and IEEE’s PhotonicsSam Ventura & Rebecca NugentWillis Chang, Carl Glazer, Farjad Zaim, Sabrina Larkin, Neha Nandakumar,

Angela NgEarly feedback:

Jeff Furman, David Hounshell, and M. Granger MorganShane Greenstein, Scott Stern and Michael Piore Lee Branstetter, Brian Kovak, and Fiona MurrayCarliss Baldwin, James Utterback, Rajshree Agarwal, Scott Stern and the other

participants of INFORMS, ISA, CCC and WTIC conferencesFinancial support:

National Science Foundation Science of Science and Innovation Policy Program

Schlumberger Foundation

Acknowledgements

20

sample overlaps

21