Changing Patterns of Productivity and Business Dynamism: Is There a Connection?

February 2017

By

John Haltiwanger, University of Maryland

This talk, without attribution, draws on joint work with Steven Davis, Ryan Decker, Jason Faberman, Lucia Foster, Cheryl Grim, Ron Jarmin, Javier Miranda, and Zoltan Wolf

.

Source: Fernald (2014)

OECDSource: Bryne et. al. (2016)

Source: Andrews et. al. (2016)

Job Reallocation Rate, U.S. Private Non-Farm (Quarterly)

Job Reallocation Rate, U.S. Private Non-Farm (Annual)

Dashed lines are Hodrick-Prescott Trends

Declining Trend in Job Reallocation Accelerated in Post-2000 Period. Trend decline continues in post-Great Recession period.

Source: BED

Source: BDS

Declining Business Dynamism in U.S. is Evident from Multiple Data Sources

Source: BED

10.011.012.013.014.015.016.017.018.019.020.0

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

20

22

24

26

28

30

32

34

36

38

1977197919811983198519871989199119931995199719992001200320052007200920112013

Startup and Exit Rates in Nonfarm Private Sector, 1981-2014

Share of Employment for Young Firms, 1981-2014, Nonfarm Private Sector

Young businesses are much morevolatile than mature businesses.The changing age distribution of businesses accounts for about 25%of the secular decline in dynamismfrom the late 1980s to mid 2000s(Decker et al. 2014).

(Young<5)

0.06

0.07

0.08

0.09

0.1

0.11

0.12

0.13

0.14

19811983198519871989199119931995199719992001200320052007200920112013

Startup and Exit Rates

Startup Rate

Exit Rate

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

1981

1984

1987

1990

1993

1996

1999

2002

2005

2008

2011

2014Source: BDS

Possible connections between indicators of business dynamism and productivity ?

1. Increase in frictions and distortions has reduced pace of dynamism and entrepreneurship. Ubiquitous finding: Large, within industry dispersion in productivity. In healthy economy, reallocation moving resources from less productive to more

productive. An increase in frictions (e.g., Hopenhayn and Rogerson (1993)) will yield a decline

in productivity How to reconcile 1990s?

2. Decline in pace of innovation/technological change (Gordon (2016)) has led to decline in dynamism/entrepreneurship (Gort and Klepper (1982) and Jovanovic (1982))

Innovation/entry Experimentation/Dispersion Reallocation/Productivity Growth3. Structural changes due to demographics, changes in business model Unclear prediction or even benign implications for productivity?

5

0

0.01

0.02

0.03

0.04

0.05

0.06

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

2009

2011

2013

Retail High Tech

0

0.05

0.1

0.15

0.2

0.25

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

2009

2011

2013

Retail High Tech

A Tale of Two Sectors: Retail vs. High TechEmployment-Weighted Startup Rates

Share of Employment at Young Firms (Age<5)

High Tech are STEM intensive industries.Includes ICT and Bio Tech.

800000000

1E+09

1.2E+09

1.4E+09

1.6E+09

1.8E+09

2E+09

2.2E+09

2.4E+09

2.6E+09

2.8E+09

3E+09

3.2E+09

1977 1982 1987 1992 1997 2002 2007

Total Real Sales by Firm Type

Single-unit Multiunit

Share of activity accounted for bySingle Unit Establishment Firms (“Mom and Pop” Firms) has declinedFrom 50 to 35 percent. Almost all of the increase in Multi-Unit Share is from Large, National Chains

Productivity Gap between Single-Unit Establishment Firms and Large, National Chains is 25 log points.

Employment-weighted annual exitRate of Single-Units is about 8 percent. About one half of one percent for Large, National Chains.

Job Reallocation Rate for Single-Units is almost 3 times larger than for National Firms.

Shift to National Chains has been productivity enhancing and reduced volatility.Source: Foster et. al. (2016)

Increases in Frictions and Distortions?

Hopenhayn and Rogerson (1993): Canonical firm dynamics model where firms face idiosyncratic

productivity shocks, endogenous entry and exit and adjustment frictions (extension of Hopenhayn (1992) with adjustment frictions). Increased adjustment frictions imply: Reduced dispersion of firm growth rates Firms with higher realizations in productivity are less likely to grow, lower

realizations in productivity are less likely to contract/exit. Reduced aggregate productivity

Illustrative Model of Increases in Adjustment Frictions

Decker et al. (2017) consider an illustrative model of adjustment frictions (consistent with Cooper and Haltiwanger (2000, 2006), Cooper, Haltiwanger and Willis (2007, 2014) and Elsby and Michaels (2013)):

𝑉𝑉 𝐸𝐸𝑖𝑖𝑖𝑖−1;𝐴𝐴𝑖𝑖𝑖𝑖 = 𝐴𝐴𝑖𝑖𝑖𝑖𝐸𝐸𝑖𝑖𝑖𝑖𝛼𝛼 − 𝑤𝑤𝑖𝑖𝐸𝐸𝑖𝑖𝑖𝑖 − 𝐶𝐶 𝐻𝐻𝑖𝑖𝑖𝑖 + 𝛽𝛽𝑉𝑉(𝐸𝐸𝑖𝑖𝑖𝑖;𝐴𝐴𝑖𝑖𝑖𝑖+1)𝐶𝐶 𝐻𝐻𝑖𝑖𝑖𝑖

= �𝛾𝛾2

𝐻𝐻𝑖𝑖𝑖𝑖𝐸𝐸𝑖𝑖𝑖𝑖−1

2

0, 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑤𝑤𝑜𝑜𝑜𝑜𝑜𝑜+ 𝐹𝐹+max(𝐻𝐻𝑖𝑖𝑖𝑖−1, 0) + 𝐹𝐹−max(−𝐻𝐻𝑖𝑖𝑖𝑖−1, 0) 𝑜𝑜𝑖𝑖 𝐻𝐻𝑖𝑖𝑖𝑖 ≠ 0

𝑎𝑎𝑖𝑖𝑖𝑖 = 𝜌𝜌𝑎𝑎𝑖𝑖𝑖𝑖 + 𝜂𝜂𝑖𝑖𝑖𝑖𝐸𝐸𝑖𝑖𝑖𝑖 = 𝐸𝐸𝑖𝑖𝑖𝑖−1 + 𝐻𝐻𝑖𝑖𝑖𝑖

Where 𝛼𝛼 < 1 due to decreasing returns or product differentiation. Calibration of this model helps illustrate different mechanisms.

Responsiveness Slides Source: Decker et. al. (2016)

00.10.20.30.40.50.60.70.8

0

0.1

0.2

0.3

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 0.9 1Adjustment Costs (Kinked)

Responses of Key Moments to Changes in Adjustment Costs (Kinked)

Reallocation

lag TFP coeff(RightAxis)

0

0.2

0.4

0.6

0.8

0

0.1

0.2

0.3

0.4

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 0.9 1Adjustment Costs (Kinked)

Responses of Key Moments to Changes in Adjustment Costs (Kinked)

STD LP

lag LP coeff (RightAxis)

With increases in adjustment frictions:1. Declining Reallocation and Responsiveness.2. Rising Dispersion of LP.

With decreases in shock dispersion:1. Declining Reallocation and Responsiveness.2. Declining Dispersion of LP.

0

0.2

0.4

0.6

0.8

1

00.10.20.30.40.50.6

0.20 0.25 0.30 0.35 0.40 0.45 0.50STD TFP

Responses of Key Moments to Changes in TFP Dispersion (Kinked Adjustment Costs)

STD LP Reallocation lag TFP coeff(Right Axis)

By Agg TFP we mean employment-weighted micro TFP. Since unweighted mean does not vary, variation in Agg TFP is isomorphic to the OP Covariance.

0

0.1

0.2

0.3

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 0.9 1Adjustment Costs (Kinked)

Responses of Key Moments to Changes in Adjustment Costs (Kinked)

Agg TFP

OP COV TFP

�𝑃𝑃𝑖𝑖 = �̅�𝑝𝑖𝑖 + 𝑐𝑐𝑜𝑜𝑐𝑐(𝜃𝜃𝑓𝑓, 𝑝𝑝𝑓𝑓 , Olley-Pakes (OP) Decomposition of industry-level productivity insightful here. OP covariance using either TFP or LP declines with increase in adjustment costs.

0

0.05

0.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 0.9 1Adjustment Costs (Kinked)

Responses of Key Moments to Changes in Adjustment Costs (Kinked)

OP COV LP (RightAxis)

0

0.2

0.4

0.6

0.8

1980s 1990s 2000s

AR(1) Coefficient

High Tech Non Tech

Dispersion and Persistence

Little evidence that changes in persistence drive patterns of reallocation

0

0.1

0.2

0.3

0.4

1980s 1990s 2000s

Std Deviation of Innovations

High Tech Non Tech

Patterns for innovations mimic overall shocks

0

0.1

0.2

0.3

0.4

0.5

1980s 1990s 2000s

Dispersion of TFP

High Tech Non Tech

Shock Processes in Manufacturing

0.000.050.100.150.200.250.300.350.400.45

1980s 1990s 2000s

Young Mature

Employment Growth Marginal Response of Plant-Level Employment Growth and Investmentto TFP for High Tech – Results from estimating plant-level regressions of outcomes on lagged TFP realizations

• Increased responsiveness during 1990s for young firm plants in High Tech

• Decreased responsiveness during 2000s for both young and mature firm plants in High Tech

0

0.05

0.1

0.15

0.2

0.25

1980s 1990s 2000s

Young Mature

Investment in Capital Equipment

Implications for Aggregate (Industry-Level) Productivity

Start with (industry) aggregate productivity:

𝑃𝑃𝑖𝑖 = �𝑖𝑖

𝜃𝜃𝑖𝑖𝑖𝑖𝑃𝑃𝑖𝑖𝑖𝑖

Reallocation contribution to prod. growth:

𝑃𝑃𝑖𝑖+1𝐶𝐶 = �𝑖𝑖

𝜃𝜃𝑖𝑖𝑖𝑖+1𝑃𝑃𝑖𝑖𝑖𝑖 𝑃𝑃𝑖𝑖+1𝐶𝐶 − 𝑃𝑃𝑖𝑖

Model-based 𝜃𝜃𝑖𝑖𝑖𝑖+1 ⇒ counterfactual 𝑃𝑃𝑖𝑖+1𝐶𝐶 − 𝑃𝑃𝑖𝑖 (with and without change in responsiveness) => Diff-in-diff

𝜃𝜃𝑖𝑖𝑖𝑖 = employment weight, 𝑃𝑃𝑖𝑖𝑖𝑖 = plant TFP, Correlation with traditional measures about 0.8

Agg. prod. growth accounted for by reallocation (essentially Change in OP covariance for fixed 𝑃𝑃𝑖𝑖𝑖𝑖)

24

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

Difference in contribution of reallocation toWithin industry Productivity Growth betweenModel estimates with trend and model Estimates with Responses in 1980

Diff-in-Diff Accounting Counterfactual for Changing Responsiveness in High Tech

Economy Wide Much more difficult to construct measures of shock processes Instead measure Revenue Labor Productivity (RLP) New Comprehensive Firm-Level Database Exclude the financial sector (private, non-farm, non-financial)

Distributions of RLP will reflect shocks and frictions (dispersion endogenous) High (low) TFPR/RLP should grow (shrink) as they will have high (low)

Marginal Revenue Products. Covariance between growth and these measures still informative. Focus on relative productivity within detailed industries.

Implication: Both changes in dispersion of measured productivity and covariance between

measured productivity and growth informative moments.

Rising Within Industry LaborProductivity Dispersion (Gross Output Per Worker) Within Age Groups

Within 6 digit NAICS Industries, 90-10Differential

Within Industry Labor Productivity Dispersion, High Tech, by Firm Age

Within Industry Labor Productivity Dispersion, All Sectors

1.5

1.7

1.9

2.1

2.3

2.5Mature

Young1.6

1.7

1.8

1.9

2

2.1

2.2

2.3

2.4

2.5

2.6 Mature

Young

Young dispersion > Mature dispersionConsistent with Young facing more frictions, engaged in learning and experimentation.

Dispersion rises within age groups post 2000.Difficult to reconcile with Gort-Klepper dynamics

Reduced Responsiveness of Employment Growth to Productivity in 1997-2013(Cov(growth,productivity) is declining)

0.15

0.2

0.25

0.3

0.35

0.4

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

Young Tech

Old Tech

Young Non Tech

Mature NonTech

-0.14

-0.12

-0.1

-0.08

-0.06

-0.04

-0.02

0

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

Young Tech

Old Tech

Young NonTechMature NonTech

Overall Net Employment Growth(inclusive) of Exit has become lessResponsive to Productivity

Exit has become lessResponsive to Productivity

Declining responsiveness is consistentWith rising dispersion in RevenueLabor Productivity

Reduction in Contribution of Reallocation to Productivity from Reduced Responsiveness, Tech vs. Nontech (Diff-in-Diff counterfactual)

Each point reflects immediategains in specified year

if responsiveness returned to 1997 rates with currentyear dispersion (latter partly

reflects accumulatedeffects of declining responsiveness)-0.1

-0.08

-0.06

-0.04

-0.02

0

1996 2001 2006 2011

Tech Nontech

Gort-Klepper Dynamics?Innovation/entry Experimentation/Dispersion Reallocation/Productivity Growth

20Source: Foster et. al. (2017)

Some evidence of Gort and Klepper Dynamics in High Tech:

1. Surge of Entry (proxy for innovative period) leadsto immediate rise in dispersion and lagged rise in productivity. 2. But these dynamics can’t account for increasein within industry dispersion post 2000 (IQRincreases by more than 10 log points for both young

and mature firms in post 2000 period). Entry is declining over this same period. Based on Gort-Klepper dynamics we would have expected a decline in dispersion.

-0.01

-0.005

0

0.005

0.01

0.015

Years 1-3 Years 4-6

Changes in Productivity Dispersion and Growth from a 1% (one time) Increase in Entry Rate,

High Tech

Dispersion Growth

In OECD, declining dynamism with Declining entrepreneurship

And

Rising Within Industry Dispersion ofRevenue Labor Productivity

Source: Andrews et. al. (2016)

Taking Stock Different dynamics across sectors: Retail Trade: Structural change yielded decline in dynamism, entrepreneurship and rise in

productivity.

High Tech: Rise and decline in entrepreneurship, dynamism and productivity. Which way does causality run?

Declining responsiveness and rising labor productivity dispersion in post 2000 period consistent with rising frictions/distortions.

Is there evidence of Gort-Klepper dynamics? Yes but dispersion rises rather than falls in post 2000 period with declining entry.

Rising Frictions/Distortions? Labor market (e.g., Occupational Licensing, Employment

at Will) Decline in competition (e.g., winner takes all sectors make

it more difficult to identify and enforce exclusionary practices) Financial market regulation (e.g., Sarbanes-Oxley, Dodd-

Frank) Zoning restrictions in information-centric locations? (Hsieh

and Moretti, 2015)

Other mechanisms/channels

24

8.0

13.0

18.0

23.0

28.0

33.0

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

Worker Reallocation (H+S)

"Excess" Churning (H-JC=S-JD)

Job Reallocation (JC+JD)

Decline in indicators of dynamism (job reallocation/entry) part of broader decline in labor market fluidity.

The latter has implications beyond thosediscussed here for productivity:1. Labor force participation2. Earnings growth3. If match quality has declined this alsohas implications for productivity.

See Davis and Haltiwanger (2014).