Test bank for advanced assessment interpreting findings and formulating diffe...
The Global Productivity Slowdown, Technology Diffusion and Public Policy
1. THE GLOBAL PRODUCTIVITY
SLOWDOWN, TECHNOLOGY
DIVERGENCE AND PUBLIC
POLICY:
A FIRM-LEVEL PERSPECTIVE
Dan Andrews *
Chiara Criscuolo +
Peter Gal *
*Economics Department
+Directorate for Science, Technology and Innovation
2. Differences in GDP per capita mostly
reflect labour productivity gaps
Percentage differences compared with the upper half of OECD countries, 2013
Source: OECD Going for Growth Database.
3. Weak productivity underpins the
collapse in OECD potential growth
Contribution to potential per capita output growth
Source: OECD Economic Outlook 2016, Volume 1.
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Capital per worker MFP
Potential employment rate Active population rate
Potential per capita growth (%)
Pre-crisis: MFP story Post-crisis: K story
4. Optimists:
• Brynjolfsson
• McAfee
• Mokyr
• Jovanovic
• …
The slowdown has ignited a
spirited debate about prospects…
Pessimists:
• Gordon
• Cowen
• Thiel
• Fernald
• …
5. 1. Technological factors
– Adoption and diffusion of general purpose technologies
(Griliches, 1957; David, 1991; Jovanovic and Rousseau, 2005)
– A “return to normal” after a decade of exceptional IT-
fueled gains (Fernald, 2014)
2. Rising resource misallocation (Gopinath et al., 2015)
3. Cyclical factors
– Demand conditions and monetary policy
(Anzoategui et al., 2016)
– Weak credit post crisis
(Riley, Bondibene and Young, 2015)
4. Measurement (Byrne, et al., 2016; Syverson, 2016)
Why did aggregate productivity
growth slowdown?
6. • The debate (e.g. Gordon vs Brynjolfsson) has centred
on innovation prospects at the global frontier (GF) but
we know little about GF firms.
• Our firm level analysis suggests:
1. Labour productivity (LP) at GF remained robust but
laggard firms increasingly fell behind.
2. LP divergence reflects MFP divergence and possibly
technological divergence, broadly defined
3. Some explanations: “winner takes all” dynamics;
stalling diffusion; decline in market dynamism and
increase in misallocation
4. Policy weakness potentially amplified MFP
divergence and the productivity slowdown
Our contribution: bring more micro
evidence to a largely macro debate
7. 1. Conceptual background
2. Data and measurement
3. Productivity divergence
4. Policy analysis: role of competition
5. Future research
Outline
9. • Widespread heterogeneity in firm productivity
(Bartelsman and Doms, 2000)
need to look beyond averages / aggregates
1. Neo-Schumpeterian growth (Aghion and Howitt, 2006)
a) The best (“global frontier”) firms innovate
b) These technologies diffuse to other firms
This raises within-firm productivity through catching-up
2. Reallocation via growth of productive firms and the
downsizing / exit of less productive ones
(Caballero and Hammour, 1994)
Conceptual background:
What drives productivity growth?
11. Data
Cross-country firm-level data Orbis
• Wide coverage
• 24 OECD countries, 1997-2014
• Both manufacturing and services
• Large and small firms
• Balance sheets and income statements
from several million company accounts
• Collected and harmonized by Bureau van Dijk
• Main limitation:
• Coverage of (small) firms uneven
20+ employees subsample
Extensive robustness checks
12. Measurement
• Substantial work to translate balance sheet information to
economic measures
• Deflation, PPP conversion, capital stock estimation, cleaning
• Several productivity measures derived
• Labour-productivity, MFP variants
• Definition of global productivity frontier:
Firms with highest levels of productivity
• More precisely: top 5% of firms in terms of productivity for
each 2-digit industry
14. The global frontier: Who are they?
• Frontier firms have more / larger
• market shares
• capital intensity
• wages
• mark-ups
• patents
• share of MNEs
• Productivity gap is also higher in services
• Frontier is composed of various countries
… and more so in services
than in manufacturing
16. Average of MFPR (Wooldridge) across each 2-digit sector (log, 2001=0)
... largely reflects MFPR divergence
Frontier
Frontier
Laggards Laggards
Capital deepening plays
less of a role
17. ... which may reflect “technological”
divergence
Average of mark-up adjusted MFPR across each 2-digit sector (log, 2001=0)
Divergence remains after correcting
for mark-ups behaviour
Frontier
Frontier
Laggards
Laggards
18. Robust to:
• Different MFP measures and mark-up corrections
• Frontier definitions (Top 100, Top 10%)
• More sustained performance (3-year averages of productivity)
• More narrowly defined industries (3 and 4 digit)
• Retaining only groups and standalone firms
• Comparing frontier with official industry aggregates
• Longer period using industry-level data: increased divergence from the
early 2000s compared to 1985-2000
Consistent with others’ findings:
• For the US using Census data (Decker et al, 2016)
and for financial returns (Furman and Orszag, 2016)
• MultiProd (Berlingieri et al, 2016)
• CompNet (ECB, 2016)
Productivity divergence…
19. Firm-level patterns consistent with
sectoral data
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
Frontier firms
Laggard firms
Sectoral data
Average of labour productivity across each 2-digit sector (log, 2001=0)
21. What does it mean?
• Best provider takes most of the market
• Second-best takes significantly less
Proximate drivers?
• Very low marginal costs easier scalability (Brynjolfsson and McAfee, 2011)
– Both production / provision and transport / communication costs
• Network effects across consumers
• Larger markets increase incentives to innovate (Acemoglu and Lynn, 2004)
Structural drivers?
• Digitalization + rise of intangibles + globalization
How do we spot it?
• Look at sectors where these are most relevant: ICT services
Winner-takes-all dynamics?
22. Technological divergence:
winner takes all dynamics?
MFP divergence
ICT-intensive services Non ICT-intensive services
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
Frontier firms
Laggards
Top 10%
Top 2%
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
Frontier firms
Laggards
Top 10%
Top 2%
24. Stronger MFPR divergence –
weaker aggregate MFP
Industry aggregate MFP and within-industry MFPR gap (1998-2007)
Data averaged across 12 OECD countries and purged of industry and year fixed effects
25. The speed of convergence to the
frontier slowed, even before the crisis
Convergence parameters estimated from a neo-Schumpeterian model
Dotted line: 95% confidence intervals
A: MFPR B: Mark-up adjusted MFPR
icstcts
j
isct
j
jj
ticst
j
j
icstFcsticst XDgapgapAA 413121 *lnln
26. Entry to the frontier has become more
entrenched amongst top firms
B: Mark-up corrected MFPR
From
top
20%
From
top
10%
From
top
5%
Manufacturing Services
0
10
20
30
40
50
60
70
80
0
10
20
30
40
50
60
70
80
From
top
20%
From
top
10%
From
top 5%
Proportion of frontier firms in time t according to their frontier status in t-2
27. Technological divergence: is declining
market contestability an issue?
Share of firms
Percent
Entry rates using
business registers from DynEmp
Declining firm turnover: fewer
young firms, while marginal firms
increasingly survive.
0
4
8
12
16
20
24
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Young firms (0-5 years)
Mature firms (6-10 years)
Non-viable old firms (older than 10 years)
Decline in entry is widespread
across countries
0
2
4
6
8
10
12
14
16
1998-2000 2001-04 2005-08 2009-13
28. Technological divergence: is declining
market contestability an issue?
Share of firms
Percent
MFPR relative to viable old firms
Log point differential
Higher productivity threshold for
entry, while marginal firms survive
despite a collapse in their MFPR
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Young firms (0-5 years)
Mature firms (6-10 years)
Non-viable old firms (older than 10 years)
Declining firm turnover: fewer
young firms, while marginal firms
increasingly survive.
0
4
8
12
16
20
24
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Young firms (0-5 years)
Mature firms (6-10 years)
Non-viable old firms (older than 10 years)
30. Pro-competitive product market reforms can promote the
diffusion of frontier technologies by:
1. Sharpening the incentives for incumbent firms to adopt
better technologies.
2. Raising managerial quality, which is complementary to
adoption.
3. Reducing entry barriers: young firms possess a
comparative advantage in commercialising leading
technologies.
4. Raising returns to “innovation” in downstream
manufacturing sectors via input-output linkages.
5. More efficient resource reallocation.
Competition, innovation and
diffusion
31. The pace of deregulation in
services has slowed
The restrictiveness of product market regulations
Notes: The horizontal line in the boxes represents the median, the upper and lower edges of each boxes reflect the 25th and
75th percentiles and the markers on the extremes denote the maximum and the minimum across countries.
A: Network industries B: Professional Services
32. Slower product market reform:
a larger increase in the gap
Selected industries; annual average change over time and across countries
Note: The figure shows the annual change in the (log) MFPR gap between the frontier and laggard firms and
the change in the (log) PMR indicator. Technical services refer to architecture and engineering.
33. Higher MFP divergence when
market reforms in services lagged
(1) (2) (3) (4)
0.205*** 0.231*** 0.332*** 0.311**
(0.065) (0.083) (0.103) (0.132)
Country fixed effects YES NO YES NO
Industry fixed effects YES YES YES YES
Year fixed effects YES NO YES NO
Country X year fixed effects NO YES NO YES
Observations 458 458 376 376
R-squared 0.201 0.323 0.327 0.463
Y: Δ MFP gap Y: Δ Mark-up corrected MFP gap
Δ Product Market
Regulations,c,t
Notes: Cluster robust standard errors (at the industry-year level) in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Both the
MFP gap and the PMR indicator are measured in log terms. The MFP gap is calculated at the country-industry-year level,
by taking the difference between the global frontier and the average of log productivity of non-frontier firms.
MFP divergence and product market regulation in services
Estimation method: five-year long differences (1998-2013)
tcstsctcs
ld
tcs
ld
tcs
ld
EPMRMFPgap ,,,,2,,10,,
34. 2.3% 2.3% 2.4%
3.8% 3.9%
1.7%
1.0%
1.4%
1.7% 1.7%
0%
1%
2%
3%
4%
5%
Transport Energy Retail Legal and
accounting
services
Technical
services
Observed increase in gap
Increase in gap due to slow deregulation
Sluggish market reform effort in
services amplified MFP divergence
Estimated contribution to the annual change in the MFP gap of the
slower pace of reform relative to the fastest reforming industry (telecoms)
MFP divergence was perhaps inevitable due to structural changes in the
global economy but policy could have worked harder
35. • The slowdown in aggregate productivity growth
masks an increasing divergence between GF
and laggard firms:
– Structural trends in the global economy unleashed
winner takes all dynamics and made adoption more
difficult.
– Thus, MFP divergence was partly inevitable but the
policy framework didn’t sufficiently adapt to these
structural trends
– Evidence of declining market contestability is a real
worry
Summary
36. • “Zombie” firms and firm exit in ECO
– The prevalence and resources trapped in marginal firms has risen,
with adverse effects for investment, capital reallocation (MFP) and
potential growth. “The Walking Dead”
– Insolvency regimes and productivity performance
• Distributed microdata projects in STI
– DynEmp (23 countries):
• New evidence on the growth dynamics of start-ups and
incumbents and its link to productivity
– MultiProd (18 countries):
• Investigates productivity patterns and the extent to which
different policy frameworks can shape productivity
• Productivity heterogeneity and Misallocation
• Granularity
• Wage inequality: “The Great Divergence(s)”
Ongoing work
37. • Future work
– Complementary factors (e.g. management, skills)
– The role of digitalisation
– Antitrust regulation need to be strengthened?
– Lobbying blocking penetration of ICT and new
business models in services
– Intellectual property regimes need updating?
Areas for future work