Nick Bloom Productivity and Reallocation Nick Bloom, Econ 247, 2015 Big Overview Economists started looking at establishment data in the 1990s (Haltiwanger, Davis, Bartelsman,
Download ReportTranscript Nick Bloom Productivity and Reallocation Nick Bloom, Econ 247, 2015 Big Overview Economists started looking at establishment data in the 1990s (Haltiwanger, Davis, Bartelsman,
Nick Bloom Productivity and Reallocation
Nick Bloom, Econ 247, 2015
Big Overview
Economists started looking at establishment data in the 1990s (Haltiwanger, Davis, Bartelsman, Bailey etc.) There was surprise over: • High levels of turnover • Heterogeneity within industries • The lumpiness of micro-economic activity • The importance of reallocation in driving productivity
Nick Bloom, Econ 247, 2015
Why should you be interested in this?
Important to understanding growth – e.g. 3/4 productivity growth is reallocation, unemployment driven by churn etc Second, this is a fertile area of research: • It is new – many open questions • It is hard – typically needs mix of empirics, simulation and modeling, so barriers to entry high Third, Stanford has a Census node. Census data is painful to access, but this also deters – so still low-hanging fruit (like my Grandma’s attic – some amazing stuff in there)
Nick Bloom, Econ 247, 2015
High levels of turnover
Heterogeneity within industries The lumpiness of micro-economic activity The importance of reallocation in driving productivity
Nick Bloom, Econ 247, 2015
Turnover
About 15% of
jobs
are destroyed and 20% created in the private sector every year. About 80% of this turnover occurs within the same SIC-4 digit industry This is robust across countries (US, Europe, Asia and SA) But, before I show data a couple of point on definitions: • This is turnover in “jobs”, defined in terms of
establishment
employment changes, e.g. CES • A linked (but distinct concept) is turnover in “employment” – which is two to three times higher – defined in terms of
workers
changes, e.g. CPS
Nick Bloom, Econ 247, 2015
Turnover in “Jobs” versus “Employment” – Expanding Firm example
Note: Worker flow=14, Job flow=4
Turnover in “Jobs” versus “Employment” – Contracting Firm example
Note: Worker flow=15, Job flow=9
Quarterly
Job
Flows in Private Sector, 1990-2005, BED data
employment ≈ change in unemployment (2) Gross flows are much bigger than net flows (3) Reduction in job churn that (in manufacturing) part of a longer trend (4) Job destruction does not necessarily mean firing – could be not hiring a replacement for a separation.
Nick Bloom, Econ 247, 2015
Source: John Haltiwanger (2005)
Updated: quarterly job flows continued falling in the Great Recession, particularly the creation margin
Change in unemployment rate Source: Business Employment Dynamics (BED) and CPS
Source:
Grimm, Haltiwanger and Foster (2013), “Reallocation in the Great Recession: Cleansing or Not?”
Young/Small plants have much higher flows
Source: Business Dynamics Statistics (BDS)
Source:
Grimm, Haltiwanger and Foster (2013), “Reallocation in the Great Recession: Cleansing or Not?”
Current recession – challenge is falling labor force participation (mainly low skilled men)
Nick Bloom, Econ 247, 2015
Job Flows and Employment Flows, total private (% of total)
Nick Bloom, Econ 247, 2015
Source: John Haltiwanger
Much of the turnover is creation/destruction in same SIC4 industry Excess reallocation = |job creation| + |job destruction| - |job creation-job destruction|
Nick Bloom, Econ 247, 2015
Source: John Haltiwanger, Changes defines as % over average base & end years
This is very much in the spirit of Schumpeter
Nick Bloom, Econ 247, 2015
This is very much in the spirit of Schumpeter
“The fundamental impulse that keeps the capital engine in motion comes from the new consumers’ goods, the new methods of production and transportation, the new markets... [The process] incessantly revolutionizes from within, incessantly destroying the old one, incessantly creating a new one. This process of Creative Destruction is the essential fact of capitalism.”
Schumpeter (p. 83, 1942)
Although probably his most famous quote was: “Early in life I had three ambitions. I wanted to be the greatest
economist in the world, the greatest horseman in Austria, and the best lover in Vienna. Well, I never became the greatest horseman in Austria
“ To which the (un-attributed) response was: “Those we knew Schumpeter as an Economist, Lover or a
Horseman presumed his skills were in the other two fields
” Nick Bloom, Econ 247, 2015
High levels of turnover
Heterogeneity within industries
The lumpiness of micro-economic activity The importance of reallocation in driving productivity
Nick Bloom, Econ 247, 2015
Heterogeneity basic facts
Typical gap between 10 th and 90 th percentiles of productivity within same industry is 200% (Syverson, 2004) These spreads are very persistent: • About 70% to 80% annual job-flows are persistent • About 60% to 70% annual productivity growth is persistent
Nick Bloom, Econ 247, 2015
Big TFP dispersion across firms: for example, US ready mix concrete plants:
Low competition High competition Nick Bloom, Econ 247, 2015
Source: Syverson (2004) 18
What could cause this heterogeneity?
One possibility is pure measurement error, but: • Productivity is strongly linked with exit and LR growth • When looking at micro-industries where we measure plant prices (e.g. boxes, bread, block ice, concrete, plywood, carbon black etc.) still see this spread (Foster, Haltiwanger and Syverson, 2008 AER)
Nick Bloom, Econ 247, 2015
Explanations of this heterogeneity?
Several possible economic models of the spread are: • Mistakes/learning (Jovanovic, 1982 Econometrica) • Mis-measurement: • “Hard” technology (e.g. R&D) • Skills • Other inputs (computers) or utilization • Management and managers
Nick Bloom, Econ 247, 2015
High levels of turnover Heterogeneity within industries
The lumpiness of micro-economic activity
The importance of reallocation in driving productivity
Nick Bloom, Econ 247, 2015
Lumpiness of growth
The share of employment growth generated by large adjustments is big (Davis and Haltiwanger, 1992 QJE) • More than 2/3 manufacturing job creation/destruction accounted for by +25% changes • For non-manufacturing even greater Same is true, but more extreme, for investment (Doms and Dunne, 1998 RED).
Suggests substantial adjustment-costs in factor changes
Nick Bloom, Econ 247, 2015
Lumpiness of employment growth
Nick Bloom, Econ 247, 2015
Source: John Haltiwanger, annual data manufacturing
High levels of turnover Heterogeneity within industries The lumpiness of micro-economic activity
The importance of reallocation in driving productivity Nick Bloom, Econ 247, 2015
Measuring productivity ( ω
i,t
)
Labor Productivity:
LP i
,
t
va i
,
t
l i
,
t
Three factor TFP:
TFP i
3 ,
t
y i
,
t
l l i
,
t
k k i
,
t
m m i
,
t
Five factor TFP:
TFP i
5 ,
t
y i
,
t
l l i
,
t
k k i
,
t
m m i
,
t
e e i
,
t
c c i
,
t
Note: va=log(value added), l=log(labor force), k=log(tangible capital), m=log(materials,
Nick Bloom, Econ 247, 2015
e=log(energy), c=log(IT). If IT included need to remove from tangible capital.
Defining industry (or aggregate) productivity
Define a simple industry productivity index: P t
P t
s i
,
t
i
,
t
Where: ω i,t is the productivity of establishment log(labor productivity) or log (TFP))
i
in period
t
(i.e.
s
i,t is the share of establishment
i
in the industry in period (i.e. the share of employment or sales in industry
t
employment or sales)
Nick Bloom, Econ 247, 2015
Industry productivity can increase through two channels
•
Within Firms (Traditional view)
– The same firms become more productive (e.g. new technology spreads quickly to all firms, like Internet) •
Between Firms (“Schumpeterian”view)
– Low TFP firms exit and resources are reallocated to high TFP firms • High TFP firms expand (e.g. more jobs) & low TFP firms contract (e.g. less jobs) • Exit/entry 27
Nick Bloom, Econ 247, 2015
These two effects are well known to cricket fans Within batsman (each batsman improves) Between batsman (more time for your best batsman)
28
Nick Bloom, Econ 247, 2015
Decomposing productivity (1)
Productivity growth for a balanced panel of establishments can be broken down into three terms:
P t
P t
1
s i s i
( (
s i s i
, ,
t t
i
1 ,
t
,
t
,
t
(
i
s i s i
,
t
,
t
,
t
1 1
s i
i
)
i
, ,
t t
)(
i
,
t
1
i
1 , 1
t
,
t
1 ) Within te rm Between te rm
i
,
t
1 ) Cross term Reallocation Within term is included in representative agent models, while the between and cross terms would not be
Nick Bloom, Econ 247, 2015
Decomposing productivity (2)
Allowing for entry and exit requires two more terms:
P t
P t
1
s i s i
( (
,
t
,
t s i s i
i
1 ,
t
,
t
(
s i Entry
,
t
,
t
i
s i s i
(
,
t
,
t
,
t i
,
t
1 1
Entry s i
i
)
)(
,
t
,
t i
1 ,
t i
1 ,
t
i i
,
t
1
) Within
1
Between
i
,
t Average
,
t
1
term term ) Cross term ) Entry term
s i Exit
,
t
(
i Exit
,
t
i Average
,
t
) Exit term
This is the Bailey, Hulten and Campbell (1992) decomposition
Nick Bloom, Econ 247, 2015
Total reallocation (between, entry and exit) accounts for about ½ of manufacturing TFP growth
* Source: John Haltiwanger *Combines 0.08 “between” and 0.34 “cross”
This is probably even an underestimate
(A) Treats all reallocation within establishments as “within” growth (large establishments in balanced panel have 500+ employees) (B) Reallocation terms most likely to be downward biased by miss measured prices (Foster, Haltiwanger and Syversson, 2008) So in manufacturing re-allocation of factors probably accounts for the majority of productivity growth
Nick Bloom, Econ 247, 2015
Reallocation (including entry) accounts for almost all Retail TFP growth 1 0.8
0.6
0.4
0.2
0 Retail
Source: Foster, Haltiwanger & Krizan (2000 and 2006)
Nick Bloom, Econ 247, 2015 Continuing Establishments Net entry
Differences in reallocation also a factor in explaining cross country TFP gaps
Source: Hsieh and Klenow (2008); mean=1 Nick Bloom, Econ 247, 2015
Nick Bloom, Econ 247, 2015
BACK-UP
Reallocation also appears to vary over the cycle: Usually higher in recessions except for the Great Recession (maybe because finance dictated growth rather than TFP during this?)
Normal is Zero Change in Unemployment, Mild is 0.01 Change, Sharp is 0.03 Change.
High Productivity is 1 std dev above mean, Low Productivity is 1 std dev below mean.
Source:
Grimm, Haltiwanger and Foster (2013), “Reallocation in the Great Recession: Cleansing or Not?”
The recession (falling output) is now over but the recovery (return to levels) is very slow
Unemployment rate, seasonally adjusted (Source BLS) Unemployment is still 4% above “normal” levels
Things look even worse in California
Unemployment rate, seasonally adjusted (Source BLS)
Unemployment is particularly a low-skill issue
Unemployment rate, seasonally adjusted
Although the recession could have been worse
Industrial production, normalized to 100 at the start of the recessions (Source FRB) December 2007 Great Recession (2007-2009) Great Depression (1929-1933) May 2009 0 10 20 30 Months since the start of the recession 40
Similar persistence of TFP & management Source: Bloom, Sadun and Van Reenen (2012), “Management as a technology: new empirics and old theories”, Stanford mimeo
JOLTS monthly
worker turnover
data Still massive churn – including quits – in depths of the recession (I quit a job in December 2001) Source: John Haltiwanger
Jolts – updated to 2012
Jolts – updated to 2012