1. 22 May 2018
PATTERNS OF FIRM LEVEL PRODUCTIVITY IN IRELAND
RESULTS FROM MULTIPROD MODEL
Javier Papa, Luke Rehill and Brendan O’Connor
NERI Labour Market Conference, 22nd May 2018

2. Outline
• High level macro picture
• Firm level productivity analysis
• Rationale
• The MultiProd model
• Data
• MultiProd Results (2006-2014)
• Concentration measures – impact of large firms
• Productivity distribution – the best vs the rest
• Efficiency of resource allocation
2

3. 3
Global trends in productivity growth
Source: Conference Board Total Economy Database
 Global productivity slowdown since
1990s
 Slowdown accelerated pre-crisis
 OECD countries slowed down the most
 Patterns are consistent across
measures of productivity (LP & MFP)
-3
-2
-1
0
1
2
3
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
OECD productivity growth
Labour Productivity MFP

4. 4
High level of labour productivity
GDP and GNI per hour worked (2015 USD - 2011 PPPs)
Source: OECD
67
53
30
35
40
45
50
55
60
65
70
75
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
France Germany Ireland (GDP) United Kingdom
United States Japan OECD - Total Ireland (GNI*)

5. 5
Decline in growth rate
-6%
-4%
-2%
0%
2%
4%
6%
8%
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Labour Productivity MFP LP trend MFP trend
Source: CSO experimental estimates of productivity (forthcoming)
Year-on-year productivity growth in Ireland

6. Need for firm-level productivity analysis
6
Aggregate productivity statistics hide underlying drivers
 Countries might display the same level, but be characterised by very
different underlying distributions
Three channels of aggregate productivity growth:
i. Innovation at the frontier
ii. Diffusion from frontier to laggard firms
iii. Resource allocation
… each of these factors may call for different policy responses.

7. MultiProd – Micro drivers of aggregate productivity
7
 The MultiProd project is based on a ‘distributed microdata’ methodology
 Harmonised software sent to countries
 Researchers in each country will run the code on their confidential microdata
 Aggregated output respect confidentiality rules – followed CSO approach
 Cross country Micro-aggregated results then analysed by the OECD
 Comparable data analysed across countries
 Productivity measured in exactly the same way across countries
 Generates non-confidential aggregate statistics to allow for cross country
analysis

8.  Produces estimates of Labour and Multi-factor productivity (MFP)
 Solow method: 𝑴𝑴𝑴𝑴𝑴𝑴 = 𝑮𝑮𝑮𝑮 − 𝜷𝜷𝑲𝑲 𝑲𝑲 − 𝜷𝜷𝑳𝑳 𝑳𝑳 − 𝜷𝜷𝑰𝑰 𝑰𝑰𝑰𝑰
 Industry specific factor shares (cross-country median)
 Wooldridge method: Regression based approach (GMM)
 Corrects for bias in estimates
 Aggregation level
 Industry (Manufacturing, Utilities, Market and Non-Market Services)
 Sectoral level (2-digit NACE)
 Basic moments are computed (e.g. mean, median, standard deviation)
 Refined by percentiles of distribution (10th, 50th, 90th), age, size, ownership
 Various measures of the efficiency of resource allocation
 Measures strength of relationship between firm-size and productivity
8
MultiProd Model - Output

9. Data
9
 CIP (1991-2014), ASI (1999-2014)
 Year, Sector, Country of Ownership, Birth Year
 Investment, Value Added, Gross Output, Intermediate Inputs
 Employment, Wages
 Business Register (2006-2014):
 Weighting to make results representative of population
 Dealing with entry/exit of firms
 Changes in industry classification
 Deflators, K/L ratios and depreciation rates based on National Accounts
 Sector level (Nace Rev. 2)
 Panel sample (2006 – 2014)
 Manufacturing & Utilities: 2,500 firms (yearly average)
 Market & Non-Market Services: 7,500 firms (yearly average)

10. 10
The MultiProd Model – cross country results
90
95
100
105
110
2006 2007 2008 2009 2010 2011 2012
Manufacturing
p10 p50 p90
90
95
100
105
110
2006 2007 2008 2009 2010 2011 2012
Services
p10 p50 p90
• Cross country results based on 18 countries (excl IE)
• Evidence of widening gap between most and least productive firms
• Results based on Orbis data show a consistent pattern

11. MultiProd Results for Ireland (2006-2014)
11

12. 12
Granularity – the contribution of largest firms (1)
Manufacturing94%
88%
0% 20% 40% 60% 80% 100%
Share of VA by sales quantile
87%
73%
0% 20% 40% 60% 80% 100%
Share of Employment by sales
quantile
Manufacturing
Services
Irish results more concentrated than the cross-country MultiProd results
• Manufacturing 80% of VA and 68% of employment in cross-country
• Services 79% of VA and 66% of employment
Source: MultiProd on the basis of CSO
data

13. 13
Granularity – the contribution of most productive firms
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2006 2007 2008 2009 2010 2011 2012 2013 2014
Manufacturing
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2006 2007 2008 2009 2010 2011 2012 2013 2014
Services
• Most productive firms in manufacturing account for 70 percent of aggregate
productivity on average over 2006-2014
• 40 percent (on average) in services, although growing over the period
Source: MultiProd on the basis of CSO
data

14. 14
Labour productivity distribution – across sectors
Textiles & apparel
Transport equipment
Rubber & plastics
Furniture & other
Food & beverages
Metal products
Wood and paper prod.
Electrical equipment
Machinery and equipment
Computer & electronics
Chemicals
Pharmaceutical
Manufacturing sectors
Hotels and restaurants
Wholesale & retail
Administration services
Transportation & storage
Real estate activities
Marketing & other
Media
Telecommunications
IT
Legal & accounting
Scientific R&D
Services sectors
• Results broadly consistent with results of the MultiProd benchmark group (excl. scientific R&D)
Source: MultiProd on the basis of CSO data

15. 15
Foreign firm Labour productivity and employment premium
Source: MultiProd on the basis of CSO
data
11%
24%
50%
61%
61%
66%
70%
83%
114%
117%
123%
Productivity
premium: 399%
(5.7)
(5.6)
(15.1)
(5.3)
(7.3)
(3.7)
(4.6)
(34.2)
(3.2)
(3.1)
(20.4)
Average foreign firm employment multiple: (2.8)
0% 50% 100% 150% 200% 250% 300% 350% 400% 450%
Electrical equipment
Textiles & apparel
Computer & electronics
Machinery and equipment
Wood and paper prod.
Chemicals
Rubber and Plastic
Furniture & other
Food & beverages
Metal products
Transport equipment
Pharmaceutical
Manufacturing 2014

16. 16
Foreign firm Labour productivity and wage premium
Source: MultiProd on the basis of CSO
data
Food & beverages
Textiles & apparel
Wood and paper prod.
Chemicals
Pharmaceutical, 399%
Rubber & plastics
Metal products
Computer & electronics
Electrical equipment
Machinery and equipment
Transport equipment
Furniture & other Wholesale & retail
Transportation & storage
Hotels and restaurants
Media
Telecommunications
IT
Real estate activities
Legal & accounting
Scientific R&D
Marketing & other
Administration services
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
200%
0% 20% 40% 60% 80% 100% 120% 140% 160%
Foreignfirmwagepremium
Foreign firm labour productivity premium

17. 17
Productivity dispersion – labour productivity
Source: MultiProd on the basis of CSO data
40
50
60
70
80
90
100
110
120
2007 2008 2009 2010 2011 2012 2013 2014
Manufacturing
p10 p50 p90 Average
40
50
60
70
80
90
100
110
120
2007 2008 2009 2010 2011 2012 2013 2014
Services
P10 p50 p90 Average

18. 18
Productivity dispersion – by country
Country
2011
(Labour Productivity) p90-p10 ratio
Manufacturing Services
Australia 6.7 7.8
Austria 7.1 11.2
Belgium 5.0 5.7
Chile 20.1 34.1
Denmark 4.3 7.1
Finland 3.2 4.0
France 3.9 6.1
Hungary 16.3 26.8
Indonesia 22.4 -
Italy 5.3 7.5
Japan 3.5 4.0
Netherlands 7.4 19.7
New Zealand 6.3 8.1
Norway 5.6 8.8
Portugal 6.6 14.2
Sweden 4.3 6.4
OECD (MultiProd) 6.6 9.2
Ireland 7.7 9.3
Ireland (95-10) 9.8 14.1
Source: MultiProd on the basis of CSO data

19. 19
Efficiency of Resource Allocation – Olley Pakes Method
56% 48% 58%
55% 66% 63% 65%
51%
54%
-
40,000
80,000
120,000
160,000
200,000
2006 2007 2008 2009 2010 2011 2012 2013 2014
Manufacturing
4% 6% 11% 5% 21%
13%
-
40,000
80,000
120,000
160,000
200,000
2006 2007 2008 2009 2010 2011 2012 2013 2014
Services
• Aggregate productivity is the weighted average of firm productivity
• Can be decomposed into unweighted firm average, and the covariance between productivity and size
• The Covariance term is known as the Olley-Pakes (OP) gap and measures efficiency in the allocation of
resources
Source: MultiProd on the basis of CSO
data

20. 20
Efficiency of Resource Allocation – cross country results 2011
63%
34%
31%
34%
41%
39%
45%
25%
29%
29%
6%
23%
10%
9%
41%
67%
48%
52%
0 20,000 40,000 60,000 80,000 100,000 120,000 140,000 160,000 180,000 200,000
IRL
BEL
NOR
DNK
SWE
NLD
AUT
FRA
FIN
IRL*
AUS
JPN
LUX
CAN
ITA
HUN
PRT
CHL
due to the efficiency of resource allocation
in 2005 USD Purchasing Power Parity terms

21.  Aggregate productivity levels comparatively high, but growth rate
declining
 Skewed distributions
 Large firms dominate value add and employment
 Most productive firms dominate aggregate productivity
 Large foreign firm productivity premium
 Productivity dispersion (i.e. ‘the gap’) is widening
 Efficiency of resource allocation driven by foreign firms (in specific
sectors)
 Efficient allocation of resources among non-MNE firms important for living
standards
21
Conclusions

22. Department of Finance
Government Buildings
Upper Merrion Street
Dublin 2
Ireland
www.finance.gov.ie
@IRLDeptFinance
This presentation is for informational purposes only.
No person should place reliance on the accuracy of the data and should not act solely on the basis of the presentation itself.
The Department of Finance does not guarantee the accuracy or completeness of information which is contained in this document and which is stated to have been obtained from or is
based upon trade and statistical services or other third party sources. Any data on past performance contained herein is no indication as to future performance.
No representation is made as to the reasonableness of the assumptions made within or the accuracy or completeness of any modelling, scenario analysis or back-testing.
All opinions and estimates are given as of the date hereof and are subject to change.
The information in this document is not intended to predict actual results and no assurances are given with respect thereto.