The Role of Production Factor Quality and Technology Diffusion in 20th Century Productivity Growth
1. The role of
production factor quality
and technology diffusion
in 20th century productivity growth
Antonin Bergeaud, Gilbert Cette, Remy Lecat
Banque de France
Global Forum on Productivity – 7 July 2016
Breakout Session A: Productivity spillovers, diffusion and
public policies
The views presented here do not necessarily reflect the ones of the Banque de France or
the Eurosystem
2. Taking a long-term macro view on technology diffusion
20th century: period of exceptional, although heterogeneous, growth
Slowdown in advanced economies since the 1970s:
Brief, limited in size and localized revival due the Information and Communication
Technologies
Current questioning on the future of innovation
Need to take a long-term view on growth drivers to address this questioning
Total Factor Productivity (TFP) explains the lion share of GDP growth since 1890
But TFP = a residual, encompassing measurement errors, in particular of factor
quality
Is it possible to identify some of the main TFP growth drivers, and hence, the
reason for TFP growth decline since 2005?
1st: try to purge TFP from the quality of production factors (labour and capital)
2nd: try to identify the role of technology diffusion through 2 emblematic
technologies, electricity and ICT
2
3. What do we do?
Analysis of TFP growth over the period 1890-2013 in 17
advanced economies:
G7 countries + Spain, Netherlands, Belgium, Finland,
Portugal, Australia, Sweden, Norway, Switzerland and
Denmark
We reconstituted a Euro Area using 8 countries: DEU,
FRA, ITA, ESP, NDL, BEL, FIN, PRT (93% of the 2010 EA
GDP)
Growth-accounting framework with Permanent
Inventory Method for capital harmonized for all countries.
Data and methodology available at:
www.longtermproductivity.com
3
4. 1st Step: Usual measurement of TFP with capital and labor
as inputs.
2nd Step: Measurement of TFP’: TFP purged from the
impact of
human capital
age of capital
3rd Step: Measurement of TFP’’: TFP’ minus the impact of
2nd industrial revolution GPT innovation (electricity)
3rd industrial revolution GPT innovation (ICT)
4
What do we do?
6. Decomposition of GDP per capita growth for 17 countries and the Euro Area
Percentage points
20th century growth:
GDP per capita growth relying on TFP and capital intensity
6
7. United States: The « One Big Wave » and the ICT shock
-1
0
1
2
3
4
5
6
Total factor productivity - US
(Whole economy, smoothed annual growth rate with HP filter, λ = 500, in %)
7Source: Bergeaud, Cette and Lecat (2016)
8. Other advanced economies: a delayed « One Big Wave »
-1
0
1
2
3
4
5
6
United States Euro Area United Kingdom Japan 8
Total factor productivity
(Whole economy, smoothed annual growth rate with HP filter, λ = 500, in %)
Source: Bergeaud, Cette and Lecat (2016)
10. Labour:
In the previous graph: number of hours worked, but educated
workers have a higher productivity
Adding the average years of education in the working-age
population.
IV estimates: +1 year of study → +5% labour productivity
Consistent with macro studies (cf. Barro and Lee, 2010)
Capital:
In the previous graph: capital stock, but innovation embodied in
new equipments
Adding the age of capital in a permanent inventory method
framework
IV estimates: 1 year older equipment → -1% labour productivity
Consistent with Wolf (1991, 1996)
10
How does factor quality impact TFP?
11. Average age of capital equipment
(In years, Whole economy)
11
3
4
5
6
7
8
9
10
United States Euro Area United Kingdom Japan
Significant but cyclical contribution of the equipment age
Source: Bergeaud, Cette and Lecat, 2016
12. 0
2
4
6
8
10
12
14
United States Euro Area United Kingdom Japan
Average number years of studies of the working age population
(In years, Whole economy)
12
Steady but heterogeneous increase in years of education
Source: Barro and Lee (2010) and van Leeuwen and van Leeuwen-Li (2014)
13. -0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
Total factor productivity - United States
(Whole economy, smoothed annual growth rate -HP filter, λ = 500- , in %)
13
US Total factor productivity - uncorrected
Source: Bergeaud, Cette and Lecat (2016)
14. -0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
TFP TFP' 14
US TFP corrected for education and capital age
Total factor productivity - United States
(Whole economy, smoothed annual growth rate -HP filter, λ = 500- , in %)
Source: Bergeaud, Cette and Lecat, 2016
16. Electricity production per capita
(Log of kWh per thousand inhabitants ,whole economy)
0
2
4
6
8
10
12
14
16
United States Euro Area United Kingdom Japan 16
Two emblematic technologies: 1/ Electricity
Source: Comin and Hobijn, 2010
17. ICT share in capital equipment
(Current value, in %)
0
2
4
6
8
10
12
United states Euro area United kingdom Japan
17
Two emblematic technologies: 2/ ICTs
Source: Cette, Clerc and Bresson, 2015
18. Diffusion as well as production of new technologies contributes
to labour productivity growth through capital intensity and TFP
Capital intensity already (although imperfectly) taken into account
Concerning technology shocks, 3 channels of TFP impact:
Direct production (TFP gains in ICT producing industries)
→ Taken into account for electricity (no possible importations),
→ not really for ICT (possible importation)
Mismeasurement of price declines of capital goods =>
Mismeasurement of capital-deepening and consequently of
TFP
→ Relevant for ICT
Externalities (“manna from heaven”)
→ Indirectly taken into account both for electricity and ICT
18
How does new technology impact TFP?
19. -0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
TFP TFP' TFP''
19
US: technologies insufficient to fully
explain the « One big wave »
Total factor productivity - United States
(Whole economy, smoothed annual growth rate -HP filter, λ = 500- , in %)
Source: Bergeaud, Cette and Lecat, 2016
20. Takeaways: the need for a broader view on
growth and productivity
TFP growth waves are the main drivers of GDP growth
TFP growth waves and TFP growth recent decline only partly
explained by factor quality adjustment and by technology
impact measurement
Major role of other channels :
improvement in the production process: assembly lines (Ford Model T in
1913);
enhanced management practices (Bloom et al., 2014);
new financing techniques (Ferguson and Wascher, 2004)…
Interactions between these channels
Hence, whatever the future course of innovation, large convergence
margins remain for most countries. 20
22. The model
1st Step:
𝑌 = 𝑇𝐹𝑃. 𝐾α. 𝐿1−α
With:
𝑌 ∶ 𝐺𝐷𝑃
𝐾: Capital stock
𝐾 = 𝐾𝐸 + 𝐾𝐵
𝐾𝐸: Equipment
𝐾𝐵: Buildings
𝐿: Number of hours worked
𝐿 = N.H
N: Number of workers
H: Average number of working hours per worker
α = O.3
Same for all countries and constant
Results robust to other values of α
22
23. The model
2nd Step:
𝑌 = 𝑇𝐹𝑃′. 𝐾. 𝑒−𝜀.𝐴 𝛼∙ 𝐿. 𝑒 𝜃.𝑆 1−𝛼
Then:
𝑇𝐹𝑃 = 𝑇𝐹𝑃′. 𝑒−𝛼.𝜀.𝐴. 𝑒 1−𝛼 .𝜃.𝑆
With:
𝐴 ∶ Average age of equipments
S: Average number of schooling years of population of more than 25 years
𝜀, 𝜃 > 0 and assumed to be the same for all countries and constant
Increase in A -> decrease of TFP for the same TFP’
Increase in S -> increase of TFP for the same TFP’
23
24. The model
3rd Step:
𝑇𝐹𝑃′
= 𝑇𝐹𝑃′′
. 𝑬𝒍𝒆𝒄𝒕 𝜼
. 𝐞 𝝁. 𝑲𝑺𝑰𝑪𝑻
With:
Elect: electricity production in Kwh per capita
KSICT: Share of ICT capital in capital equipment, at current prices
𝜂, 𝜇 > 0 and assumed to be the same for all countries and constant
Increase in Elect -> increase of TFP’ and TFP for the same TFP’’
Increase in KSICT -> increase of TFP’ and TFP for the same TFP’’
24
30. Literature review
Abundant related literature
Looking at the main factors of growth over the long-run and their impact has
been a widely studied subject.
Some examples (among numerous, see the paper):
On all aspects, Aghion and Howitt (1998); numerous following
publications
Education: Madsen (2014) finds that education explains 0.48 points of the
1.87% average growth in gdp per capita. A large related literature has
focused on estimating the return of education on growth over the long run
(Barro and Lee, 1992, 2010; Morisson and Murtin, 2010; Cohen and Soto,
2007; …)
Institutions: Prados (2015) build a new database on economic freedom
and institution over the long run
Technology: Comin and Mestieri (2013) and Comin and Hobijn (2010)
stressed the importance of technology adoption to explain divergence
between economies. Madsen (2010) looked at the “Anatomy of growth” in
the 20th century to highlight the importance of technology spillover using
patent data
… 30
Hinweis der Redaktion
Very fast GDP per capita growth, lowest in countries with high initial GDP per capita level such as Australia, Belgium or the UK; very strong in Japan, with a very low initial level.
Over the whole period, growth relying on TFP and capital intensity; negative contribution of hours worked, also for all sub-periods
Disaprate contribution of employment rate