Productivity and income differences in the 20 century

Productivity and income differences in the 20th century

Robert Inklaar and Daniel Gallardo Albarrán (University of Groningen)World KLEMS Conference, June 4–5 2018

Development accounting

• What can account for international income differences?�Maddison’s proximate causes of growth

• Hsieh and Klenow (2010):!"#"= %"

&&'( )"

!"

(&'( ℎ"+"

#"GDP per capita Capital/output Human capital/capita

Income differences are predominantly traceable to productivity differences

-4-2

02

α⁄ (1

−α) x

log

of re

lativ

e ca

pita

l/out

put r

atio

-4 -3 -2 -1 0 1log of CGDPo/capita (USA=1)

Physical capital

-4-2

02

log

of re

lativ

e la

bor i

nput

per

cap

ita


Human capital

-4-2

02

1⁄ (1

−α) x

log

of re

lativ

e TF

P


Productivity

Source: PWT 9.0, Feenstra et al. (2015)Notes: Development accounting across 115 countries

Development accounting for 2011

But has that been a constant feature?

• 20th century has mostly been one of income divergence• Asian ‘perspiration vs. inspiration’: has rapid physical capital

accumulation given the expected income pay-off?• Diminishing development returns to education:

1950 2010India Spain India Spain

Av. years of schooling 0.9 4.9 5.4 9.3GDP per capita (US=1) 5.8% 24.2% 8.9% 65%

Context in the literature

• Comparing productivity growth over the long run of considerable interest� Bergeaud, Cette and Mojon (2016), Allen (2012), Madsen (2010)

• Yet comparing relative productivity levels requires a differentperspective�Measuring growth differs from measuring comparative levels

(Feenstra, Inklaar and Timmer, 2015; Inklaar and Diewert, 2016; Inklaar and Rao, 2017)

Ingredient #1:Maddison Project Database 2018

• Documented in Bolt, Inklaar, de Jong and van Zanden (2018)�Distinguish a ‘growth’ data series and a ‘levels’ data series

• Rely on multiple cross-country income comparisons� Post-1950: PPPs from International Comparison Program; see also PWT� Pre-1950: Historical income comparisons, work of many economic

historians, e.g. Ward and Devereux, van Zanden & co-authors

Methodological choice with important consequences

Extrapolation-based measure

New 'cross-country' measure

0.5

11.

5U

nite

d St

ates

=1

Brazil China India France NetherlandsArgentina UKSource: Maddison Project Database 2018, Bolt et al. (2018)

Real GDP per capita in 1900, US=1

Using the growth series toextrapolate from a modern income comparison can lead to notably different relative levels than using the most recent historical income comparison

Ingredient #2:New estimates of physical capital• Builds on recent work

� Bergeaud, Cette and Mojon (2016), Allen (2012), Madsen (2010)• Broader range of countries, at different levels of development

� 38 countries in Europe, North America, Latin America and Asia• Account for changes in relative prices of equipment and

structures• Implement a novel initialization method

�Country-specific but without relying on steady-state assumption• Note: estimates of total stock, not capital services

The change in physical capital

ARG

AUS

AUT

BEL

BRA

CANCHECHL

COL

DEU

DNK

ECU

ESPFIN

FRA

GBRIDN

IND

ITA

JPN

NLD

NORNZL

PER

SGP

SWE

URY

USA

VEN

ARG

AUSAUT

BEL

BOL

BRA

CANCHE

CHL

COL

CRI

DEUDNK

ECU

ESPFIN

FRA

GBRHND

IDN

IND

ITAJPN

KOR

MEX

NIC NLDNOR

NZL

PER

PRT

SGP

SLV

SWE

TWNURY

USA

VEN

1900

2011

01

23

4Ca

pita

l/out

put r

atio

6 7 8 9 10 11log of GDP per capita

Source: GDP per capita from Maddison Project Database 2018 (Bolt et al. 2018)

Capital/output ratio and income levels: 1900 & 2011

Considerable variation in 1900 capital/output ratios:• Far away from a steady state• Physical capital seems more

important for development accounting in 1900 than 2011

The change in human capital

Average years of schooling also showed more variation in 1900 than in 2011

Development accounting: 2011-2

-1.5

-1-.5

0.5

α⁄ (1

−α) x

log

of re

lativ

e ca

pita

l/out

put r

atio

-3 -2 -1 0 1log of CGDPo/capita (USA=1)

Physical capital

-2-1

.5-1

-.50

.5lo

g of

rela

tive

labo

r inp

ut p

er c

apita


Human capital

-2-1

.5-1

-.50

.51⁄

(1−α

) x lo

g of

rela

tive

TFP


Productivity

Notes: Development accounting across 38 countries


Development accounting: 1900-1

-.50

.51

α⁄ (1

−α) x

log

of re

lativ

e ca

pita

l/out

put r

atio

-2.5 -2 -1.5 -1 -.5 0log of CGDPo/capita (USA=1)

Physical capital

-1-.5

0.5

1lo

g of

rela

tive

labo

r inp

ut p

er c

apita


Human capital

-1-.5

0.5

11⁄

(1−α

) x lo

g of

rela

tive

TFP


Productivity

Notes: Development accounting across 23 countries


The rise of TFP

Productivity

Human capital

Fixed capital0.2

.4.6

.8C

oeffi

cien

t on

GD

P pe

r cap

ita

1900 1920 1940 1960 1980 2000 2020

Discussion

• Development is hard: putting in more physical and humancapital will not automatically lead to higher productivity• Increasing mismeasurement of capital?

�Capital services vs. stocks�Omission of intangible assets� Ignoring dimensions of human capital unrelated to schooling

Productivity and income differences in the 20 century

Documents

Transcript of Productivity and income differences in the 20 century