Productivity Revisited


BOX 1.1 Are the Current Productivity Lags Just the Calm before the Next


Download 4.27 Mb.
Pdf просмотр
bet4/20
Sana15.12.2019
Hajmi4.27 Mb.
1   2   3   4   5   6   7   8   9   ...   20

BOX 1.1
Are the Current Productivity Lags Just the Calm before the Next 
Productivity Storm?
There is considerable evidence that information technologies (IT) were a key force behind produc-
tivity acceleration episodes in the late twentieth century (see, for example, Jorgenson, Ho, and 
Stiroh 2008), but these growth spurts appeared with a long lag, after a 20-year period (roughly 
1974–94) of brisk investment in IT. The absence of an impact of IT on productivity during this 
period prompted Robert Solow to make his famous observation, “You can see the computer age 
everywhere but in the productivity statistics” (Solow 1987, 36). Recent works like Fernald 2014 
and Byrne, Oliner, and Sichel 2013 have presented evidence that these IT-based gains have slowed 
over the past decade. But perhaps this represents a respite between periods of fast productivity 
growth. This view implies that while highly effective new technologies are being created, there 
are implementation and adjustment costs that must be paid before they can be effectively 
deployed to obtain noticeable improvements in productivity (see, for example, van Ark 2016). 
History may offer some guidance on the issue of productivity gains from general-purpose 
technologies like IT. Syverson (2013) shows that the productivity growth from “portable power” 
technologies like electrification and the internal combustion engine arrived in two waves sepa-
rated by a decade-long slowdown. While this prior diffusion hardly implies that a second IT wave 
is imminent, it does show that productivity accelerations from general-purpose technologies do 
not have to be one-off events. Just because their resultant productivity growth sped up in the late 
1990s and early 2000s does not mean it cannot speed up again.

The Elusive Promise of Productivity 
9
FIGURE 1.4
  The Number of Global Researchers Has Doubled since 1995, with Most 
Growth in the Developing World
Source: Elaborations using UNESCO data.
1995
0
2
4
6
8
Number of researchers (millions)
2000
2005
2010
2015
World
Developing countries
Advanced economies
Source: Branstetter, Li, and Veloso 2013.
FIGURE 1.5
  Most of the U.S. Patent and Trademark Office Patents in China and India 
Have Been Co-invented and Sponsored by Multinational Firms 
Purely Chinese (panel a)/Indian (panel b) invented and assigned to other types of organizations
Purely Chinese (panel a)/Indian (panel b) invented and assigned to indigenous firms
Purely Chinese (panel a)/Indian (panel b) invented and assigned to multinational firms
Co-invented and assigned to multinational firms
Co-invented and assigned to indigenous firms
Co-invented and assigned to other types of organizations
3,000
2,000
Number of patents
1,000
0
1980
1990
2000
Grant year
a. China
2010
b. India
Grant year
1970
1990
1980
2000
Number of patents
2010
1,500
1,000
500
0

10 
Productivity Revisited
advanced economies may provide more sophisticated, creative, and high-level intel-
lectual input. Hence, while new discoveries may be taking more and more research 
effort, the size and global productivity of that effort is increasing.
Are Measurement Issues Obscuring True Productivity Growth?
Some analysts have suggested that the deceleration in productivity is substantially illu-
sory and represents a problem with measurement.
9
 This “mismeasurement hypothesis” 
argues that products emerged in the mid-2000s, such as Google search, that are highly 
valued, but whose contributions to economic activity are difficult to capture in stan-
dard economic statistics because they are consumed at a zero price—or at least at a very 
low price relative to their social value. Hence the gains in true productivity are not 
reflected in the prevailing economic output statistics. 
However, this hypothesis starts to falter when confronted by its implications for 
what can be measured in the data. Multiple recent systematic analyses using varied 
approaches and data have found that the slowdown is not primarily a mismeasure-
ment phenomenon (see Cardarelli and Lusinyan 2015; Byrne, Fernald, and Reinsdorf 
2016; Nakamura and Soloveichik 2015; and Syverson 2016). As one example, for the 
OECD, figure 1.6 plots the relationship between the size of a country’s productivity 
slowdown (on the vertical axis) and two measures of the importance of IT products 
in that country’s economy: on the demand side, the fraction of the country’s house-
holds that have a broadband connection; and on the supply side, the share of 
value added accounted for by ICT-producing industries. Figure 1.6 reveals no obvious 
relationship to the eye, and a regression analysis confirms this.
10
 
Brynjolfsson, Rock, and Syverson (2017) alternatively postulate that the delays in 
building the necessary intangible complements to innovations such as AI—including 
R&D, patents, trademarks and copyrights, or organizational or entrepreneurial capital—
can distort the measurement of TFP because these complements are not well counted in 
the imperfect notion of gross domestic product (GDP). The early investments in labor 
and capital for these complements, which are not yet counted in GDP and whose output 
lies in the future, will appear as a decline in TFP. Brynjolfsson and his coauthors cite 
Brookings Institution research that investments in autonomous vehicles exceeded $80 
billion from 2014 to 2017, with little consumer adoption of these technologies yet. This 
amounts to 0.44 percent of 2016 GDP (spread over three years). Adding in equally costly 
labor inputs would lower estimated labor productivity by 0.1 percent per year over the 
last three years.
11
Another form of mismeasurement may arise from not taking important changes in 
industrial organization into account in the estimation of productivity. De Loecker and 
Eeckhout (2018) argue that the increase in industrial concentration of the U.S. econ-
omy (and in Europe as well) has increased markups and led to a corresponding fall in 

The Elusive Promise of Productivity 
11
FIGURE 1.6
  There Is No Obvious Relationship between the Productivity Slowdown 
and the Prominence of Information Technology
AUS
BEL
CAN
CHE
CZE
DEU
DNK
ESP
FIN
FRA
GBR
HUN
IRL
ISL
ITA
JPN
KOR
LUX
NLD
NOR
POL
PRT
SWE
TUR
USA
–3
–2
–1
0
1
Percentage change in labor productivity growth 
20
40
60
80
100
Percentage of households with broadband
a. Productivity slowdown and household broadband usage
–3
–2
–1
0
1
Percentage change in labor productivity growth
4
6
8
10
12
ICT percentage of value added
b. Productivity slowdown and share of value added from ICT production
BEL
CAN
CHE
CZE
DEU
DNK
ESP
FIN
FRA
GBR
GRC
HUN
IRL
ISL
ITA
JPN
KOR
LUX
NLD
NOR
POL
PRT
SWE
USA
Source: Syverson 2016. 
Note: This figure plots the size of the measured productivity slowdown from 1995–2004 to 2005–14 in a country versus the share of the 
country’s households with a broadband connection (panel a) or the value-added share of information and communication technologies 
(ICT) producers in the country (panel b). All data are from the Organisation for Economic Co-operation and Development. See chapter 
text and Syverson 2016 for details.
the share of income going to labor and capital. Taking these into effect, they find that 
productivity growth has increased since 1980 and has hovered around 3  percent–4  percent 
since the 2007 crisis. This, however, cannot explain why labor productivity, defined as 
output divided by number of workers, should also fall. Furthermore, De Loecker and 
Eeckhout (2018) show that Latin American markups have remained high, but have not 
increased over the last decades. Estimates from six representative countries do not sug-
gest an increase in industrial concentration (figure 1.7) (see also Díez, Leigh, and 
Tambunlertchai 2018). 

12 
Productivity Revisited
Hence, on balance, there are some reasons to think that production data are under-
measuring at least the current and future impact of new technologies. In addition, 
there is something incongruous about the simultaneous concerns, on the one hand, 
that productivity growth is a thing of the past and on the other, that rapid progress in 
robotics and AI will displace masses of workers through productivity gains. Across 
the centuries, disruptive technologies have eliminated certain types of jobs (manual 
weavers being the iconic example) while creating entirely new professions (computer 
 programmers, heart surgeons, automobile assemblers) that pay substantially better. 
The labor data do suggest that important employment effects of automation and 
 
technology-facilitated outsourcing are likely responsible for the polarization of 
advanced-economy labor markets.
12
However, evidence of this effect for developing countries is still scattered and weak, 
perhaps partly because technology has not yet arrived in all these countries or because 
the intensity of use of these technologies has diverged (Comin and Mestieri 2018). 
Figure 1.8, using census data, confirms that for advanced economies, employment for 
workers engaged as machinery operators or in crafts has declined or stagnated com-
pared with higher-end professional employment or lower-skilled clerks, service work-
ers, or elementary occupations. Figure 1.9 suggests that, in fact, in advanced economies, 
more robots are associated with fewer jobs for manufacturing operators, although not 
necessarily more total jobs. 
FIGURE 1.7
  Industrial Concentration Has Not Increased in a Sample of 
Emerging Markets
0.2
0.4
0.6
0.8
1.0
Top four share of industry sales
1997
2000
2003
2006
2009
2012
Chile
Ethiopia
China
Indonesia
Côte d’Ivoire
Morocco
Source: Authors’ elaboration using census data. 
Note: The figure plots the average fraction of total manufacturing industry sales that is accounted for by the largest four firms within 
that industry in selected developing countries. Concentration ratios are calculated for each four-digit industry and then averaged 
across manufacturing using industry weights.

The Elusive Promise of Productivity 
13
The Weakness of Economic Convergence
The second area of preoccupation prompting new work on productivity is the con-
tinued failure of economic convergence over the long term. Even if advances were to 
come to a halt in the advanced economies, productivity in the follower nations would 
lag the frontier nations, offering the prospect of massive gains from technological 
catch-up. 
The enduring question in development economics is why this catch-up has not 
been happening. The average GDP per capita of the richest 10 percent of countries in 
2000 was 40 times higher than that of the poorest 10 percent of countries—meaning 
that the average person in an advanced economy produces in just over nine days what 
the average person in a follower country produces in an entire year, Restuccia (2013) 
finds (see also Caselli 2005).
 
Numerous studies have documented that roughly half of 
this difference in income cannot be explained by differences in capital or other tan-
gible factors of production and hence is attributed to differences in the efficiency 
with which they are combined—that is, to TFP (Klenow and Rodríguez Clare 1997; 
FIGURE 1.8
  Labor Markets Are Becoming More Polarized in Advanced Economies, 
but Not in Developing Countries
Source: Maloney and Molina 2016. Calculations are based on IPUMS (Integrated Public Use Microdata Series) data.
Note: The figure plots the percent change in employment before and after 2000. Data span 1979−2012. Horizontal I-bars show 
 confidence  intervals. 
Legislators and managers
Professionals
Technicians
Operators and assemblers
Crafts and related
Clerks
Skilled agricultural and fishery
Service workers
Elementary occupations
−1.0
0
1.0
2.0
−1.0
0
1.0
2.0
a. Advanced economies
b. Developing countries
Percent change in employment by occupation

14 
Productivity Revisited
FIGURE 1.9
  Are Robots Displacing or Creating Manufacturing Jobs?
0
2
4
6
8
10
Operators and assemblers per 100 inhabitants
20
40
60
80
Robots per 100,000 inhabitants
Advanced economies
Less-developed countries
Source: Maloney and Molina 2016. Calculations are based on IPUMS (Integrated Public Use Microdata Series) data.
Easterly and Levine 2001). The potential contribution to raising global productivity 
and reducing poverty of achieving convergence are immense. 
Yet despite early arguments for a natural process of convergence among the now-
frontier countries (Baumol 1986), it has subsequently proven statistically elusive (see, 
for example, De Long 1988). Furthermore, Pritchett (1997), among others, documents 
a “Great Divergence” of the last two centuries where, instead of follower countries 
catching up, advanced economies, with few exceptions, continue to pull ahead. The 
emergence of different “convergence clubs” (Quah 1996; Maasoumi, Racine, and 
Stengos 2007), in which follower countries converge to clumps of similar levels of 
income far from the frontier, has been documented to be largely a matter of differences 
in productivity growth. Convergence seems to be weakening even in the regions where 
it was assumed. A recent World Bank study, “Growing United (Ridao-Cano and 
Bodewig 2018), shows that the productivity gap between southern and northern mem-
ber states of the European Union has been widening since the early 2000s. Convergence 
is elusive at the subnational level as well.
13
 What is clear is that reducing the between-
country differences in productivity would contribute massively to global productivity 
growth. Yet convergence does not appear to be an inevitable natural force, like gravity.
This fact underlies what the previous volume in this series (Ciera and Maloney 
2017) called the innovation paradox, focusing on the puzzle of why rates of 

The Elusive Promise of Productivity 
15
technology adoption are low in developing countries, and how that inhibits con-
vergenceThe gains from adopting existing products, processes, and management 
techniques from abroad are thought to be large. Indeed, the radiation of ideas, 
products, and technologies to developing countries represents an externality of 
truly historic proportions. In fact, Comin and Hobijn (2004, 2010) and Comin and 
Ferrer (2013) argue that it is precisely the differences in the rate and intensity of 
adoption of new technologies that drives the magnitude of the Great Divergence.
14
 
Comin and Mestieri (2018) argue that a reduction in the average adoption lag by 
one year is associated with a 3.8 percent higher per capita income. Cutting the 
adoption lag faced by a country from 50 years longer than the United States to the 
U.S. level is associated with an increase in per capita income by a factor of seven!
15
 
Recent estimates of the returns to one type of innovation investment, R&D, for the 
United States and Spain put them at a striking 40 percent to 60  percent annually.
16 
Griffith, Redding, and Van Reenen (2004) and Goñi and Maloney (2017) show that 
returns rise much higher (potentially to the triple digits) with increased distance 
from the technological frontier for a while, reflecting the gains from Schumpeterian 
catch-up afforded to follower countries.
 
Yet countries do not seem to exploit these 
potential gains. 
Ironically, the lack of polarization observed in panel b of figure 1.8 for developing 
countries may be due partly to this low rate of technological adoption. The underly-
ing data suggest perhaps incipient effects observed in more advanced economies like 
Mexico and Brazil, but the majority of countries in the sample show nothing. In fact, 
the exact reverse effect is observed in Vietnam and China, which may be because, as 
figure 1.9 suggests, in developing countries, robot density is associated with higher 
employment of operators and assemblers, in contrast to advanced economies. This 
may arise in cases where large-scale offshoring also involves the introduction of auto-
mation. As Maloney and Molina (2016) discuss, there are many reasons why automa-
tion and robots would be adopted more slowly in the majority of developing 
countries, ranging from the country’s technological absorptive capacity to the skill of 
the workforce, its ability to mobilize resources for large capital investments, the 
capacity for maintenance, and attention to tolerances. As these problems are 
redressed, today’s advanced-economy problems may, in fact, become those of the 
developing world tomorrow.
The Mechanisms of Productivity Growth: Second-Wave Analysis
In sum, there is no firm consensus on the first puzzle of the global productivity slow-
down and, in fact, the causes may differ between advanced economies and develop-
ing countries. There appears not to be a pronounced fall in dynamism or increase in 
industrial concentration or shift toward lower-productivity services (see box 1.2) in 
developing countries that have been forwarded as explanations, for instance, 

16 
Productivity Revisited
BOX 1.2
Structural Transformation Decompositions
Following in the spirit of economists from Kuznets to Chenery who have detailed the movement of 
labor from agriculture to manufacturing, McMillan and Rodrik (2011) and McMillan, Rodrik, and 
Verduzco-Gallo (2014) decompose labor productivity growth into two components: one that holds 
labor shares in different sectors constant but allows changes in average labor productivity, and 
another that holds sectoral productivity constant and allows for observed reallocation of labor. 
They find the latter component plays an important role in many high-growth countries. 
Rogerson (2017), for this volume, examines data for Asia and broadly confirms their findings. 
Structural transformation—effectively the analog to the “between” dimension discussed earlier—
does in some cases account for half of productivity growth (China, Thailand), although in many 
cases it accounts for less than 10 percent. Little structural transformation has occurred in high-
growth Malaysia, and in both India and Indonesia structural transformation has been far less 
extensive than in China or Thailand (figure B1.2.1). 
A second important takeaway is that the importance of reallocation diminishes with the level 
of income, accounting for very little in more advanced economies. In the United States, structural 
transformation never accounts for more than 0.1 percent of growth, while in Japan; Taiwan, China; 
and the Republic of Korea, the contribution is low or negative, despite having been important during 
their miracle periods. Baumol’s cost disease, in which progressively more spending goes into sec-
tors such as services where productivity is lower and slower, may partly explain these rates.
As the following chapters demonstrate, the interpretation of these patterns is not clear for 
policy. They are neither obviously capturing movements from high- to low-productivity sectors nor 
illustrating the underlying drivers of such movements.
FIGURE B1.2.1
   The Percentage of Productivity Growth Contributed by 
Structural Transformation Varies Widely by Country and 
over Time 
Source: Rogerson 2017.
−50
0
50
100
150
Percentage of productivity growth
contributed by structural change
China
India
Indonesia
Japan
Korea, Rep.
Malaysia
Philippines
Taiwan, China
Thailand
United States
1950s
1960s
1970s
1980s
1990s
2000s

The Elusive Promise of Productivity 
17
in the United States. The second puzzle remains a long-standing analytical challenge 
that goes to the core of the World Bank’s mandate. 
This said, while the literature has not offered a definitive explanation for why the 
productivity engine has not regained its previous force or why followers are so slow in 
catching up to the leaders, it has, over the last 20 years, dramatically increased our 
understanding of the functioning of the underlying dynamics and mechanisms. More 
profoundly, it has revolutionized the conceptual and analytical tools for analyzing pro-
ductivity and its determinants. 
Productivity growth—both of the countries pushing the frontier and of those 
unevenly catching up—can be broken down mechanically into three components, as 
shown in figure 1.10.
Improved firm performance (within firm). At the center of productivity analysis is the 
firm. The within component is related to individual firms becoming more productive: 
that is, increasing the amount of output they produce with a constant amount of inputs 
(such as labor, capital, land, raw materials, and other intermediate inputs) because they 
have increased their internal capabilities, including managerial skills, workforce skills, 
innovation capacity, and technology-absorption capability. 
Improved allocation of factors of production across firms (between-firm). Ideally, the 
most productive firms would attract the most resources, thereby ensuring the greatest 
possible output. However, myriad distortions—including poorly designed legislation 
or political patronage that prevents resources from moving from less efficient firms—
can have large effects. The between-component is associated with the reallocation of 
factors of production and economic activity toward more efficient firms.
Improved entry and exit of firms (selection). Aggregate productivity growth can also 
be explained by the entrance of high-productivity firms (relative to the industry aver-
age) and the exit of low-productivity firms (again, relative to the industry average). 
Examining the factors that affect the entry of higher-quality firms moves into the study 
of entrepreneurship. Understanding the disincentives and barriers to exit involves 
issues of business climate and potentially social norms.


Do'stlaringiz bilan baham:
1   2   3   4   5   6   7   8   9   ...   20


Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2019
ma'muriyatiga murojaat qiling