Popular in Economics

Central Monetary Authorities

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Central Monetary Authorities

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Economic History

Golden Fetters

UC Berkeley Barry Eichengreen

Economic Inequalities

Visions of Inequality

City University of New York Branko Milanovic

Economics

School Finance Equalization Increases Intergenerational Mobility

Yale University Barbara Biasi

Economics

Olivier Blanchard on U.S. Pandemic-Era Inflation

Peterson Institute for International Economics Olivier Blanchard

Deindustrialization and Industry Polarization

View more insights in Social Sciences - Economics - Development Economics

View more figures from Southern Methodist University

View the page of Michael Sposi

Read the Study

Next in Development Economics

Linkage between Income and Government Expenditure at Indian Sub-nationals

Dinesh Kumar Nayak | National Institute of Public Finance and Policy

Labor supply responses to rainfall shocks

Pushkar Maitra | Monash University

Michael Sposi suggests deindustrialization and increasing industry polarization are driven by sector-biased productivity growth and sectoral trade integration.

Image courtesy of interviewee. November 5, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the concept of "industry polarization," which refers to the increasing dispersion of manufacturing shares in countries' economic activity over time. The research highlights how, as countries develop, they typically shift economic activity from agriculture to manufacturing and then to services. However, more recent observations indicate that countries today tend to achieve lower peaks in manufacturing employment shares compared to previous decades, a phenomenon referred to as "premature deindustrialization," initially noted by economist Dani Rodrik.

The authors explore the changing landscape of economic structure through a quantitative approach, examining data across various countries and comparing trends over several decades. They identify key factors contributing to this phenomenon, including technological advancements leading to increased productivity in manufacturing but lesser gains in the service sector, and the declining relative prices of manufacturing outputs.

The analysis demonstrates that while technological change accounts for a significant portion of deindustrialization, the increasing concentration of manufacturing production in fewer countries is primarily driven by declining trade costs. The paper emphasizes that trade and technology interact, with trade functioning as a mechanism that amplifies the effects of technological change, contributing to overall structural shifts in economies.

The findings suggest important implications for industrial policy, particularly for lower-income countries entering the global market later. This video concludes by indicating that future research could investigate how trade dynamics influence the development and adoption of new technologies, which could further inform our understanding of structural changes in the global economy.

Language:

Export as PDF

Export as PDF
Here are five key points extracted from the video regarding industry polarization and deindustrialization:

1. **Definition and Concept of Industry Polarization**: Industry polarization refers to the increasing dispersion in the sectoral composition of economic activity among countries, particularly concerning the share of manufacturing in their economies. This phenomenon indicates that some countries now have a more varied (or polarized) mix of economic activities compared to the past.

2. **Structural Change Theory**: The literature on structural change analyzes how the mix of economic activity shifts among the primary (agriculture), secondary (manufacturing), and tertiary (services) sectors as countries develop. Initially, industrialization sees a shift from agriculture to manufacturing, but in more advanced economies, there is now a transition from manufacturing to services (de-industrialization).

3. **Premature Deindustrialization**: Observations indicate that countries today tend to reach lower peaks in manufacturing share before transitioning into deindustrialization compared to countries in the past. This trend, identified by economist Dani Rodrik, raises questions about the efficiency and underlying causes of this premature deindustrialization.

4. **Role of International Trade and Technology**: The video discusses how international trade and technological advancements contribute to changes in economic structures. Technological growth in manufacturing often outpaces that in services, which affects relative prices and consumer spending patterns. The research also highlights how international trade amplifies the effects of technological change on manufacturing shares and overall economic activity.

5. **Policy Implications**: The findings suggest that countries with lower income levels should reconsider their industrial policies. Simply aiming to replicate the industrial growth seen in countries like South Korea may not be feasible due to changes in the global economic landscape that exert less pressure on new entrants, resulting in inherently lower manufacturing shares. Thus, a nation’s lower manufacturing share might not reflect inefficiency but rather the influence of global economic conditions.

Export as PDF
Here are some key quotes and paraphrased segments from the video related to the concept of industry polarization, structural change, and deindustrialization:

1. **Definition of Industry Polarization**:
"Industry polarization is a term that describes a particular phenomenon regarding the sectoral composition of economic activity, particularly with respect to manufacturing share."

2. **Historical Convideo**:
"Simon Kuznets in the 1950s noted that as countries develop, the share of economic activity in agriculture decreases over time, transitioning into manufacturing and services, leading to what's known as industrialization."

3. **Current Trends in Manufacturing**:
"Today, particularly in advanced economies, we observe a transition out of manufacturing and into services. Manufacturing as a share of economic activity is shrinking."

4. **Premature Deindustrialization**:
"Danny Rodrik referred to this phenomenon as 'premature deindustrialization,' where countries do not attain as large a peak manufacturing share before deindustrializing."

5. **Different Trends Across Countries**:
"Countries today are attaining lower peaks in manufacturing share compared to their predecessors a few decades ago. For example, while the UK or US might have had a peak in the 1950s or 60s, a country like India might only peak at 20-25%."

6. **Evidence from Data**:
"We find that the manufacturing hump is declining over time and varies across countries. The highest manufacturing share has remained around 40% since 1970, but the lowest has fallen from about 10% to half that."

7. **International Trade**:
"The concentration of manufacturing production in fewer countries and the associated trade dynamics are key factors in understanding deindustrialization. As manufacturing production is increasingly concentrated, these countries are taking on more global production and exporting it."

8. **Mechanisms of Structural Change**:
"Our model includes mechanisms like changes in income (which affects consumption patterns) and changes in relative prices (driven by technological advancements), both of which influence the structural changes in economies."

9. **Technological Change and Deindustrialization**:
"We found that technological change contributed about 60% to the observed deindustrialization phenomenon. This drives the manufacturing productivity growth at a faster rate than in services."

10. **Policy Implications**:
"Our results suggest that countries entering the global stage later will inherently have a lower manufacturing share, which is not necessarily a bad outcome. The focus should be on the productivity of sectors rather than just their shares."

11. **Future Research Directions**:
"An interesting path forward is to consider how trade itself affects the technologies used within countries, acknowledging that integration into the global economy significantly impacts technological advancements."

These quotes and ideas encapsulate the main themes of industry polarization, the historical convideo of structural change, the significance of trade and technology, and their implications for policy and future research.

X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:90000

1
00:00:00.285 --> 00:00:03.965
Industry polarization is, is a term that we kind of,

2
00:00:03.965 --> 00:00:06.685
just came up with to describe a particular phenomenon,

3
00:00:06.985 --> 00:00:11.085
that, the sectoral composition

4
00:00:11.085 --> 00:00:13.725
of economic activity, and particularly with respect

5
00:00:13.825 --> 00:00:17.085
to manufacturing share in a country's economic activity.

6
00:00:17.865 --> 00:00:21.005
If we look at, those shares across countries

7
00:00:21.545 --> 00:00:24.285
and measure the dispersion, the observation is

8
00:00:24.285 --> 00:00:28.285
that dispersion has, has increased over time.

9
00:00:28.465 --> 00:00:30.925
And so we're referring to that particular phenomenon

10
00:00:31.305 --> 00:00:32.685
as industry polarization.

11
00:00:33.045 --> 00:00:36.205
Countries are sometimes more polarized in terms of,

12
00:00:36.205 --> 00:00:38.285
the mix of economic activity,

13
00:00:38.295 --> 00:00:39.485
today than they were in the past.

14
00:00:39.785 --> 00:00:42.445
The literature on structural change is, is

15
00:00:42.445 --> 00:00:44.725
where our research, sits.

16
00:00:45.385 --> 00:00:48.285
And, and that literature is, is,

17
00:00:48.285 --> 00:00:49.605
focused on explaining

18
00:00:49.605 --> 00:00:52.845
and understanding why within a country the mix

19
00:00:52.845 --> 00:00:53.885
of economic activity,

20
00:00:53.885 --> 00:00:56.285
particularly across three broad sectors.

21
00:00:56.415 --> 00:00:58.765
Think of a primary sector like agriculture,

22
00:00:58.845 --> 00:01:00.845
a secondary sector like manufacturing,

23
00:01:00.945 --> 00:01:03.005
and a tertiary sector like services.

24
00:01:04.065 --> 00:01:08.005
And the observation going back to conet in,

25
00:01:08.005 --> 00:01:12.485
Simon Conet in the 1950s, is that as countries develop,

26
00:01:12.675 --> 00:01:15.325
initially when their income per capita is relatively low,

27
00:01:15.325 --> 00:01:17.805
there's a large share of economic activity

28
00:01:17.805 --> 00:01:19.725
that's engaged in the agriculture sector.

29
00:01:20.265 --> 00:01:22.565
And over time, as the economy transitions

30
00:01:22.665 --> 00:01:27.285
and grows, resources, shift away from agriculture

31
00:01:27.285 --> 00:01:29.285
and into manufacturing and services,

32
00:01:29.745 --> 00:01:31.685
and that's what's known as industrialization.

33
00:01:31.705 --> 00:01:34.445
So you can imagine the industrial revolution is kind

34
00:01:34.445 --> 00:01:38.565
of like a, kind of where, you know, this phenomenon,

35
00:01:38.625 --> 00:01:41.485
initially, originated

36
00:01:43.055 --> 00:01:45.675
and eventually over time, what we see,

37
00:01:45.675 --> 00:01:49.355
particularly now in advanced economies is while agriculture

38
00:01:49.355 --> 00:01:51.395
share has, is low and kind remains low,

39
00:01:51.855 --> 00:01:53.715
but now there's a transition out

40
00:01:53.715 --> 00:01:55.075
of manufacturing and into services.

41
00:01:55.075 --> 00:01:57.355
Manufacturing is as a share

42
00:01:57.575 --> 00:01:59.395
of economic activity is shrinking.

43
00:01:59.575 --> 00:02:00.795
So if you look at the whole picture,

44
00:02:01.375 --> 00:02:04.595
you first have the rise in manufacturing industrialization,

45
00:02:04.595 --> 00:02:06.795
and then you have the decline in manufacturing share,

46
00:02:06.805 --> 00:02:08.715
which is de-industrialization.

47
00:02:08.895 --> 00:02:10.995
That's the other word in our title.

48
00:02:11.895 --> 00:02:15.835
Um, so that's the backdrop, for it.

49
00:02:15.935 --> 00:02:18.635
Um, thinking about these manufacturing shares,

50
00:02:18.635 --> 00:02:20.115
when we think about industry polarization,

51
00:02:20.135 --> 00:02:23.115
so let me give a little more context, to connect it.

52
00:02:23.115 --> 00:02:26.155
Why we think industry polarization is useful

53
00:02:26.155 --> 00:02:28.315
for thinking about the structural change phenomenon.

54
00:02:29.935 --> 00:02:32.635
If we, as I just described,

55
00:02:32.805 --> 00:02:35.435
these manufacturing humps over time, as countries develop,

56
00:02:35.575 --> 00:02:37.835
if we look at countries that experienced

57
00:02:38.365 --> 00:02:40.755
these humps earlier on

58
00:02:40.815 --> 00:02:44.355
and compare them to their predecessors later on

59
00:02:44.355 --> 00:02:48.035
that are experiencing these humps, now what we see is

60
00:02:48.035 --> 00:02:52.425
that these humps are kind of lower, even if you

61
00:02:53.295 --> 00:02:56.465
control for the level of economic, income per capita.

62
00:02:57.685 --> 00:03:00.725
The idea being that, say in the 1950s

63
00:03:00.905 --> 00:03:03.365
or 1960s, the UK

64
00:03:03.425 --> 00:03:06.285
or the US would've attained a peak manufacturing share

65
00:03:06.285 --> 00:03:08.965
of say, 35% of its GDP.

66
00:03:09.785 --> 00:03:12.565
Uh, whereas today, if you look at a, a country like India,

67
00:03:12.625 --> 00:03:13.805
for example, you know,

68
00:03:14.385 --> 00:03:17.045
it may be attaining a peak at a much lower level,

69
00:03:17.385 --> 00:03:19.565
say somewhere between 20 and 25%.

70
00:03:19.945 --> 00:03:22.685
And this is kind of a very systematic,

71
00:03:22.855 --> 00:03:24.845
trend over time is that,

72
00:03:25.165 --> 00:03:26.965
countries are attaining lower and lower peaks.

73
00:03:27.265 --> 00:03:29.885
Now, this observation, this is not novel to us.

74
00:03:29.895 --> 00:03:34.085
We're not, the first ones to look at this, this was,

75
00:03:34.155 --> 00:03:37.485
initially, pointed out by Danny Roderick in,

76
00:03:37.545 --> 00:03:39.045
in a 2016 paper.

77
00:03:39.945 --> 00:03:41.685
And he referred to this phenomenon

78
00:03:41.745 --> 00:03:43.805
as premature deindustrialization.

79
00:03:43.905 --> 00:03:47.005
The idea being is that, you know, countries today,

80
00:03:47.005 --> 00:03:50.365
they're not attaining as large of a peak manufacturing share

81
00:03:50.365 --> 00:03:51.725
before they de-industrialized.

82
00:03:51.985 --> 00:03:56.005
And that sense, it might be premature, we're, we've tried

83
00:03:56.005 --> 00:03:57.805
to avoid a little bit using the word premature

84
00:03:58.045 --> 00:03:59.525
'cause it feels a little bit loaded

85
00:03:59.665 --> 00:04:02.845
and might suggest that there's something,

86
00:04:02.915 --> 00:04:05.405
some inefficiency in the background or some kind of problem.

87
00:04:05.505 --> 00:04:10.105
And so we're approaching this question kind of,

88
00:04:10.495 --> 00:04:13.185
without taking any position on whether something might be

89
00:04:13.185 --> 00:04:14.785
wrong or or not, we're trying

90
00:04:14.805 --> 00:04:17.025
to understand why it's the case, right?

91
00:04:17.085 --> 00:04:19.305
Can we explain kind of what are the drivers

92
00:04:19.335 --> 00:04:20.705
that lead to this phenomenon?

93
00:04:20.705 --> 00:04:22.465
And then we could begin to think about, you know,

94
00:04:22.465 --> 00:04:25.265
whether there's room for policy to, you know, if there is,

95
00:04:25.485 --> 00:04:26.625
if there's, if there's a problem

96
00:04:26.625 --> 00:04:27.865
that needs to be fixed, per se.

97
00:04:28.965 --> 00:04:31.385
Um, so,

98
00:04:32.165 --> 00:04:34.505
how does industry polarization come into this?

99
00:04:34.895 --> 00:04:36.505
Well, this observation

100
00:04:36.895 --> 00:04:39.025
that these manufacturing piece getting lower

101
00:04:39.045 --> 00:04:43.065
and lower, we can kind of summarize that in a way, is if we,

102
00:04:44.125 --> 00:04:46.885
if we take a bunch of countries,

103
00:04:47.385 --> 00:04:49.365
and particular our sample, we're kind

104
00:04:49.365 --> 00:04:52.445
of tracking countries over four decades, starting in 1970,

105
00:04:53.385 --> 00:04:55.445
and it's a mix of OECD countries

106
00:04:55.445 --> 00:04:56.885
and some emerging economies.

107
00:04:57.545 --> 00:05:00.965
And the OECD countries since 1970, think of the us, uk,

108
00:05:01.175 --> 00:05:03.125
Japan, we kind of see the declining part

109
00:05:03.125 --> 00:05:04.125
of the manufacturing hump.

110
00:05:04.125 --> 00:05:07.445
And some of the emerging economies like Indonesia, China,

111
00:05:07.735 --> 00:05:10.285
India, we see kind of the rising part of the hump.

112
00:05:10.705 --> 00:05:13.085
And then we have a few countries that kind of had,

113
00:05:13.585 --> 00:05:16.325
growth miracles, let's say over this period, Korea

114
00:05:16.585 --> 00:05:18.325
and Taiwan, you see kind of the full hump.

115
00:05:18.745 --> 00:05:21.245
And so you could piece all these data points together,

116
00:05:22.225 --> 00:05:24.645
and if you do that and kind of look at them against

117
00:05:25.265 --> 00:05:28.445
the respective income per capita, at each one

118
00:05:28.485 --> 00:05:30.805
of those points, you can kind of trace out the hump.

119
00:05:31.225 --> 00:05:35.325
And now if we look at this manufacturing hump in only,

120
00:05:35.665 --> 00:05:39.205
say the pre 1990 period, say if we split our sample in half,

121
00:05:39.595 --> 00:05:42.645
that hump tends to be higher than in the post 1990 period.

122
00:05:42.745 --> 00:05:45.365
So that's what we're saying is that even at the same level

123
00:05:45.365 --> 00:05:47.765
of income countries today

124
00:05:48.385 --> 00:05:51.485
are at a lower manufacturing employment share than

125
00:05:52.075 --> 00:05:54.565
what their predecessor would've been a few decades ago.

126
00:05:55.665 --> 00:05:56.765
So that observation

127
00:05:56.825 --> 00:05:58.845
Is about the expected path, right?

128
00:05:58.845 --> 00:06:01.045
If we trace out this path, it's an expected path.

129
00:06:01.665 --> 00:06:03.725
Um, so it's kind of an average in some sense.

130
00:06:04.525 --> 00:06:06.845
Industry polarization is about the variance. Okay?

131
00:06:07.465 --> 00:06:09.645
So there could be a couple things going on

132
00:06:09.705 --> 00:06:11.845
for why this expected path is falling.

133
00:06:11.865 --> 00:06:16.015
It could just be globally in every country, you know,

134
00:06:16.015 --> 00:06:18.015
manufacturing share is just shrinking.

135
00:06:18.195 --> 00:06:21.815
So in later years, the whole average has shifted down

136
00:06:21.815 --> 00:06:25.445
because every, every, every individual's country's,

137
00:06:25.445 --> 00:06:27.405
manufacturing share has fallen, okay?

138
00:06:27.945 --> 00:06:30.205
But what we actually find is that's not the case.

139
00:06:31.065 --> 00:06:33.645
And the fact that the variance is changing

140
00:06:34.265 --> 00:06:37.005
in these manufacturing shares across countries,

141
00:06:37.105 --> 00:06:39.765
that's telling us something about what's going on

142
00:06:39.765 --> 00:06:41.205
underneath, behind the average.

143
00:06:41.625 --> 00:06:42.685
So what we actually see,

144
00:06:42.685 --> 00:06:44.765
if we look at the actual distribution

145
00:06:44.765 --> 00:06:46.565
of manufacturing shares across countries,

146
00:06:47.905 --> 00:06:51.005
the highest in any year has been around 40%

147
00:06:51.695 --> 00:06:53.045
since 1970.

148
00:06:53.435 --> 00:06:55.085
Okay? Now that doesn't mean

149
00:06:55.635 --> 00:06:58.285
it's the same country every single year

150
00:06:58.285 --> 00:07:00.085
that occupies that top share.

151
00:07:00.625 --> 00:07:03.965
In 1970, this could be, you could think of, you know, Japan

152
00:07:04.105 --> 00:07:06.485
or, or Korea in the eighties,

153
00:07:06.785 --> 00:07:10.005
but then in later years, it's, you think

154
00:07:10.005 --> 00:07:12.725
of like Indonesia say, in more recent decades, right?

155
00:07:12.725 --> 00:07:14.925
So it's a different country occupying that top position,

156
00:07:15.185 --> 00:07:18.685
but the top position is pretty consistently around 40%.

157
00:07:19.415 --> 00:07:21.395
If we look at other parts of the, of,

158
00:07:21.415 --> 00:07:23.355
of this distribution across countries

159
00:07:23.355 --> 00:07:25.955
of these manufacturing shares, in 1970,

160
00:07:26.135 --> 00:07:28.635
the lowest in our sample was around 20%.

161
00:07:29.455 --> 00:07:31.555
By the end of the sample, four decades later,

162
00:07:31.895 --> 00:07:33.795
the lowest was about 10%.

163
00:07:34.025 --> 00:07:35.075
It's fallen by half.

164
00:07:35.735 --> 00:07:38.715
And again, the identity of the country

165
00:07:38.715 --> 00:07:41.355
that's in the lowest position is not necessarily

166
00:07:41.415 --> 00:07:42.515
the same year by year.

167
00:07:42.945 --> 00:07:44.995
Okay? And we also see the median

168
00:07:45.215 --> 00:07:49.675
of this distribution declining from about 30% in 1970

169
00:07:50.215 --> 00:07:52.675
to about just around 25%

170
00:07:52.675 --> 00:07:55.595
or a little under that, by 2011.

171
00:07:55.735 --> 00:07:59.435
So that tells us that this decline in the average

172
00:07:59.435 --> 00:08:02.875
manufacturing share is really capturing the fact

173
00:08:02.875 --> 00:08:05.035
that there's a, there's a bulk of countries,

174
00:08:05.035 --> 00:08:07.115
particularly the bottom of the distribution

175
00:08:07.545 --> 00:08:09.875
that are pulling the average down, okay?

176
00:08:10.295 --> 00:08:11.715
But there's countries,

177
00:08:12.255 --> 00:08:13.875
or the, the countries in the top position

178
00:08:13.875 --> 00:08:14.995
aren't necessarily fault to.

179
00:08:15.015 --> 00:08:16.315
So it's not a phenomenon

180
00:08:16.315 --> 00:08:18.955
that every country's manufacturing share is declining.

181
00:08:20.095 --> 00:08:23.755
And so, so we bring that, to the table to re

182
00:08:24.095 --> 00:08:28.675
to reinvestigate this question of indu deindustrialization

183
00:08:28.695 --> 00:08:32.075
or prematurity, deindustrialization taking into account

184
00:08:32.225 --> 00:08:35.235
this, this, increasing dispersion in the manufacturing,

185
00:08:35.575 --> 00:08:36.715
value added shares.

186
00:08:37.015 --> 00:08:39.235
And it motivated us to investigate

187
00:08:40.105 --> 00:08:42.755
what might international trade be doing here.

188
00:08:43.635 --> 00:08:47.575
The idea being that is manufacturing production is

189
00:08:47.575 --> 00:08:49.415
increasingly concentrated in fewer

190
00:08:49.415 --> 00:08:53.475
and fewer countries, then it must be presumably the case

191
00:08:54.265 --> 00:08:57.015
that these countries are taking on more

192
00:08:57.015 --> 00:08:58.175
and more of global production

193
00:08:58.275 --> 00:09:00.095
and then exporting it to the rest of the world.

194
00:09:00.355 --> 00:09:02.455
And so we wanna investigate that

195
00:09:02.555 --> 00:09:06.735
and ask, can that particular mechanism, international trade,

196
00:09:06.735 --> 00:09:09.135
in this increasing concentration of production, can

197
00:09:09.135 --> 00:09:11.055
that explain, or how much can explain

198
00:09:11.635 --> 00:09:15.535
of the deindustrialization phenomenon our work is, is,

199
00:09:15.955 --> 00:09:18.495
it, it's centered in the structural change literature,

200
00:09:18.495 --> 00:09:20.095
but particularly in this direction,

201
00:09:20.095 --> 00:09:21.175
this more recent direction

202
00:09:21.175 --> 00:09:22.975
of understanding deindustrialization

203
00:09:22.995 --> 00:09:26.695
or premature deindustrialization as, as Danny Roderick,

204
00:09:26.875 --> 00:09:29.255
had kind of, originated.

205
00:09:30.115 --> 00:09:33.895
Um, now the way we approach the question is we take

206
00:09:35.145 --> 00:09:36.745
a quantitative approach.

207
00:09:36.845 --> 00:09:39.945
In other words, we're gonna take data and,

208
00:09:40.005 --> 00:09:42.385
but we're gonna, we're gonna examine the data

209
00:09:42.385 --> 00:09:44.225
through the lens of a theoretical model.

210
00:09:45.485 --> 00:09:49.305
And in the model, we're gonna incorporate some

211
00:09:50.025 --> 00:09:52.865
standard ingredients from the theoretical

212
00:09:53.185 --> 00:09:54.225
structural change literature.

213
00:09:54.525 --> 00:09:58.425
And the key ones are due to,

214
00:09:59.545 --> 00:10:02.145
Sergio Rebello and co-authors in,

215
00:10:02.645 --> 00:10:04.465
in a paper in the early two thousands,

216
00:10:05.675 --> 00:10:08.175
and another one by, way in PIs

217
00:10:08.795 --> 00:10:11.055
in another paper in, 2007.

218
00:10:11.115 --> 00:10:13.055
And these, what are these key ingredients?

219
00:10:13.055 --> 00:10:17.815
The re rebello et all paper has this framework

220
00:10:18.435 --> 00:10:22.095
in which structural change could be generated,

221
00:10:22.245 --> 00:10:24.575
through changing income through what,

222
00:10:25.195 --> 00:10:28.165
what economists call non homoe preferences.

223
00:10:28.185 --> 00:10:31.365
That's just a fancy word for saying,

224
00:10:32.115 --> 00:10:36.125
changes in your income affect the mix of consumption

225
00:10:36.125 --> 00:10:37.205
that you wanna undertake.

226
00:10:37.545 --> 00:10:41.205
The idea being that as you become, let's say you're richer,

227
00:10:41.795 --> 00:10:44.725
your relative demand for agriculture might fall.

228
00:10:44.755 --> 00:10:47.725
Doesn't mean you, let's say consume less food, for example,

229
00:10:48.185 --> 00:10:50.365
but at some point you're kind of satiated with food

230
00:10:50.385 --> 00:10:53.045
and an extra extra dollar of earnings that you get,

231
00:10:53.045 --> 00:10:54.885
you're gonna spend elsewhere, say,

232
00:10:54.885 --> 00:10:56.845
on luxury things like services, like,

233
00:10:56.855 --> 00:10:59.045
which would include things like health and education

234
00:10:59.505 --> 00:11:00.885
or travel and dining out

235
00:11:00.885 --> 00:11:02.125
at restaurants and things like this.

236
00:11:02.585 --> 00:11:04.685
And so the idea being that as your income grows,

237
00:11:04.705 --> 00:11:06.925
the additional dollars that you're getting in income are

238
00:11:06.925 --> 00:11:09.925
increasingly spent on services and less on agriculture.

239
00:11:09.925 --> 00:11:12.365
And that kind of generates decline in agriculture share

240
00:11:12.365 --> 00:11:13.645
and a rise in services share.

241
00:11:14.065 --> 00:11:15.565
And then the byproduct is kind

242
00:11:15.565 --> 00:11:16.605
of this hump in manufacturing.

243
00:11:17.145 --> 00:11:19.765
Um, so that's one kind of key mechanism

244
00:11:19.765 --> 00:11:20.845
that's been estimated

245
00:11:21.225 --> 00:11:24.765
and studied, very extensively in the literature,

246
00:11:24.945 --> 00:11:26.365
and it's kind of a staple ingredient.

247
00:11:26.505 --> 00:11:27.765
So we take that on board.

248
00:11:28.065 --> 00:11:31.525
The other ingredient is, way anti's,

249
00:11:32.745 --> 00:11:34.125
mechanism.

250
00:11:34.235 --> 00:11:37.615
This is sometimes called a relative price effect

251
00:11:37.615 --> 00:11:38.735
or a bowel effect.

252
00:11:38.915 --> 00:11:41.255
Uh, but essentially what that is, is,

253
00:11:42.575 --> 00:11:44.515
is generating a mechanism

254
00:11:44.515 --> 00:11:46.635
through which changes in relative prices

255
00:11:47.995 --> 00:11:49.495
affect your expenditures.

256
00:11:50.035 --> 00:11:54.045
And, and those changes in relative prices are di driven

257
00:11:54.505 --> 00:11:57.365
by changes in technology.

258
00:11:58.245 --> 00:12:03.175
The idea is this, what we see in the data over time is that

259
00:12:03.895 --> 00:12:06.175
productivity growth has generally

260
00:12:06.685 --> 00:12:08.695
been more prevalent in the manufacturing sector

261
00:12:08.885 --> 00:12:10.055
than in the service sector.

262
00:12:10.515 --> 00:12:14.455
You could think of things like, you know, more automation,

263
00:12:14.875 --> 00:12:18.215
things that just increase the output for a given amount

264
00:12:18.215 --> 00:12:20.455
of workers through this technological change.

265
00:12:20.665 --> 00:12:23.575
Where in the service sectors, the standard example,

266
00:12:23.635 --> 00:12:24.815
and this might be a bit extreme,

267
00:12:24.815 --> 00:12:26.855
but just to kind of fix ideas, like,

268
00:12:26.865 --> 00:12:30.145
let's say producing a haircut, you know, it,

269
00:12:30.285 --> 00:12:32.585
it takes a half hour to produce a haircut today,

270
00:12:32.605 --> 00:12:33.865
and it took a half hour to produce

271
00:12:33.865 --> 00:12:35.745
that same haircut 50 years ago, dude.

272
00:12:36.125 --> 00:12:39.755
And so, so what, what does this do well over time?

273
00:12:41.265 --> 00:12:44.365
If manufacturing productivity growth is exceeding that

274
00:12:44.365 --> 00:12:48.415
and services growth, then that's gonna lead to lower costs

275
00:12:48.515 --> 00:12:50.455
of producing an additional unit

276
00:12:50.455 --> 00:12:53.655
of manufacturing output relative to say, a haircut.

277
00:12:54.275 --> 00:12:55.615
Um, and therefore the price

278
00:12:55.615 --> 00:12:57.975
of manufacturing would decline relative to services.

279
00:12:58.405 --> 00:13:00.975
Okay? Now, how does this feed back into structural change?

280
00:13:01.935 --> 00:13:04.625
Well, here we

281
00:13:06.515 --> 00:13:10.255
have this notion of, say, manufacturing

282
00:13:10.355 --> 00:13:13.655
and services being complimentary, which is just a fancy word

283
00:13:13.715 --> 00:13:18.215
for saying, you, you tend to not be able to substitute one

284
00:13:18.315 --> 00:13:22.065
for the other, much like you would say, you know,

285
00:13:22.215 --> 00:13:24.065
Coke versus Pepsi, right?

286
00:13:24.065 --> 00:13:26.665
If the, if the price of one goes up to the relative

287
00:13:26.665 --> 00:13:28.945
to the other, you just consume more of the cheaper one.

288
00:13:29.065 --> 00:13:30.305
'cause they're very similar, okay?

289
00:13:30.765 --> 00:13:34.385
But when we're thinking about broad classifications of,

290
00:13:34.565 --> 00:13:38.105
of sectors, it's not as easy to substitute, say,

291
00:13:38.265 --> 00:13:40.225
a haircut for a new computer, right?

292
00:13:40.245 --> 00:13:42.385
So if the price of haircuts are increasing relative

293
00:13:42.385 --> 00:13:46.305
to computers, you spend more on haircuts and,

294
00:13:46.535 --> 00:13:47.585
because you just,

295
00:13:48.145 --> 00:13:49.985
a computer's not a viable substitute for it.

296
00:13:49.985 --> 00:13:51.265
So you just spend more on the thing

297
00:13:51.265 --> 00:13:53.145
that's experiencing higher relative price.

298
00:13:53.525 --> 00:13:56.345
Or the flip side of that is you spend less on the thing

299
00:13:56.615 --> 00:13:58.825
with, the decreasing relative price.

300
00:13:59.415 --> 00:14:01.465
Okay? And so what, what does that do?

301
00:14:01.465 --> 00:14:04.585
Well, if we have this technological change happening

302
00:14:04.845 --> 00:14:07.545
that's driving down the price of manufacturing relative

303
00:14:07.545 --> 00:14:10.065
to services, that's gonna give rise to

304
00:14:11.325 --> 00:14:14.005
a change in a shift in economic economic activity

305
00:14:14.555 --> 00:14:16.285
away from manufacturing to services.

306
00:14:16.945 --> 00:14:21.245
And so that's really a mechanism which, helps us explain

307
00:14:21.825 --> 00:14:26.215
why in the more advanced economies, the manufacturing

308
00:14:26.805 --> 00:14:28.655
hump is decreasing.

309
00:14:28.945 --> 00:14:30.775
We're talking about the decreasing end in services,

310
00:14:30.775 --> 00:14:31.855
share continues rising.

311
00:14:32.315 --> 00:14:35.695
So when you combine these income effects, as I described,

312
00:14:35.695 --> 00:14:38.095
which is really gonna be important for ex

313
00:14:38.095 --> 00:14:41.375
explaining why agriculture share is shrinking, we combine

314
00:14:41.375 --> 00:14:43.895
that with these relative price effects, which is going

315
00:14:43.895 --> 00:14:46.615
to give us a strong mechanism for,

316
00:14:46.725 --> 00:14:48.215
service share to increase.

317
00:14:48.715 --> 00:14:50.255
Um, then you could kind

318
00:14:50.255 --> 00:14:52.965
of imagine is the manufacturing hump is kind

319
00:14:52.965 --> 00:14:54.365
of a residual of these two things.

320
00:14:54.505 --> 00:14:56.725
As agriculture shrinks, manufacturing rises,

321
00:14:56.905 --> 00:14:58.965
and as a relative price of manufacturing falls,

322
00:14:59.195 --> 00:15:01.285
then eventually manufacturing share is gonna fall.

323
00:15:01.705 --> 00:15:03.365
So that's kind of the key ingredients

324
00:15:03.365 --> 00:15:04.405
we take into our model.

325
00:15:04.905 --> 00:15:07.045
We add a few things to it. Okay?

326
00:15:07.065 --> 00:15:10.845
So the first is, as, as I talked about

327
00:15:10.845 --> 00:15:13.165
before, is international trade.

328
00:15:13.395 --> 00:15:17.285
Okay? This is gonna give us another mechanism which alters

329
00:15:17.285 --> 00:15:18.605
the mix of economic activity.

330
00:15:19.635 --> 00:15:23.535
In particular, you might imagine, say, think

331
00:15:23.535 --> 00:15:26.575
of Japan in the 1970s and eighties, right?

332
00:15:26.725 --> 00:15:30.375
It's really coming onto the world stage, producing a ton

333
00:15:30.375 --> 00:15:34.015
of manufactured goods, Nintendos and Toyotas

334
00:15:34.015 --> 00:15:35.335
and Hondas, right?

335
00:15:35.435 --> 00:15:37.175
And doing it very efficiently.

336
00:15:37.875 --> 00:15:41.215
Um, that's gonna grow its manufacturing sector

337
00:15:41.215 --> 00:15:43.735
because it's exporting to the rest of the world, okay?

338
00:15:43.885 --> 00:15:48.055
Independent of what's going on with these other mechanisms

339
00:15:48.055 --> 00:15:49.535
that I just, discussed.

340
00:15:51.165 --> 00:15:54.385
The other thing that we bring on board is a little more

341
00:15:54.385 --> 00:15:58.305
subtle, but it has to do with the underlying

342
00:15:59.435 --> 00:16:02.745
production, techniques, if you will.

343
00:16:03.805 --> 00:16:07.505
And so, when we look closely at the mix

344
00:16:08.205 --> 00:16:13.185
of intermediate goods that are used in production, this,

345
00:16:13.195 --> 00:16:14.785
these, the mixes are

346
00:16:15.305 --> 00:16:17.545
changing in a very systematic way, okay?

347
00:16:19.425 --> 00:16:22.645
As an example, you might imagine the production

348
00:16:22.645 --> 00:16:24.125
of an automobile, okay?

349
00:16:24.195 --> 00:16:26.685
Suppose you're thinking of, you know,

350
00:16:27.075 --> 00:16:29.845
general Motors in the 1950s, okay?

351
00:16:30.305 --> 00:16:32.325
How was a car produced? Okay?

352
00:16:32.345 --> 00:16:34.365
Now, I'm not an expert on car production,

353
00:16:34.825 --> 00:16:38.125
but if we look at the mix of, of stuff

354
00:16:38.125 --> 00:16:40.325
that's used in the auto sector, we can look at that data.

355
00:16:40.385 --> 00:16:43.445
And what we see is we look at manufacturing production,

356
00:16:44.105 --> 00:16:47.605
it would've used manufacturing inputs extensively.

357
00:16:48.025 --> 00:16:51.445
So to produce a car which would fall into manufacturing

358
00:16:51.505 --> 00:16:54.525
sector, you need to get lots of components

359
00:16:54.555 --> 00:16:56.045
that are also manufactured.

360
00:16:56.865 --> 00:17:00.435
Carburetors engines, metal fabricated metal,

361
00:17:01.105 --> 00:17:04.635
plastic rubbers, a whole bunch of stuff all coming out

362
00:17:04.635 --> 00:17:05.755
of the manufacturing sector.

363
00:17:05.855 --> 00:17:07.075
So as an input, okay?

364
00:17:07.075 --> 00:17:08.555
So manufacturing share

365
00:17:08.655 --> 00:17:11.515
as an intermediate input would've been very high.

366
00:17:12.015 --> 00:17:15.435
Now we look at the production of a car today, think

367
00:17:15.435 --> 00:17:17.585
of a Tesla, okay?

368
00:17:18.485 --> 00:17:21.525
It has a lot of manufacturing,

369
00:17:21.605 --> 00:17:26.325
manufactured intermediate inputs as well, steel batteries,

370
00:17:26.925 --> 00:17:28.365
whatever it may be.

371
00:17:28.945 --> 00:17:31.805
But there's not a greater share of stuff that's

372
00:17:32.325 --> 00:17:33.325
produced from the service sector.

373
00:17:33.375 --> 00:17:37.045
Think of all the software technology that's on board, okay?

374
00:17:37.235 --> 00:17:39.445
That wouldn't have been there in the 1950s.

375
00:17:39.585 --> 00:17:43.005
So services share in the final value of

376
00:17:43.005 --> 00:17:45.525
that car is much higher today than it was before.

377
00:17:46.145 --> 00:17:48.845
So we kind of, in some sense have a structural change

378
00:17:49.585 --> 00:17:52.285
in terms of intermediate input mix, okay?

379
00:17:52.305 --> 00:17:53.845
And we see this across the board,

380
00:17:53.845 --> 00:17:56.085
whether we look at production of manufactured things

381
00:17:56.625 --> 00:17:58.645
or production of agriculture output.

382
00:17:58.865 --> 00:18:02.045
We see the mix of intermediate inputs used even within

383
00:18:02.115 --> 00:18:05.005
agriculture, shifting more and more towards services.

384
00:18:05.745 --> 00:18:08.405
And then finally, you look at the production of services,

385
00:18:08.675 --> 00:18:12.285
they also are increasingly using their own sector's

386
00:18:12.305 --> 00:18:13.605
output as an input service.

387
00:18:13.625 --> 00:18:18.345
So you see services, services,

388
00:18:18.345 --> 00:18:21.305
sector are not only becoming a larger

389
00:18:21.305 --> 00:18:23.065
and larger share of say, consumer demand,

390
00:18:23.175 --> 00:18:26.745
they're becoming an increasing share of producer demand

391
00:18:26.745 --> 00:18:28.225
through the intermediate input mix.

392
00:18:28.725 --> 00:18:31.705
And so we bring that on board into the paper too,

393
00:18:32.035 --> 00:18:34.625
which is gonna give us additional channels

394
00:18:34.625 --> 00:18:38.305
through which structural change is gonna under is gonna un,

395
00:18:38.445 --> 00:18:39.545
un, un unroll.

396
00:18:40.365 --> 00:18:44.505
So the idea is this, suppose the final consumer

397
00:18:45.905 --> 00:18:47.485
has more income for whatever reason,

398
00:18:47.505 --> 00:18:49.005
and they wanna consume more services.

399
00:18:49.275 --> 00:18:50.725
Okay? So if you have kind

400
00:18:50.725 --> 00:18:53.005
of old fashioned structural change, like, you know,

401
00:18:53.005 --> 00:18:54.205
from the early two thousands,

402
00:18:54.515 --> 00:18:58.125
when you add this additional mechanism in what happens?

403
00:18:58.195 --> 00:19:01.445
Well, to produce that additional unit of services, I need

404
00:19:01.445 --> 00:19:04.445
to increasingly use services itself as an input.

405
00:19:05.215 --> 00:19:08.075
Um, that's gonna amplify the role

406
00:19:08.175 --> 00:19:10.475
of the underlying technological change

407
00:19:10.475 --> 00:19:13.235
and the underlying changes in income per capita in terms of

408
00:19:13.235 --> 00:19:17.515
how much it can, how much it will contribute to the,

409
00:19:17.565 --> 00:19:19.395
underlying structural change that takes place.

410
00:19:19.525 --> 00:19:22.675
We're looking into the window spanning four decades,

411
00:19:23.035 --> 00:19:25.955
starting in, 1970 or 1971.

412
00:19:26.095 --> 00:19:30.315
Um, that, that window of data is, mostly dictated by,

413
00:19:30.375 --> 00:19:33.105
you know, our requirements for

414
00:19:33.105 --> 00:19:34.505
what we need to investigate to question.

415
00:19:35.085 --> 00:19:38.385
Um, and this span of countries includes a lot

416
00:19:38.385 --> 00:19:42.545
of OECD countries, mostly US, Japan, Western Europe,

417
00:19:42.965 --> 00:19:46.225
and then a handful of, emerging economies.

418
00:19:46.845 --> 00:19:49.345
And so the combination of the advanced economies,

419
00:19:49.345 --> 00:19:51.065
which are kind of on a downwards part

420
00:19:51.065 --> 00:19:52.185
of the manufacturing hump,

421
00:19:52.185 --> 00:19:53.425
and then the emerging economies

422
00:19:53.425 --> 00:19:55.585
that are on the increasing part of the hump, we can kind

423
00:19:55.585 --> 00:19:58.745
of trace out along, we can kind of trace out a path

424
00:19:58.775 --> 00:20:02.145
with respect to a comfort capita if we keep track of

425
00:20:02.145 --> 00:20:03.905
where those countries, lie in terms

426
00:20:03.905 --> 00:20:04.985
of economic development.

427
00:20:05.445 --> 00:20:09.945
Um, and so our, our strategy is empirically we've,

428
00:20:10.125 --> 00:20:12.545
we look at that average path in both the pre 90

429
00:20:12.565 --> 00:20:14.305
and post 90 periods.

430
00:20:14.405 --> 00:20:18.905
And, and find as Roderick also found that the,

431
00:20:19.255 --> 00:20:21.745
that average path is lower in the post 90 period.

432
00:20:22.125 --> 00:20:25.265
And then we add to that the, this observation of,

433
00:20:25.605 --> 00:20:28.305
of increasing dispersion across countries.

434
00:20:28.805 --> 00:20:31.585
Um, in the manufacturing value added share,

435
00:20:31.645 --> 00:20:33.625
we use the regression to,

436
00:20:33.685 --> 00:20:35.505
to trace out these average paths.

437
00:20:36.525 --> 00:20:38.385
Um, so in the regression,

438
00:20:38.445 --> 00:20:42.945
we would take on the left hand side, the manufacturing

439
00:20:43.635 --> 00:20:48.045
share in a country's GDP, at a particular point in time.

440
00:20:48.065 --> 00:20:50.365
And then on the right hand side, we would have

441
00:20:50.365 --> 00:20:52.285
that country's income per capita.

442
00:20:52.785 --> 00:20:57.095
Um, and we would include, allow for a nonlinearity

443
00:20:57.095 --> 00:21:00.215
and have, you know, a quadratic, specification there.

444
00:21:00.595 --> 00:21:04.615
Um, importantly we would also include country fixed effects.

445
00:21:05.235 --> 00:21:07.455
Uh, the reason is just,

446
00:21:08.085 --> 00:21:10.215
some countries are perhaps more inclined

447
00:21:10.215 --> 00:21:13.335
to have large manufacturing shares compared to others, say,

448
00:21:13.335 --> 00:21:14.415
because of endowments.

449
00:21:14.565 --> 00:21:16.615
Yeah. Or the particular geography of the country.

450
00:21:16.715 --> 00:21:20.055
So we wanna cleanse the calculation from those particular,

451
00:21:20.195 --> 00:21:22.135
you know, kind of country characteristics

452
00:21:22.515 --> 00:21:26.175
and focus on over time what happens within a country

453
00:21:26.435 --> 00:21:28.775
as its income per capita changes, is what we're trying

454
00:21:28.795 --> 00:21:30.415
to get at using that regression.

455
00:21:30.835 --> 00:21:35.095
And, and then what our results, kind of point out is

456
00:21:35.095 --> 00:21:37.735
that, you know, that, in the post 90 period,

457
00:21:38.115 --> 00:21:40.055
the change over time would've,

458
00:21:40.205 --> 00:21:42.055
been less than in the pre 90 period.

459
00:21:42.595 --> 00:21:43.895
We take that on board,

460
00:21:43.895 --> 00:21:46.175
and then we put it into this quantitative,

461
00:21:46.175 --> 00:21:47.255
theoretical model.

462
00:21:47.475 --> 00:21:52.335
And we, we ask if we, if we take the observed volume

463
00:21:52.395 --> 00:21:57.375
of trade, that we have in the data, right?

464
00:21:57.375 --> 00:21:59.615
And we construct measures of trade frictions

465
00:21:59.615 --> 00:22:01.935
and things like this in order for the model

466
00:22:01.955 --> 00:22:03.935
to reproduce the trade flows that we observe.

467
00:22:04.715 --> 00:22:08.215
And if we measure sectoral productivity,

468
00:22:09.375 --> 00:22:11.795
in each country and each year directly from the data,

469
00:22:11.935 --> 00:22:14.235
we take that as an input into the model.

470
00:22:14.335 --> 00:22:17.155
So we kind of make the model match

471
00:22:17.975 --> 00:22:20.755
the underlying technological change that we observe

472
00:22:21.095 --> 00:22:23.955
and the underlying trade patterns that we observed.

473
00:22:24.055 --> 00:22:27.235
And then we ask, when we put these two things in,

474
00:22:27.415 --> 00:22:31.445
can we reproduce the rodricks fact, right?

475
00:22:31.465 --> 00:22:34.605
Or this premature de-industrialization phenomenon

476
00:22:34.985 --> 00:22:37.245
and the industry polarization.

477
00:22:38.105 --> 00:22:40.685
And we find that the model could generate a majority of it,

478
00:22:40.685 --> 00:22:42.645
it could generate all of the industry polarization

479
00:22:42.645 --> 00:22:44.365
because that's directly related to trade.

480
00:22:44.365 --> 00:22:45.445
It's kind of not surprising.

481
00:22:45.445 --> 00:22:48.245
And we've kind of cooked up the model to be consistent

482
00:22:48.245 --> 00:22:49.245
with the trade patterns.

483
00:22:49.515 --> 00:22:53.845
It's less trivial that the model would reproduce the,

484
00:22:54.225 --> 00:22:56.485
the hump shape, and in particular the fact

485
00:22:56.485 --> 00:22:58.685
that the hump was higher in the pre 90 period

486
00:22:58.685 --> 00:22:59.765
than the post 90 period.

487
00:23:00.065 --> 00:23:01.645
So the model generates that.

488
00:23:01.645 --> 00:23:04.165
And then we follow up by asking, well,

489
00:23:04.185 --> 00:23:06.245
how much did international trade

490
00:23:06.495 --> 00:23:08.125
contribute to these two things?

491
00:23:08.465 --> 00:23:10.245
And how much did technological change

492
00:23:10.415 --> 00:23:11.565
contribute to these things?

493
00:23:11.565 --> 00:23:13.405
We wanna kind of flip the switch on

494
00:23:13.405 --> 00:23:16.685
or off of either one of those two forces that we fed in.

495
00:23:17.705 --> 00:23:19.965
And what we find, this is kind of the, the,

496
00:23:19.965 --> 00:23:21.765
the main finding now is that

497
00:23:22.465 --> 00:23:26.525
the underlying technological change contributed about 60%

498
00:23:27.065 --> 00:23:29.205
to the de-industrialization as measured

499
00:23:29.205 --> 00:23:33.125
by the change in the pre 90 hump and the post 90 hump.

500
00:23:33.125 --> 00:23:35.565
So about two thirds of that almost is due

501
00:23:35.565 --> 00:23:36.645
to technological change.

502
00:23:38.225 --> 00:23:42.515
Um, it generated essentially enough technological change,

503
00:23:42.535 --> 00:23:45.725
didn't on its own contribute much to industry polarization.

504
00:23:46.105 --> 00:23:48.885
The flip side of that is industry polarization.

505
00:23:48.915 --> 00:23:51.405
This increasing concentration of production is due mostly

506
00:23:51.945 --> 00:23:53.925
to declining trade costs over time.

507
00:23:54.395 --> 00:23:59.225
Okay? Now, we find that on its own declining trade costs

508
00:24:00.655 --> 00:24:04.085
don't explain much of the premature deindustrialization,

509
00:24:05.175 --> 00:24:09.155
but I think the most interesting takeaway that we found is

510
00:24:09.155 --> 00:24:11.835
that international trade does matter,

511
00:24:11.935 --> 00:24:14.955
but only in the presence of

512
00:24:15.695 --> 00:24:17.275
the particular technological change.

513
00:24:18.065 --> 00:24:21.745
Um, and so let me just take a minute to explain that.

514
00:24:22.285 --> 00:24:26.895
The idea is that you can imagine globally what's going on is

515
00:24:26.895 --> 00:24:29.815
that manufacturing productivity growth tends

516
00:24:29.815 --> 00:24:31.855
to exceed services productivity growth

517
00:24:32.155 --> 00:24:33.535
that's gonna induce a fall on

518
00:24:33.535 --> 00:24:34.815
the relative price of manufacturing.

519
00:24:35.845 --> 00:24:39.835
Okay? So imagine you're experiencing that on your own

520
00:24:39.935 --> 00:24:43.275
as an isolated economy, and then you start trading with me,

521
00:24:43.275 --> 00:24:45.155
and I'm also experiencing the same kind

522
00:24:45.155 --> 00:24:46.595
of underlying technological change.

523
00:24:46.595 --> 00:24:49.115
And then we start trading with each other, right?

524
00:24:49.505 --> 00:24:50.955
What prices do you pay for?

525
00:24:50.955 --> 00:24:53.395
The things that I produce is gonna be a function

526
00:24:53.415 --> 00:24:55.635
of the technological change that I'm undergoing.

527
00:24:56.175 --> 00:24:59.275
So international trade then serves as a mechanism

528
00:24:59.335 --> 00:25:02.585
to propagate this technological change.

529
00:25:02.845 --> 00:25:06.145
And so when you combine it international trade with the,

530
00:25:06.655 --> 00:25:09.665
with the technological change, now you can account

531
00:25:09.685 --> 00:25:14.345
for almost all of the, observed,

532
00:25:14.645 --> 00:25:16.975
deindustrialization.

533
00:25:17.805 --> 00:25:20.425
So it's a little bit of a subtle effect of how trade works.

534
00:25:20.645 --> 00:25:22.465
It works as an amplification mechanism,

535
00:25:22.565 --> 00:25:25.455
but it, on its own, it would've through the lens

536
00:25:25.455 --> 00:25:27.375
of our model, wouldn't have contributed much

537
00:25:27.595 --> 00:25:28.695
to deindustrialization,

538
00:25:29.525 --> 00:25:31.065
but it doesn't mean it wasn't important.

539
00:25:31.605 --> 00:25:34.145
So that's kind of the main, that's the main punchline

540
00:25:34.145 --> 00:25:36.625
of the paper policy implications, I think,

541
00:25:37.625 --> 00:25:42.105
I think a lot is here, especially for, countries,

542
00:25:42.275 --> 00:25:44.505
maybe at lower levels of, in income per capita

543
00:25:44.735 --> 00:25:48.465
that are focusing on industrial policy, these debates

544
00:25:49.085 --> 00:25:53.665
almost always entail, kind of,

545
00:25:55.065 --> 00:25:58.805
can we follow the role of what South Korea did?

546
00:25:58.865 --> 00:26:00.725
Should we focus on export, right?

547
00:26:00.865 --> 00:26:02.725
So how do we grow manufacturing?

548
00:26:02.905 --> 00:26:04.325
So there's, there seems

549
00:26:04.345 --> 00:26:08.365
to be this sense in which growing manufacturing is,

550
00:26:08.785 --> 00:26:10.245
is the goal to begin with.

551
00:26:10.745 --> 00:26:13.765
And so I think what our, what our results suggest is that,

552
00:26:13.825 --> 00:26:18.245
you know, even in the absence of any inefficiencies

553
00:26:18.245 --> 00:26:22.365
or any kind of distortions, if you're kind of entering

554
00:26:23.145 --> 00:26:25.485
the global stage at a later date,

555
00:26:26.885 --> 00:26:29.495
just the underlying nature of the world

556
00:26:29.495 --> 00:26:32.735
that you're interacting with is going to, is going

557
00:26:32.735 --> 00:26:35.375
to put lower pressure on, in some sense,

558
00:26:35.875 --> 00:26:37.935
result in you having a lower manufacturing share

559
00:26:37.935 --> 00:26:39.055
than you otherwise would.

560
00:26:39.235 --> 00:26:42.375
And so that on its own is not necessarily a bad thing.

561
00:26:42.835 --> 00:26:44.165
And again, I just wanna kind

562
00:26:44.165 --> 00:26:46.045
of reiterate the way we measure shares.

563
00:26:46.215 --> 00:26:48.645
These are measured at price.

564
00:26:48.825 --> 00:26:50.005
It doesn't mean that the quantity

565
00:26:50.065 --> 00:26:51.365
of manufacturing is falling.

566
00:26:51.545 --> 00:26:53.725
It means that the price times quantity is falling.

567
00:26:53.905 --> 00:26:56.845
So you can have a low share because the price is low

568
00:26:56.845 --> 00:26:58.045
and the relative price could be low

569
00:26:58.045 --> 00:27:01.565
because you have a very productive technology there.

570
00:27:02.025 --> 00:27:05.325
And so, so I think that's an important thing

571
00:27:05.385 --> 00:27:08.205
to keep in mind in, in these policy debates

572
00:27:08.205 --> 00:27:10.365
of industrial policy, in particular.

573
00:27:11.105 --> 00:27:13.485
And now the other, the other part

574
00:27:13.485 --> 00:27:15.725
of your question was about where do we go from here?

575
00:27:16.235 --> 00:27:19.525
Well, I think very interesting path forward is

576
00:27:19.985 --> 00:27:21.845
to think about how trade itself

577
00:27:22.475 --> 00:27:24.645
affects the technologies that you use.

578
00:27:24.665 --> 00:27:27.765
The calculations that we did were kind of saying we kind

579
00:27:27.765 --> 00:27:31.085
of take technological changes given we didn't really ask

580
00:27:31.345 --> 00:27:32.765
how it got there, right?

581
00:27:32.875 --> 00:27:34.605
What gave rise to this higher

582
00:27:34.605 --> 00:27:36.125
manufacturing productivity growth?

583
00:27:36.505 --> 00:27:40.845
And so you might be imagining, for example, you know, well,

584
00:27:40.845 --> 00:27:42.565
how should I think about a country like China?

585
00:27:43.455 --> 00:27:46.045
Would it have the technologies that it has today

586
00:27:46.625 --> 00:27:51.015
if it weren't integrated in the global economy, be a trade

587
00:27:51.015 --> 00:27:52.095
for the last couple of decades?

588
00:27:52.635 --> 00:27:53.935
You know, I think most people would,

589
00:27:54.725 --> 00:27:57.275
their prior would be probably not okay.

590
00:27:57.575 --> 00:27:59.755
And so that's not something that we've taken

591
00:28:00.615 --> 00:28:02.515
on board into our analysis,

592
00:28:03.015 --> 00:28:06.275
and I think that's a clearly an important, aspect

593
00:28:06.275 --> 00:28:08.755
that would need to be added to get a better picture

594
00:28:08.775 --> 00:28:09.755
of what's really going on.

Understanding Industry Polarization and Structural Change

00:00:00 - Industry polarization is, is a term that we kind of, just came up with to describe a particular phenomenon, that, the sectoral composition of economic activity, and particularly with respect to manufacturing share in a country's economic activity. If we look at, those shares across countries and measure the dispersion, the observation is that dispersion has, has increased over time. And so we're referring to that particular phenomenon

00:00:33 - Countries are sometimes more polarized in terms of, the mix of economic activity, today than they were in the past. The literature on structural change is, is where our research, sits. And, and that literature is, is, focused on explaining and understanding why within a country the mix of economic activity, particularly across three broad sectors. Think of a primary sector like agriculture, a secondary sector like manufacturing, and a tertiary sector like services.

00:01:08 - Simon Conet in the 1950s, is that as countries develop, initially when their income per capita is relatively low, there's a large share of economic activity that's engaged in the agriculture sector. And over time, as the economy transitions and grows, resources, shift away from agriculture and into manufacturing and services, and that's what's known as industrialization. So you can imagine the industrial revolution is kind

Transition from Manufacturing to Services: Understanding De-Industrialization

00:01:38 - initially, originated and eventually over time, what we see, particularly now in advanced economies is while agriculture share has, is low and kind remains low, but now there's a transition out of manufacturing and into services. Manufacturing is as a share of economic activity is shrinking. So if you look at the whole picture, you first have the rise in manufacturing industrialization,

00:02:06 - which is de-industrialization. That's the other word in our title. so that's the backdrop, for it. thinking about these manufacturing shares, when we think about industry polarization, so let me give a little more context, to connect it. Why we think industry polarization is useful for thinking about the structural change phenomenon. If we, as I just described, these manufacturing humps over time, as countries develop,

00:02:38 - these humps earlier on and compare them to their predecessors later on that are experiencing these humps, now what we see is that these humps are kind of lower, even if you control for the level of economic, income per capita. The idea being that, say in the 1950s or 1960s, the UK or the US would've attained a peak manufacturing share

Understanding Premature Deindustrialization

00:03:09 - whereas today, if you look at a, a country like India, for example, you know, it may be attaining a peak at a much lower level, say somewhere between 20 and 25%. And this is kind of a very systematic, trend over time is that, countries are attaining lower and lower peaks. Now, this observation, this is not novel to us. We're not, the first ones to look at this, this was, initially, pointed out by Danny Roderick in,

00:03:39 - And he referred to this phenomenon as premature deindustrialization. The idea being is that, you know, countries today, they're not attaining as large of a peak manufacturing share before they de-industrialized. And that sense, it might be premature, we're, we've tried to avoid a little bit using the word premature 'cause it feels a little bit loaded and might suggest that there's something, some inefficiency in the background or some kind of problem. And so we're approaching this question kind of,

00:04:13 - wrong or or not, we're trying to understand why it's the case, right? Can we explain kind of what are the drivers that lead to this phenomenon? And then we could begin to think about, you know, whether there's room for policy to, you know, if there is, if there's, if there's a problem that needs to be fixed, per se. so, how does industry polarization come into this? Well, this observation that these manufacturing piece getting lower and lower, we can kind of summarize that in a way, is if we,

Tracking Manufacturing Trends Across Economies

00:04:47 - and particular our sample, we're kind of tracking countries over four decades, starting in 1970, and it's a mix of OECD countries and some emerging economies. And the OECD countries since 1970, think of the us, uk, Japan, we kind of see the declining part of the manufacturing hump. And some of the emerging economies like Indonesia, China, India, we see kind of the rising part of the hump. And then we have a few countries that kind of had, growth miracles, let's say over this period, Korea

00:05:18 - And so you could piece all these data points together, and if you do that and kind of look at them against the respective income per capita, at each one of those points, you can kind of trace out the hump. And now if we look at this manufacturing hump in only, say the pre 1990 period, say if we split our sample in half, that hump tends to be higher than in the post 1990 period. So that's what we're saying is that even at the same level of income countries today

00:05:52 - what their predecessor would've been a few decades ago. So that observation Is about the expected path, right? If we trace out this path, it's an expected path. so it's kind of an average in some sense. Industry polarization is about the variance. Okay? So there could be a couple things going on for why this expected path is falling. It could just be globally in every country, you know, manufacturing share is just shrinking. So in later years, the whole average has shifted down

Trends in Global Manufacturing Shares: Declines and Variations

00:06:25 - manufacturing share has fallen, okay? But what we actually find is that's not the case. And the fact that the variance is changing in these manufacturing shares across countries, that's telling us something about what's going on underneath, behind the average. So what we actually see, if we look at the actual distribution of manufacturing shares across countries, the highest in any year has been around 40% since 1970.

00:06:55 - it's the same country every single year that occupies that top share. In 1970, this could be, you could think of, you know, Japan or, or Korea in the eighties, but then in later years, it's, you think of like Indonesia say, in more recent decades, right? So it's a different country occupying that top position, but the top position is pretty consistently around 40%. If we look at other parts of the, of, of this distribution across countries of these manufacturing shares, in 1970,

00:07:29 - By the end of the sample, four decades later, the lowest was about 10%. It's fallen by half. And again, the identity of the country that's in the lowest position is not necessarily the same year by year. Okay? And we also see the median of this distribution declining from about 30% in 1970 to about just around 25% or a little under that, by 2011. So that tells us that this decline in the average

Reevaluating Deindustrialization: The Role of International Trade

00:08:02 - that there's a, there's a bulk of countries, particularly the bottom of the distribution that are pulling the average down, okay? But there's countries, or the, the countries in the top position aren't necessarily fault to. So it's not a phenomenon that every country's manufacturing share is declining. And so, so we bring that, to the table to re to reinvestigate this question of indu deindustrialization or prematurity, deindustrialization taking into account

00:08:35 - value added shares. And it motivated us to investigate what might international trade be doing here. The idea being that is manufacturing production is increasingly concentrated in fewer and fewer countries, then it must be presumably the case that these countries are taking on more and more of global production and then exporting it to the rest of the world. And so we wanna investigate that

00:09:06 - in this increasing concentration of production, can that explain, or how much can explain of the deindustrialization phenomenon our work is, is, it, it's centered in the structural change literature, but particularly in this direction, this more recent direction of understanding deindustrialization or premature deindustrialization as, as Danny Roderick, had kind of, originated. now the way we approach the question is we take

Examining Structural Change Through Theoretical Models

00:09:36 - In other words, we're gonna take data and, but we're gonna, we're gonna examine the data through the lens of a theoretical model. And in the model, we're gonna incorporate some standard ingredients from the theoretical structural change literature. And the key ones are due to, Sergio Rebello and co-authors in, in a paper in the early two thousands,

00:10:08 - in another paper in, 2007. And these, what are these key ingredients? The re rebello et all paper has this framework in which structural change could be generated, through changing income through what, what economists call non homoe preferences. That's just a fancy word for saying, changes in your income affect the mix of consumption

00:10:37 - The idea being that as you become, let's say you're richer, your relative demand for agriculture might fall. Doesn't mean you, let's say consume less food, for example, but at some point you're kind of satiated with food and an extra extra dollar of earnings that you get, you're gonna spend elsewhere, say, on luxury things like services, like, which would include things like health and education or travel and dining out at restaurants and things like this. And so the idea being that as your income grows, the additional dollars that you're getting in income are

Shifts in Economic Sectors: Agriculture to Services and the Role of Technology

00:11:09 - And that kind of generates decline in agriculture share and a rise in services share. And then the byproduct is kind of this hump in manufacturing. so that's one kind of key mechanism that's been estimated and studied, very extensively in the literature, and it's kind of a staple ingredient. So we take that on board. The other ingredient is, way anti's, mechanism. This is sometimes called a relative price effect

00:11:38 - but essentially what that is, is, is generating a mechanism through which changes in relative prices affect your expenditures. And, and those changes in relative prices are di driven by changes in technology. The idea is this, what we see in the data over time is that productivity growth has generally been more prevalent in the manufacturing sector

00:12:10 - You could think of things like, you know, more automation, things that just increase the output for a given amount of workers through this technological change. Where in the service sectors, the standard example, and this might be a bit extreme, but just to kind of fix ideas, like, let's say producing a haircut, you know, it, it takes a half hour to produce a haircut today, and it took a half hour to produce that same haircut 50 years ago, dude. And so, so what, what does this do well over time?

Shifting Economic Dynamics: Manufacturing vs. Services

00:12:44 - and services growth, then that's gonna lead to lower costs of producing an additional unit of manufacturing output relative to say, a haircut. and therefore the price of manufacturing would decline relative to services. Okay? Now, how does this feed back into structural change? Well, here we have this notion of, say, manufacturing and services being complimentary, which is just a fancy word

00:13:18 - for the other, much like you would say, you know, Coke versus Pepsi, right? If the, if the price of one goes up to the relative to the other, you just consume more of the cheaper one. 'cause they're very similar, okay? But when we're thinking about broad classifications of, of sectors, it's not as easy to substitute, say, a haircut for a new computer, right? So if the price of haircuts are increasing relative to computers, you spend more on haircuts and,

00:13:48 - a computer's not a viable substitute for it. So you just spend more on the thing that's experiencing higher relative price. Or the flip side of that is you spend less on the thing with, the decreasing relative price. Okay? And so what, what does that do? Well, if we have this technological change happening that's driving down the price of manufacturing relative to services, that's gonna give rise to a change in a shift in economic economic activity away from manufacturing to services.

Shifts in Economic Activity: Decreasing Manufacturing and Rising Services

00:14:21 - why in the more advanced economies, the manufacturing hump is decreasing. We're talking about the decreasing end in services, share continues rising. So when you combine these income effects, as I described, which is really gonna be important for ex explaining why agriculture share is shrinking, we combine that with these relative price effects, which is going to give us a strong mechanism for, service share to increase.

00:14:50 - of imagine is the manufacturing hump is kind of a residual of these two things. As agriculture shrinks, manufacturing rises, and as a relative price of manufacturing falls, then eventually manufacturing share is gonna fall. So that's kind of the key ingredients we take into our model. We add a few things to it. Okay? So the first is, as, as I talked about before, is international trade. Okay? This is gonna give us another mechanism which alters the mix of economic activity.

00:15:23 - of Japan in the 1970s and eighties, right? It's really coming onto the world stage, producing a ton of manufactured goods, Nintendos and Toyotas and Hondas, right? And doing it very efficiently. that's gonna grow its manufacturing sector because it's exporting to the rest of the world, okay? Independent of what's going on with these other mechanisms that I just, discussed.

Evolution of Production Techniques in the Automotive Industry

00:15:54 - subtle, but it has to do with the underlying production, techniques, if you will. And so, when we look closely at the mix of intermediate goods that are used in production, this, these, the mixes are changing in a very systematic way, okay? As an example, you might imagine the production

00:16:24 - Suppose you're thinking of, you know, general Motors in the 1950s, okay? How was a car produced? Okay? Now, I'm not an expert on car production, but if we look at the mix of, of stuff that's used in the auto sector, we can look at that data. And what we see is we look at manufacturing production, it would've used manufacturing inputs extensively. So to produce a car which would fall into manufacturing sector, you need to get lots of components

00:16:56 - Carburetors engines, metal fabricated metal, plastic rubbers, a whole bunch of stuff all coming out of the manufacturing sector. So as an input, okay? So manufacturing share as an intermediate input would've been very high. Now we look at the production of a car today, think of a Tesla, okay? It has a lot of manufacturing, manufactured intermediate inputs as well, steel batteries,

Shifts in Economic Structure: The Rising Role of Services

00:17:28 - But there's not a greater share of stuff that's produced from the service sector. Think of all the software technology that's on board, okay? That wouldn't have been there in the 1950s. So services share in the final value of that car is much higher today than it was before. So we kind of, in some sense have a structural change in terms of intermediate input mix, okay? And we see this across the board, whether we look at production of manufactured things or production of agriculture output.

00:18:02 - agriculture, shifting more and more towards services. And then finally, you look at the production of services, they also are increasingly using their own sector's output as an input service. So you see services, services, sector are not only becoming a larger and larger share of say, consumer demand, they're becoming an increasing share of producer demand through the intermediate input mix. And so we bring that on board into the paper too,

00:18:34 - through which structural change is gonna under is gonna un, un, un unroll. So the idea is this, suppose the final consumer has more income for whatever reason, and they wanna consume more services. Okay? So if you have kind of old fashioned structural change, like, you know, from the early two thousands, when you add this additional mechanism in what happens? Well, to produce that additional unit of services, I need to increasingly use services itself as an input.

Investigating Structural Change Over Four Decades

00:19:08 - of the underlying technological change and the underlying changes in income per capita in terms of how much it can, how much it will contribute to the, underlying structural change that takes place. We're looking into the window spanning four decades, starting in, 1970 or 1971. that, that window of data is, mostly dictated by, you know, our requirements for what we need to investigate to question. and this span of countries includes a lot

00:19:42 - and then a handful of, emerging economies. And so the combination of the advanced economies, which are kind of on a downwards part of the manufacturing hump, and then the emerging economies that are on the increasing part of the hump, we can kind of trace out along, we can kind of trace out a path with respect to a comfort capita if we keep track of where those countries, lie in terms of economic development. and so our, our strategy is empirically we've,

00:20:12 - and post 90 periods. And, and find as Roderick also found that the, that average path is lower in the post 90 period. And then we add to that the, this observation of, of increasing dispersion across countries. in the manufacturing value added share, we use the regression to, to trace out these average paths. so in the regression, we would take on the left hand side, the manufacturing

Modeling Manufacturing Shares and Income Dynamics

00:20:48 - And then on the right hand side, we would have that country's income per capita. and we would include, allow for a nonlinearity and have, you know, a quadratic, specification there. importantly we would also include country fixed effects. the reason is just, some countries are perhaps more inclined to have large manufacturing shares compared to others, say, because of endowments.

00:21:16 - So we wanna cleanse the calculation from those particular, you know, kind of country characteristics and focus on over time what happens within a country as its income per capita changes, is what we're trying to get at using that regression. And, and then what our results, kind of point out is that, you know, that, in the post 90 period, the change over time would've, been less than in the pre 90 period. We take that on board, and then we put it into this quantitative,

00:21:47 - And we, we ask if we, if we take the observed volume of trade, that we have in the data, right? And we construct measures of trade frictions and things like this in order for the model to reproduce the trade flows that we observe. And if we measure sectoral productivity, in each country and each year directly from the data, we take that as an input into the model. So we kind of make the model match

Impact of Trade and Technology on De-Industrialization

00:22:21 - and the underlying trade patterns that we observed. And then we ask, when we put these two things in, can we reproduce the rodricks fact, right? Or this premature de-industrialization phenomenon and the industry polarization. And we find that the model could generate a majority of it, it could generate all of the industry polarization because that's directly related to trade. It's kind of not surprising. And we've kind of cooked up the model to be consistent

00:22:49 - It's less trivial that the model would reproduce the, the hump shape, and in particular the fact that the hump was higher in the pre 90 period than the post 90 period. So the model generates that. And then we follow up by asking, well, how much did international trade contribute to these two things? And how much did technological change contribute to these things? We wanna kind of flip the switch on or off of either one of those two forces that we fed in. And what we find, this is kind of the, the,

00:23:22 - the underlying technological change contributed about 60% to the de-industrialization as measured by the change in the pre 90 hump and the post 90 hump. So about two thirds of that almost is due to technological change. it generated essentially enough technological change, didn't on its own contribute much to industry polarization. The flip side of that is industry polarization. This increasing concentration of production is due mostly

Interaction of Trade and Technological Change in Deindustrialization

00:23:54 - Okay? Now, we find that on its own declining trade costs don't explain much of the premature deindustrialization, but I think the most interesting takeaway that we found is that international trade does matter, but only in the presence of the particular technological change. and so let me just take a minute to explain that.

00:24:26 - that manufacturing productivity growth tends to exceed services productivity growth that's gonna induce a fall on the relative price of manufacturing. Okay? So imagine you're experiencing that on your own as an isolated economy, and then you start trading with me, and I'm also experiencing the same kind of underlying technological change. And then we start trading with each other, right? What prices do you pay for? The things that I produce is gonna be a function

00:24:56 - So international trade then serves as a mechanism to propagate this technological change. And so when you combine it international trade with the, with the technological change, now you can account for almost all of the, observed, deindustrialization. So it's a little bit of a subtle effect of how trade works. It works as an amplification mechanism, but it, on its own, it would've through the lens

Policy Implications of Manufacturing Trends

00:25:27 - to deindustrialization, but it doesn't mean it wasn't important. So that's kind of the main, that's the main punchline of the paper policy implications, I think, I think a lot is here, especially for, countries, maybe at lower levels of, in income per capita that are focusing on industrial policy, these debates almost always entail, kind of, can we follow the role of what South Korea did?

00:26:00 - So how do we grow manufacturing? So there's, there seems to be this sense in which growing manufacturing is, is the goal to begin with. And so I think what our, what our results suggest is that, you know, even in the absence of any inefficiencies or any kind of distortions, if you're kind of entering the global stage at a later date, just the underlying nature of the world

00:26:32 - to put lower pressure on, in some sense, result in you having a lower manufacturing share than you otherwise would. And so that on its own is not necessarily a bad thing. And again, I just wanna kind of reiterate the way we measure shares. These are measured at price. It doesn't mean that the quantity of manufacturing is falling. It means that the price times quantity is falling. So you can have a low share because the price is low and the relative price could be low because you have a very productive technology there.

Exploring the Interplay between Trade and Technological Advancement

00:27:05 - to keep in mind in, in these policy debates of industrial policy, in particular. And now the other, the other part of your question was about where do we go from here? Well, I think very interesting path forward is to think about how trade itself affects the technologies that you use. The calculations that we did were kind of saying we kind of take technological changes given we didn't really ask how it got there, right?

00:27:34 - manufacturing productivity growth? And so you might be imagining, for example, you know, well, how should I think about a country like China? Would it have the technologies that it has today if it weren't integrated in the global economy, be a trade for the last couple of decades? You know, I think most people would, their prior would be probably not okay. And so that's not something that we've taken on board into our analysis,

00:28:08 - that would need to be added to get a better picture of what's really going on.

Faculti AI can make mistakes

Research

Policy

Analysis