Economy: a Network

Yüklə 1,97 Mb.
ölçüsü1,97 Mb.
  1   2   3   4   5   6   7

Interconnections of the Industries in the Economy: a Network Approach

Evgeniya Goryacheva a,
a Economics Discipline Group, Business School, University of Technology Sydney, Australia

This paper uses a network approach to investigate the interrelation be- tween industries in the economy. I consider the economy as a network of the industries connected with each other through the product flows. I pro- pose dynamic directed weighted network formation model. According to the model industry’s position in a network and productivity level affect its probability of a new link adoption. I apply a empirical methodology to the data obtained from the input-output tables in order to identify the key factors affecting the changes of the technological network.
Keywords: Network; Input-Output Analysis; Intersectoral Linkages.


  1. Introduction

I use a network approach to describe U.S. economy. I build a network using U.S. input-output tables. Observing the U.S. input-output tables in different years, we can conclude that this network is not stable. The product flows and inputs’ shares are changing over time. New links are created between industries, some old links disappear, and the weights of the links are also changing over time. The main research question of this paper is how the industries’ network changes over time. Which factors affect the network formation process?

I propose the rules of the industries’ network evolution that have economic interpretation and test their validity empirically. One of these rules is that indus- tries’ positions in a network are closely related to economies of scale. Industries with high centrality have many consumers. They have cost advantages compared to other industries, because their fixed costs are spread out over more product flows. Lower costs lead to the lower prices of the outputs, this attracts new con- sumers and also increases existing product flows. As a result, these industries become even more influential over time.

Obtained empirical results support this idea. The U.S. panel data estima- tion shows that the industry’s eigenvector centrality affects its probability of a new link creation and its share in intermediate inputs of other connected industries.

I propose the hypothesis that the industry’s productivity influences the industry’s position in the network. The network evolves in the following way. More productive industries have low costs and as a result low prices of their outputs. Many other industries adopt these industries as suppliers for this reason. At the same time industries increase the product flows from the productive industries because of the low costs. This makes productive industries more influential over time. I test this hypothesis using empirical data and obtained results show that more productive industries become more central over time.

The growth of the industry’s productivity also increases the industry’s indegree in the period between 1972 and 1992. Because of the competition, indus- tries try to improve their products. As a result innovations lead to the changes in their production functions and to the adoption of new suppliers. For example, at some point the car manufacturing industry started using plastic instead of steel for some car parts. The car manufacturing adopted plastic manufacturing as its new supplier because of the technological innovation.

I propose a dynamic weighted network formation model which is a mod- ification of the model introduced in Barrat, Barth´elemy, and Vespignani (2004a). This model captures above evolution rules and leads to the power-law distribution

of the industries’ weighted outdegree.

It is important to investigate the process of the network formation, be- cause the structure of the network affects stability of the whole system. As was shown in Acemoglu, Carvalho, Ozdaglar, and Tahbaz-Salehi (2012), weighted out- degree of the U.S. industries have a power-law distribution.12 This means that even small shocks can lead to the large system volatility.

The structure of the network can help predict how shocks will propagate in the economy. Many models try to predict the shocks’ propagation in the network. It is worth mentioning that the input-output analysis was popular for investigating effects of different kinds of shocks on the economy in the previous century. The main advantage of the input-output analysis is its simplicity. It is a clear mathematical procedure to calculate multipliers and then to use them to get an impact of the change in one industry on the economic activity of another indus- try. But at the same time there is a limitation of this approach, because it is valid only for short-term prediction. Because it makes prediction under an assumption that the network is static, which is not true in the long run. Many models that use a network approach to describe shocks’ propagation in the economy also assume

that industries’ network is fixed.

In reality, the structure of the network is changing over time, and if these changes are large the models with network stability assumption can not be used for long-term predictions. In this case it is necessary to take into account the evolution of the network’s structure; otherwise, the prediction of the shocks’ influence on the economy would be incorrect. Therefore I investigate how industries’ network is changing over time and what causes these changes.

Using data of the U.S. input-output tables I build a network, where nodes are industries and links represent product flows between them. Each industry uses a part of other industries’ outputs as inputs for its own production. The weights of the links represent the size of the product flows.

For example, the aircraft manufacturing industry uses inputs from differ- ent industries: rubber, iron and steel, electronics, metalworking machinery, space vehicles, motor vehicle parts manufacturing industries. At the same time some of these industries such as electronics, metalworking machinery, space vehicles and motor vehicle parts manufacturing industries also use part of the output produced by the aircraft manufacturing as inputs in their production. This interconnection

1The network has a power-law weighted outdegree distribution if P (k) ckβ , where P (k)

- the empirical counter-cumulative distribution function, k - a weighted outdegree, c, β - the

constants and β > 1.

2Industry’s weighted outdegree is a sum of the industry’s shares in intermediate inputs of other industries.

Figure 1: Industries-suppliers and industries-consumers of the aircraft manufac- turing industry

is illustrated in Figure 1.

Figure 2 illustrates the U.S. economy in 2014 as a network. This directed graph includes the most important links (with large product flows) between ag- gregated sectors. The size of the node reflects the importance of the industry’s position in the network. The sectors with a higher eigenvector centrality have a larger size of the nodes that represent them.

In reality this network is not fixed over time. Industries start to use inputs from new industries-suppliers, while some industries may lose their industries- consumers. These changes can be described as new links’ adoption and links’ destruction in the network. At the same time the size of the product flows between industries is also changing, which again can be described as a change of the links’ weights in the network. I describe the rules of the industries’ network evolution and their economic interpretation.

Figure 2: Aggregated Input-Output Network of the U.S. industries in 2014

Yüklə 1,97 Mb.

Dostları ilə paylaş:
  1   2   3   4   5   6   7

Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur © 2020
rəhbərliyinə müraciət

    Ana səhifə