20
Work in Progress NA-242 Globalization Study Network Analysis Concepts for Technology Diffusion Through Trade Tom Wood - PNNL DOE-NA-242 Multilateral Team Meeting

Tom Wood - PNNL DOE-NA-242 Multilateral Team Meeting

  • Upload
    ismail

  • View
    46

  • Download
    1

Embed Size (px)

DESCRIPTION

Work in Progress NA-242 Globalization Study Network Analysis Concepts for Technology Diffusion Through Trade. Tom Wood - PNNL DOE-NA-242 Multilateral Team Meeting. Globalization Project. The Issues. - PowerPoint PPT Presentation

Citation preview

Page 1: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Work in ProgressNA-242 Globalization Study

Network Analysis Concepts for Technology Diffusion Through Trade

Tom Wood - PNNL

DOE-NA-242 Multilateral Team Meeting

Page 2: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Globalization Project

P h ase 1Tota l trad e

P h ase2S en s it ive Trad e

Trad e F lows Task In d u s try O u tlook Task O th er P R A P p rod cu c ts

G lob a liza tion P ro jec t

Page 3: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

The Issues

• Which countries represent high risks for secondary transfers (transshipment, re-export, reverse engineering, etc.) of sensitive technology?

• How is this dependent on [1] country of origin, [2] ultimate destination, [3] nature of technology?

• What is globalization doing to the technology transfer network?

• What sort of control regime can effectively address this problem?

Page 4: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Premise

• Network models best represent the relevant features of diffusion of nuclear and other sensitive technology– Large number of actors

– Immense number of transactions

– Well-defined but mutable “structure” in network

– Barriers, partitions, and clusters

– Scalable – applicable to entities at individual, firm, industry, and country scales

– Can represent transmission of hardware, specs, information, etc.

Page 5: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Latent Distance MeasuresBackground and Relevance

• Concept of “latent” (unobservable but estimable) distances and spaces is emerging in social network analysis literature (Hoff, Raferty, and Handcock 2001)

• Published approaches (that I have seen) assume a “normed” space – i.e. with a metric measure of distance.

• Extension of this concept to non-metric spaces captures the essence of our technology diffusion problem.

• We are using this as a descriptive (not an inferential) framework for now.

Page 6: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Network Identification for Trade Flows Problem – iteration 1

Select DataFit or Select

Relational Model

Screen for Relevance

Code to HTS or other system

Sensitive Technology Lists

Set Constraints

Critical Examination

Transactions

UN-OCED

DOTS

transaction

Sectoral

Calculate Distances

Solve Shortest Path Problem

Structure and

Parse Network

Page 7: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Figure 1. Scheme for Preliminary Calculations

Page 8: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Gravity Models and Economic Distances

Gravity Models are the standard for predicting bi-lateral trade flows.The general form is:

Tij = [ (Ai * Aj ) ]/ dij

or dij = [ (Ai * Aj)/Tij]

1/

 

Where A is some measure of activity (“economic mass”) and d is distance. The term gravity model is from the analogy to Newton’s formula for gravitational force.

Typically, these models are fit to minimize error in an estimate of Tij.

We take T and A as given to calculate dij.

In this context, dij represents more than physical distance – ie “effective” or “implicit” or “latent” economic distance.

Page 9: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Internal Average SCL Economic Distance - Gravity Model

1000

11000

21000

31000

41000

51000

61000

71000

81000

91000

Russia

n Fed

erat

ion

Tajikis

tan

Turkm

enist

an

Uzbek

istan

Kazak

hsta

n

Ukrai

ne

Belar

us

Syrian

Ara

b Rep

ublic

Azerb

aijan

Isra

el

China

, P.R

.

Geo

rgia

Sudan

Pakist

an

Mol

dova

Kyrgy

z Rep

ublic

Armen

iaIn

dia

China

, Hon

g Kon

g

Iran,

Islam

ic Rep.

Alger

ia

Figure 4. Internal SCL Economic Distances

Page 10: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Avg Economic Distance - NSG to SCL Countries - Gravity Model

1000

11000

21000

31000

41000

51000

61000

71000

81000

91000

Figure 5. NSG to SCL Distances – Sorted by NSG

Page 11: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Figure 6. Direct vs. Indirect Distances

Indirect ED <= Direct ED

Direct ED < Indirect ED

Page 12: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Histogram – Number of Intermediate Countries in Shortest Paths from NSG to SCL case 138.b14

Page 13: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

“Joining Tree” for World Economy – Based on ED relationships

Page 14: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Shortest Paths for Four Selected NSG to SCL Pairs

Page 15: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Sensitivity of Shortest Paths to Model Parameters

Page 16: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Shortest Path AnalysisSome Preliminary Observations

• In all cases, the set of intermediate countries on the shortest path set is compact, and moderately to highly skewed. Typically, a dozen countries account for about 70% of the total instances.

• In most cases, the set of intermediate countries is highly disjoint from both the NSG and the SCL.

• In many cases, scale effects appear to give implausible shortest paths. Very small countries with a high degree of economic dependence (e.g. islands) have very short distances both among themselves and to more developed counties (particularly their former colonial masters).

Page 17: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

GEOGRAPHIC AND ECONOMIC DISTANCE ELASTICITIES BY SECTOR

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

-1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0

Geographic Distance Regression Coefficient

Ec

on

om

ic D

ista

nc

e R

eg

res

sio

n C

oe

ffic

ien

t

Dairy

Pharmaceuticals

precision, optical inst.

Scientific & Medical Equip.

Chemicals, petrol refining

Industrial chems

Radio, TV

Aircraft

Non-ferrousMetals

Paper

Automotive

Plastics

Tubes, Pipes

Cement

Shipbuilding

Soaps, detergents

Wire

Tires

Agriculturalmachines

MeatYarns

Lime, Clay products

Stone Products

Knitting

Iron & Steel Foundries

Tobacco

Sector Namexxx

Sector Geographic/Economic Distance Coefficients

Page 18: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Next Steps

• Build the data set for the “Sensitive Technology sector” based on LANL coding of lists to HTS.

– Which lists do we wish to use? (all, all nuclear, etc.)– Parse country list to exclude very small and or backward countries –

• Unless of special interest – let me know

• Characterize the R&D intensity and concentration of this sector – (Pick the economic model on which to base the distance measures)

• Calculate measures, solve and analyze shortest paths • MLT workshop to explore policy implications

• .

Page 19: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Some Possible Policy Ideas

• “Special Status” for selected countries posing high risk for re-transfer

– Unilaterally

– Multilaterally

• “Export Substitution” – A policy of identification and incentivized transfer of technology which has a high substitutability in legitimate uses and low substitutability in weapons uses.

Page 20: Tom Wood  - PNNL DOE-NA-242 Multilateral Team Meeting

Selected Literature

Morstein, Jennifer Hunt and Wayne D. Perry. 2000. Commercial Nuclear Trading Networks as Indicators of Nuclear Weapons Intentions. The Nonproliferation Review/Fall-Winter 2000, Monterey, California.  Conley, Timothy G. and Ethan Ligon. Economic Distance and Cross-Country Spillovers. Unpublished Monograph, University of Chicago Department of Economics. 2001.  Van Houtum, Henk. Borders, Distances, and Spaces – A typology of borders in terms of distances. Proceedings of European Regional Science Association. Dublin 1999.  Boisso, dale and Michael Ferrantino. 1997 Economic Distance, Cultural Distance, and Openness in International Trade: Empirical Puzzles. Journal of Economic Integration.  Sen, Ashish and Tony Smith. 2002. Gravity models of Spatial Interaction Behavior. Springer-Verlag. Berlin. [LCN 95035034]  Haynes, Kingsly E. Gravity and Spatial Interaction Models. Sage Publications. 1984.[LCCN 84050799]

Handcock, M.S. 2002. Degeneracy and Inference for Social Network Models. Working paper, Center for Statistics and the Social Sciences. University of Washington [csss.washington.edu] Priebe, Volker. 2001. Average Case Complexity of Shortest Path Problems. Dissertation. University of Saarlandes