Using panel VAR to analyze international knowledge spillovers

• Published date: 01 November 2019
• Author: Norman Sedgley,Nune Hovhannisyan
• Date: 01 November 2019
• DOI: http://doi.org/10.1111/roie.12438

Rev Int Econ. 2019;27:1633–1660. wileyonlinelibrary.com/journal/roie

|

1633

© 2019 John Wiley & Sons Ltd

Received: 29 May 2018

|

Revised: 10 June 2019

|

Accepted: 10 July 2019

DOI: 10.1111/roie.12438

ORIGINAL ARTICLE

Using panel VAR to analyze international knowledge

spillovers

NuneHovhannisyan

|

NormanSedgley

Loyola University Maryland, Baltimore,

Maryland

Correspondence

Nune Hovhannisyan, Department of

Economics, Sellinger School of Business

and Management, Loyola University

Maryland, 4501 North Charles Street,

Baltimore, MD 21210.

Email: nhovhannisyan@loyola.edu

Abstract

Technology diffusion often plays a critical role in models

of trade and economic growth. Most existing empirical tests

for international technology spillovers suggest some role

for spillovers in explaining productivity growth. It has been

relatively difficult, however, to identify separate roles for

the direct and indirect channels of knowledge spillovers.

The influence of these channels is often confounded owing

to the focus on total‐factor productivity (TFP) and R&D

spending within a cross‐section or panel data setting. This

paper employs an alternative methodology to investigate the

role of direct knowledge spillovers. Using citation‐weighted

domestic patents, citation‐weighted foreign patents and

value added for 14 U.S. manufacturing industries over

the period 1977 to 2004 a panel VAR methodology is em-

ployed to investigate the dynamic role of direct and indirect

knowledge spillovers. Evidence for the role of the direct

knowledge spillovers channel is found—an increase in cita-

tion‐weighted patents abroad directly increases the measure

of domestic citation‐weighted patents, after accounting for

the influence of productivity/value added. The role of for-

eign innovative activity, however, is small relative to the

role of U.S. innovative activity in explaining the dynamics

of industry value added.

JEL CLASSIFICATION

O33; F43; C32; C33; L6

1634

|

HOVHANNISYAN ANd SEdGLEY

1

|

INTRODUCTION

International technology spillovers play a potentially important role in economic growth and produc-

tivity/income convergence. Technology diffusion often plays a critical role in theoretical models of

international trade and economic growth and economists’ understanding of the relationship between

trade, knowledge creation, and growth. The literature recognizes roles for both direct channels and

indirect channels of technological diffusion. Direct channels work through the direct benefits to re-

searchers in one economy from knowledge generated abroad. The role of this channel is diminished to

the degree that spillovers are geographically localized and/or the receiving economy lacks absorptive

capacity to utilize the knowledge. Indirect channels for diffusion rely on trade, often in intermediate

goods, and foreign direct investment (FDI). It has been relatively difficult to identify separate roles for

the direct and indirect channels of knowledge spillovers. The influence of these channels is often con-

founded owing to the literature's focus on total‐factor productivity (TFP) and R&D spending within a

cross‐section or panel data setting. This paper employs an alternative methodology to identify the role

of direct knowledge spillovers.

The evidence provided in this paper is of significant importance in formulating an understand-

ing of trade, innovation and growth. Consider the extremes encountered in the literature. Eaton and

Kortum (2001) provide a theoretical framework where technology transfer is limited by trade and

therefore geographic distance. An implication of Eaton and Kortum is that foreign R&D and/or pat-

ents are reflected in TFP, but since the effect is indirect, there would be no impact on domestic R&D

and/or patents. This view promotes the indirect channel of spillovers. In contrast, Acemoglu, Aghion,

and Zilibotti (2006), Ertur and Koch (2011), and Howitt and Mayer‐Foulkes (2005), promote a view

that completely abstracts from trade and explains convergence in income growth rates via a direct

spillover of knowledge across international borders, thus promoting the importance of the direct chan-

nel for spillovers. In this view convergence does not depend on openness or trade, but instead simply

on the existence of innovator/entrepreneurs who actively engage in innovation, thus benefiting from

the public goods nature of knowledge created internationally.

An empirical literature has developed and provides important insights into the role of knowl-

edge/R&D spillovers. While each of the existing studies has relative strengths and weaknesses, they

are each specified in a manner that makes it difficult to discern the role of the direct channel for spill-

overs. Despite the inability to identify the channels through which spillovers operate, this literature

does provide evidence of a role for knowledge spillovers in general, conditional on absorptive capacity

(in addition to the papers above see, e.g., Belderbos, Ito, & Wakasugi, 2008; Coe & Helpman, 1995;

Coe, Helpman, & Hoffmaister, 2009; Hu & Jefferson, 2009; Kugler, 2006; Lee, 2006; Mancusi, 2008;

Yang, 2003). Furthermore, it does seem clear that knowledge is geographically localized to a signif-

icant degree. Keller (2002), for example, estimates the impact of domestic and foreign R&D on TFP.

The effectiveness of foreign R&D negatively depends on the bilateral geographic distance between

the source and the recipient country. While this suggests that the role of direct knowledge spillovers

in innovation and growth is limited by geography, it by no means rules out or subordinates a direct

channel for knowledge transfer. On this issue Branstetter (2001) finds no evidence of international

knowledge spillovers and argues that the explanation is the geographic concentration of knowledge.

In contrast, using a dynamic OLS (DOLS) methodology, Lee (2006) reports significant evidence

of direct knowledge spillovers. Eberhardt, Helmers, and Strauss (2013) find evidence of knowledge

spillovers and unobserved cross‐sectional dependencies and argue that ignoring spillovers can lead to

inflated estimates of private returns to R&D.

The existing literature, summarized below, is largely centered on methodologies that make use of

panel data with fixed effects or time series models that do not take advantage of the potential benefits