theory and empirics of capability accumulation

20
Research Policy 50 (2021) 104258 Available online 12 April 2021 0048-7333/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Contents lists available at ScienceDirect Research Policy journal homepage: www.elsevier.com/locate/respol Theory and empirics of capability accumulation: Implications for macroeconomic modeling Matthias Aistleitner a , Claudius Gräbner a,b,c ,, Anna Hornykewycz a a Institute for Comprehensive Analysis of the Economy (ICAE), Johannes Kepler University Linz, Austria b Institute for Socio-Economics, University of Duisburg–Essen, Germany c ZOE. Institute for Future-Fit Economies, Germany ARTICLE INFO JEL classification: B41 B52 D2 M2 O3 O33 Keywords: Capability accumulation Complexity Economic development Innovation Technological change Agent-based modeling Endogenous growth Knowledge accumulation and learning ABSTRACT The accumulation of new technological capabilities is of high empirical relevance, both for the development of countries and the business success of firms. This paper aims to delineate strategies for the consideration of processes of capability accumulation in comprehensive macroeconomic models. It comprises a methodological discussion about the preconditions for macroeconomic models to consider insights from the applied business literature, as well as an extensive review of the literature specialized on capability accumulation on the firm and aggregated level. In doing so, it summarizes evidence on various determinants and mechanisms of capability accumulation and aligns them with the current representation of capability accumulation in macroeconomic models. Based on these results, it offers some suggestions on how macroeconomists may integrate these determinants derived from the specialized literature into their models. 1. Introduction Numerous empirical investigations stress that the accumulation of technological capabilities is essential for the growth and development of countries and regions, as well as for the business success of firms. Concerning the national level, for instance, Hidalgo and Hausmann (2009) show ‘‘that countries tend to approach the level of income associated with the capability set available in them’’ (p. 10570). And on the firm level technological capabilities were found to be ‘‘[. . . ] central to [a firm’s] identity, its strategies, and its potential for suc- cess’’ (Aharonson and Schilling, 2016, p. 81). Despite the fact, however, that scholars from various disciplines agree on the relevance of capa- bilities, and have produced various empirical insights on the factors driving capability accumulation (CA, hereafter), many of these insights Supported by funds of the Österreichische Nationalbank (Austrian Central Bank, Anniversary Fund, project number: 18144). MA and CG acknowledge funding by the Austrian Science Fund (FWF) under grant number ZK 60-G27. Open access funding by the FWF is acknowledged by all authors. The authors also gratefully acknowledge support of the Economic and Social Research Council (UK) under grant reference number ES/R00787X/1. We also acknowledge the role of the National Institute of Economic and Social Research (NIESR) in hosting the Rebuilding Macroeconomics Network. Corresponding author at: Institute for Socio-Economics, University of Duisburg–Essen, Germany. E-mail addresses: [email protected] (M. Aistleitner), [email protected] (C. Gräbner), [email protected] (A. Hornykewycz). URL: https://claudius-graebner.com (C. Gräbner). from the applied literature specialized on CA have not made their way into macroecononmic models. This leaves macroeconomic researchers with both a challenge and an opportunity: on one hand, there are empirical insights available that can inform the construction of new macroeconomic models that promise a deeper understanding of economic growth and development. On the other hand, transferring the fine-grained empirical results on CA into a comprehensive macroeconomic framework is troublesome and often conflicts with the task of keeping the overall complexity of these models manageable. This challenge is aggravated by the fact that, in economics, there is much less concrete work on which mechanisms impact on CA under particular circumstances. The situation is different in management, organization and innovation studies, where the mech- anisms of CA have been at center stage for a long time — yielding https://doi.org/10.1016/j.respol.2021.104258 Received 11 June 2020; Received in revised form 26 March 2021; Accepted 28 March 2021

Upload: others

Post on 09-Apr-2022

4 views

Category:

Documents


0 download

TRANSCRIPT

Research Policy 50 (2021) 104258

A0

Contents lists available at ScienceDirect

Research Policy

journal homepage: www.elsevier.com/locate/respol

Theory and empirics of capability accumulation: Implications formacroeconomic modeling✩

Matthias Aistleitner a, Claudius Gräbner a,b,c,∗, Anna Hornykewycz a

a Institute for Comprehensive Analysis of the Economy (ICAE), Johannes Kepler University Linz, Austriab Institute for Socio-Economics, University of Duisburg–Essen, Germanyc ZOE. Institute for Future-Fit Economies, Germany

A R T I C L E I N F O

JEL classification:B41B52D2M2O3O33

Keywords:Capability accumulationComplexityEconomic developmentInnovationTechnological changeAgent-based modelingEndogenous growthKnowledge accumulation and learning

A B S T R A C T

The accumulation of new technological capabilities is of high empirical relevance, both for the developmentof countries and the business success of firms. This paper aims to delineate strategies for the consideration ofprocesses of capability accumulation in comprehensive macroeconomic models. It comprises a methodologicaldiscussion about the preconditions for macroeconomic models to consider insights from the applied businessliterature, as well as an extensive review of the literature specialized on capability accumulation on thefirm and aggregated level. In doing so, it summarizes evidence on various determinants and mechanismsof capability accumulation and aligns them with the current representation of capability accumulation inmacroeconomic models. Based on these results, it offers some suggestions on how macroeconomists mayintegrate these determinants derived from the specialized literature into their models.

1. Introduction

Numerous empirical investigations stress that the accumulation oftechnological capabilities is essential for the growth and developmentof countries and regions, as well as for the business success of firms.Concerning the national level, for instance, Hidalgo and Hausmann(2009) show ‘‘that countries tend to approach the level of incomeassociated with the capability set available in them’’ (p. 10570). Andon the firm level technological capabilities were found to be ‘‘[. . . ]central to [a firm’s] identity, its strategies, and its potential for suc-cess’’ (Aharonson and Schilling, 2016, p. 81). Despite the fact, however,that scholars from various disciplines agree on the relevance of capa-bilities, and have produced various empirical insights on the factorsdriving capability accumulation (CA, hereafter), many of these insights

✩ Supported by funds of the Österreichische Nationalbank (Austrian Central Bank, Anniversary Fund, project number: 18144). MA and CG acknowledge fundingby the Austrian Science Fund (FWF) under grant number ZK 60-G27. Open access funding by the FWF is acknowledged by all authors. The authors also gratefullyacknowledge support of the Economic and Social Research Council (UK) under grant reference number ES/R00787X/1. We also acknowledge the role of theNational Institute of Economic and Social Research (NIESR) in hosting the Rebuilding Macroeconomics Network.∗ Corresponding author at: Institute for Socio-Economics, University of Duisburg–Essen, Germany.

E-mail addresses: [email protected] (M. Aistleitner), [email protected] (C. Gräbner), [email protected] (A. Hornykewycz).URL: https://claudius-graebner.com (C. Gräbner).

from the applied literature specialized on CA have not made their wayinto macroecononmic models.

This leaves macroeconomic researchers with both a challenge andan opportunity: on one hand, there are empirical insights availablethat can inform the construction of new macroeconomic models thatpromise a deeper understanding of economic growth and development.On the other hand, transferring the fine-grained empirical results onCA into a comprehensive macroeconomic framework is troublesomeand often conflicts with the task of keeping the overall complexity ofthese models manageable. This challenge is aggravated by the fact that,in economics, there is much less concrete work on which mechanismsimpact on CA under particular circumstances. The situation is differentin management, organization and innovation studies, where the mech-anisms of CA have been at center stage for a long time — yielding

vailable online 12 April 2021048-7333/© 2021 The Authors. Published by Elsevier B.V. This is an open access a

https://doi.org/10.1016/j.respol.2021.104258Received 11 June 2020; Received in revised form 26 March 2021; Accepted 28 Ma

rticle under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

rch 2021

Research Policy 50 (2021) 104258C. Gräbner et al.

comparatively more detailed insights. It, thus, seems to be an obviousnext step to facilitate an interdisciplinary discourse, which is attractivenot only for macroeconomists, who would benefit from insights aboutan essential driver of economic dynamics. It is attractive also forinnovation scholars, who could see their results (that mostly concernthe firm-level) being placed into a more comprehensive analysis of theeconomy as a whole, and who could benefit from the economic insightson macroeconomic phenomena that feed back in important ways onprocesses of CA.

The aim of this paper is to facilitate such discourse and to addressthe challenge of considering CA within macroeconomic models. Indoing so, however, a methodological problem arises: the integration ofdifferent perspectives on capabilities and their accumulation in macroe-conomics and innovation studies. These disciplines do not necessarilyagree on what the main repositories of capabilities are, how theywork, and how they evolve. This meta-theoretical incompatibility is oneimportant reason for the wedge between macroeconomics and the otherdisciplines concerned with CA. For a fruitful integration of the insightsfrom the different disciplines, however, this methodological challengeneeds to be addressed by adopting a common framework of analysis.

Given the extensive empirical findings on the determinants of CA,such a framework must depart in some important ways from thecommon standards of macroeconomic analysis, but orient itself more onthe work in the tradition of Nelson and Winter (1982), who have shownhow economic analysis can be aligned with the business, managementand innovation literature in a more complementary and constructiveway. Once such a framework is adopted, progress on understandingthe systemic determinants and implications of capabilities and theiraccumulation will become more likely.

To facilitate this progress, the present paper is meant to identify themost influential insights from the empirical literature on CA, to describehow CA has been operationalized in specialized models that have theirmain focus on CA, and to derive suggestions on how these insightscan be most constructively integrated into macroeconomic models. Thiswould allow to develop a more comprehensive view on how CA shapesand is being shaped by the dynamics of economies as a whole.

Thus, the paper proceeds follows: Section 2 clarifies the meaningof the term ‘capabilities’, explicates the relevance of different theoriesof the firm as a basis for their analysis, and delineates a frameworkfor the comprehensive integration of CA into macroeconomic anal-ysis. Section 3 provides an overview on the implementation of CAin macroeconomic models to date. The current state of the empiricalliterature on CA, as well as its operationalization in specialized modelsis then summarized in Section 4. Section 5 synthesizes the findings anddiscusses implications for macroeconomic modeling. Finally, Section 6summarizes and concludes the paper.

2. On the theory of capabilities

The term ‘capabilities’ is potentially ambiguous. This section ismeant to distinguish its different interpretations and to clarify its usethroughout this paper. It will be argued that (1) it is justified to seecapabilities as being first and foremost properties of firms (and notindividuals or countries), and (2) it is necessary to consider the factorsgoverning the accumulation of capabilities on both the level of firmsand countries and, thereby, to conceptualize the overall process of CAwithin a framework of ‘systemism’ (e.g. Gräbner and Kapeller, 2015).

2.1. Firms as the locus of capabilities

Since capabilities are related to ‘knowledge’ or ‘know-how’, onecould simply treat the terms as synonyms and operationalize capa-bilities via traditional measures of human capital. Capabilities wouldthen be located on the level of the individual and the question of

2

how a firm or a country could improve upon its capabilities would

boil down to the question of how its employees or citizens, respec-tively, could accumulate more human capital. Such an approach tocapabilities is implicit in many approaches that take a radical indi-vidualistic starting point, such as the neoclassical theory of the firm,which is also at the center of the endogenous growth theory (seeSection 3. These theories focus on prices as the main devices forcoordination, and the idea of profit maximizing firms governed byutility maximizing managers. This makes their application to real firms,which are faced by the existence of fundamental uncertainty and theinfeasibility of optimization, as well as the absence of price-basedcoordination for intra-firm allocation problems, problematic (Teece,2017). Such approaches also tend to assume firms to be homogeneous.If firm heterogeneity is allowed for – as in recent general equilibriummodels in the spirit of Melitz (2003), such as Costinot et al. (2020)– this comes down to simply assuming heterogeneity in terms of sizeand productivity. There is, however, considerable evidence that thisis not what matters, neither for the continuous success of the firm,nor for growth and development on the country level (e.g. Sutton,2012; Teece, 2017). Therefore, the relevance of the traditional theoryof the firm to scholars outside economics, especially management,remains disputed (e.g. Teece, 2017). Critiques argue that the theoryabstracts from much that is of high relevance for practitioners andmanagement scholars, such as direct competition between firms, thedistributional and organizational choices managers have to make, aswell as the organizational routines firms use when confronted withKnightian uncertainty. This impairs the explanation of how exactlyfirms actually accumulate capabilities to innovate, how they developthe competitive advantage against each other and, thereby, where themassive heterogeneity of firm performance originates (e.g. Dosi andRoventini, 2019; Teece, 2017). Alternative approaches, mainly frommanagement literature, distinguish themselves by placing ‘capabilities’at the center of their attention, and it is this stream of literature wherethe main inspiration for the definition of capabilities used in the presentpaper comes from.

Among the most prominent of these accounts are the ‘resource-based view of the firm’ (RBV, see Barney, 1991) and the ‘dynamiccapabilities view of the firm’ (see Teece et al., 1997; Teece, 2017).They can be distinguished from the neoclassical view of the firm bytheir more dynamic nature and the fact that the ultimate carriers ofcapabilities are, in both cases, firms. The central idea of the RBVis that the competitive success of a firm is ultimately determinedby its internal resources. Resources encompass all assets (tangible orintangible) that a firm effectively controls, including its equipmentand the human capital of its employees. Capabilities, then, are theorganizational abilities to use these resources to achieve a particulargoal (Helfat and Peteraf, 2003). Teece et al. (1997) introduced a distinc-tion between ‘ordinary’ and ‘dynamic’ capabilities, with the latter being‘higher order capabilities’ that allow organizations to reconfigure theirordinary capabilities and to adapt to a changing market environment.The idea of dynamic capabilities adds an element of dynamics, yet someargue that ordinary capabilities may also change without the use ofdynamic capabilities (see, e.g., Helfat and Peteraf, 2003). In any case,ordinary and dynamic capabilities are considered to be collections ofroutines (Helfat and Peteraf, 2003), which suggests the ultimate carrierof capabilities to be not the individual, but the firm. This allows totrace the heterogeneity of firms to differences in the level of accu-mulated capabilities, but it contrasts with the individualistic world ofthe neoclassical approach, where capabilities are rooted exclusively inthe abilities of individuals, or get mixed into an aggregated productionfunction.

2.2. The need for a systemist view on capability accumulation

Although the locus of capabilities is the firm, one cannot under-

stand CA by looking at firms alone. Rather, a ‘systemist’ view, which

Research Policy 50 (2021) 104258C. Gräbner et al.

tm

dM

aahi

3

attcostfma

w‘

t

t2TKtco

a

explicitly recognizes the interplay of different ontological levels, is nec-essary (see Bunge, 2004; Gräbner and Kapeller, 2015).1 The systemistapproach, which is ‘‘a viewpoint or strategy for designing researchprojects’’ (Bunge, 2004, p. 191), starts from the ontological premisethat ‘‘everything in the universe is [...] a system or a component ofone’’ (Bunge, 2004, p. 190). To describe and explain a system, one hasto explicate its parts, the relations among its parts, its environment, andits characteristic mechanisms (Bunge, 2004). Systems feature a ‘layered’ontology in the sense that they are made up of different levels – suchas the micro, meso, or macro – and that these levels are connected via‘bridging mechanisms’. The latter may operate from lower to higherlevels (bottom-up mechanisms) or vice versa (top-down mechanisms),which is why systemism can be considered a golden middle-ground ascompared to holism and invidividualism (Bunge, 2004, p. 190).

What are the reasons for adopting a systemist approach? First,its layered ontology with (at least) a micro, meso, and macro level,each characterized by actors, their relational structure, and particu-lar mechanisms, allows for a straightforward way to describe actors,structures and mechanisms relevant in the context of CA, and to distin-guish clearly the ontological level on which they respectively operate.Second, in contrast to the extreme views of holism or individualism,systemism explicitly allows for ‘bridging mechanisms’ linking the microto the macro and vice versa. This is decisive in the present case: onhe one hand, there is clear evidence for the relevance of bottom-upechanisms since the accumulation of capabilities by firms inter alia

determines to a significant part the wealth of a nation and, thereby,its policy space (e.g. Nelson and Winter, 1982; Hidalgo et al., 2007;Sutton, 2012). At the same time, the action space for the individualfirm – which may be owned by individuals, groups or the state – aswell as their economic and non-economic environment depends to asignificant extent on institutions set by the state and, thereby, on thecountry level (e.g. Nelson, 2008; Lee and Malerba, 2018). Thus, thereare also important top-down effects when CA is concerned.

The need to consider these different ontological levels simultane-ously has been recognized in the evolutionary literature, as evidenced,for instance, by the work on national innovation systems (Nelson, 1993;Lundvall, 1995; Freeman, 1995, hereafter NIS), which is concernedwith identifying those institutional settings that enable the ‘‘production,iffusion, and use of new and economically useful knowledge’’ (Lee andalerba, 2018, p. 188).

In all, a systemist view on CA does justice to the existing literaturend represents an intuitive way to disentangle the various factorsnd mechanisms determining CA on distinct ontological levels. It alsoighlights its important role for macroeconomic dynamics and, thereby,ts place in macroeconomic models.

. CA in macroeconomic models

The challenge for macroeconomists is to take into account CAs an empirically relevant process, while at the same time keepinghe overall complexity of their models manageable. This comes withrade-offs: a more complete depiction of the CA process makes it,eteris paribus, more difficult to also include an adequate descriptionf the central bank, the government, household behavior etc.2 Thisection reviews how this challenge is currently addressed. To keephe overview concise, it is focused on three representative modelingrameworks: endogenous growth models – which stand for the economicainstream and in many ways also provide the building blocks for CGE

nd DSGE models –, Post-Keynesian, and agent-based models. Together,

1 The term ‘systemist’ is used instead of ‘system’ to avoid a confusionith Niklas Luhmann’s ‘systems theory’. The latter differs substantially from

systemism’ in the sense of Bunge (2004).2 Models that strictly focus on CA processes and do, therefore, not face this

rade-off are dealt with in Section 4.3.

3

M

they not only account for a considerable share of macroeconomicmodels concerned with topics related to CA, but they also illustratethe relevance of different methodological vantage points: as will beshown below, the fact that endogenous growth models are largelybased on the neoclassical theory of the firm means that they havedifficulties to incorporate newer insights on CA from the managementand organizational literature (see Section 2.1) . Similarly, while beingan important source of inspiration, much of the Post-Keynesian macroliterature does not incorporate the micro and meso levels on whichimportant processes of CA take place. Thus, the stage is set for the classof macroeconomic agent-based models, which are consistent with thesystemist view of CA advocated in Section 2 and, therefore, seem tobe the most promising device for a closer integration of the empiricalinsights summarized later.

3.1. Endogenous growth models

Endogenous growth models (EGMs) are the most common approachto study macroeconomic dynamics and development in the long run.Many central ideas related to the macroeconomic study of CA, suchas the idea of directed technological change (e.g. Acemoglu, 2002a) orinnovation networks (e.g. Acemoglu et al., 2016) either originated, or atleast were discussed extensively, within this framework. The historicalmotivation to develop these models was the desire to study the roleof CA for economic development endogenously, rather than treating itas an exogenous factor as common in the previously prominent neo-classical growth models: in EGMs, CA is the outcome of the investmentactivities of (profit maximizing) agents, as well as the assumed marketenvironment (for an extensive overview see, e.g., Acemoglu, 2009).They highlight different mechanisms according to which CA takes placeand operates, most of which are related to the R&D decisions of firms.Table 1 gives a short summary; for a more detailed explanation, seeAppendix A.

EGMs necessarily take a bird’s eye perspective on the process of CA:due to their commitment to the neoclassical (and new institutional)theory of the firm and the neglect of any true uncertainty, they abstractfrom most of the particular learning activities on the firm and employeelevel. This helps to focus on the macroeconomic implications of CA, yetit also shallows the concrete mechanisms underlying CA on the microlevel. What is more, this framework aggravates any consideration of theheterogeneity of firms in terms of their performance and competitive-ness, which the management literature generally puts at center stage oftheir analysis.3

3.2. Keynesian models

In contrast to EGMs, Post-Keynesian models usually do not featureoptimizing firms and do take true uncertainty into account (for anoverview see, e.g. Blecker and Setterfield, 2019; Foley et al., 2019).4The main strength of these models is to highlight how technologicalchange interacts with macroeconomic processes such as balance-of-payments constraints, stock-flow consistency, exchange rates or tradepatterns. Setterfield (2002), for instance, studies the endogenous emer-gence of technological lock-ins that constrain technological change on

3 Some newer model variants also feature agent heterogeneity, especiallyhe Heterogeneous New Keynesian Models (HANK, see Kaplan and Violante,018). However, heterogeneity here only refers to households, not firms.here are also attempts to integrate firm heterogeneity (e.g. Melitz, 2003;han and Thomas, 2008; Ottonello and Winberry, 2020), yet, not only is

his heterogeneity limited, there also are no microeconomic underpinnings inapabilities, such that the origins of heterogeneity remain vague, at best (forne of the very few exceptions see Gabaix and Landier, 2008).

4 New-Keynesian models, by contrast, are similar to EGMs but also take intoccount short-run fluctuations, rather than only long-term growth patterns.

ost of what has been said in the previous section also applies to these models.

Research Policy 50 (2021) 104258C. Gräbner et al.

lmaaotA

Table 1Implementation of the CA process in selected macroeconomic endogenous growth models. For a more detailed description of the mechanisms see Appendix A.

Building blocks of CA in macroeconomic models I: endogenous growths models (EGMs)

Consideration of CA Implication Seminal papers

Firms invest into R&D to acquire better inputs such as machines, processes andskills

Reduction in production costs Rivera-Batiz and Romer(1991)

Firms invest into R&D to develop new variants of products, which may comewith temporary monopoly rents due to consumers’ preferences for variety

Temporary monopoly rents Grossman and Helpman(1991a,b)

Firms invest into R&D, but the carriers of CA (e.g. scientists) are ‘scarce’, sosustained growth is possible only due to knowledge spillovers: CA happenscumulatively and via interpersonal knowledge diffusion

Increased productivity Romer (1990)

Firms invest into R&D and recombine existing knowledge to advanceproductivity; CA happens cumulatively and mainly via recombination of existing,rather than the creation of new knowledge as such.

Increased productivity Weitzman (1998)

Entrant firms invest into R&D and replace incumbents, mostly by offeringhigher-quality versions of existing products; in effect, CA is enforced by thedanger of creative destruction

Temporary monopoly rents(for successful innovators)

Aghion and Howitt (1992)

An important set of growth models formalizes barriers to technology adoption(such as extractive institutions) and focuses on why CA does not take place

Persistent differences in levelsof income and capabilitiesacross the world

Howitt (2000), Acemogluet al. (2007)

People learn to use particular technologies, thereby closing the possible mismatchbetween knowledge and technologies; in effect, CA happens cumulatively

Slow or no CA taking place;this implies lack of catch-upacross countries

Atkinson and Stiglitz (1969),Acemoglu and Zilibotti(2001), Acemoglu (2002b)

ctdaiagt

Soo

a(pi

the macro level, and Ribeiro et al. (2016) show how technological andstructural change is affected by exchange rate policies.

These models can be used to work out the implications of technolog-ical change, such as lock-ins, endogenous polarization, and changes inthe functional income distribution. All these topics are directly relatedto the way that technological change takes place and is determinedby CA. The actual roots for these dynamics, however, also involvethe cumulative way that firms accumulate new capabilities (see Sec-tion 4.2.2 below, but also Dosi et al. (2019) or Gräbner et al. (2020)).Yet, because the traditional focus of these models has been on themacroeconomic level, their results are derived from assumptions aboutthe aggregated level of technological change. Moreover, the evolution offirm heterogeneity, which is central to the treatment of capabilities inthe business literature, cannot play a role in these models since they donot explicitly consider the heterogeneity of firms in general.

Since the micro and macro levels are largely absent from thesemodels, they cannot serve as an immediate vehicle to study CA froma systemist angle as suggested in Section 2. From the perspectiveof the present paper, the main contribution of Keynesian models is,therefore, to serve as a central inspiration of how CA depends onaggregated factors, such as aggregate demand (see Dosi et al., 2010)and other structural macroeconomic constraints, such as balance-of-payments constraints and stock-flow consistency (Blecker and Setter-field, 2019).

3.3. Agent-based models

Agent-based models (ABMs) have grown in popularity over theprevious years due to their success in replicating a respectable numberof stylized facts on the micro, meso and macro level (for a surveysee, e.g., Dawid and Delli Gatti, 2018; Dosi and Roventini, 2019).These models consist of various types of agents – such as households,firms, banks, a national bank and a government – whose inter- andintra-class interactions are simulated directly and analyzed numericallywithout an a priori equilibrium assumption. One of the main advantagesies in the straightforward consideration of agent heterogeneity (whichatters because agent heterogeneity affects diffusion of knowledge,

nd countries and firms differ considerably in terms of their level ofccumulated capabilities Hidalgo et al., 2007; Syverson, 2011) and out-f-equilibrium dynamics (which matter in the context of innovation orechnological change, see, e.g., Dosi and Roventini, 2019). In effect,BMs have become a popular tool to study CA.

4

a

While many of the ABMs discussed below are – at least in part – in-spired by Post-Keynesian models, they model the micro and meso levelof the economy explicitly and give firms a more prominent role in theiranalysis. In contrast to EGMs, they necessarily contain explicit protocolsfor all processes that periodically, take place within the model. Onone hand, this might entail the danger of ad hoc specifications ofthese processes. On the other, the flexibility of ABMs and the requiredexplicitness might facilitate the inclusion of results of the specializedliterature on CA. In the following, we ask to what extent this inclusionhas already taken place. To this end, we discuss five different models,each representative for a particular ‘family’ of macroeconomic ABMs.5An overview is given in Table 2.

Macroeconomic ABMs usually distinguish a production sector forconsumption and one for capital goods. Dosi et al. (2019, representativefor the ‘Keynes-meets-Schumpeter’ models) and Caiani et al. (2019,representative for the ‘Benchmark’ models) locate CA processes onlyin the consumption good sector. Firms invest into R&D to increase theprobability of successful innovative or imitative activities. If the formeris successful, the firm may discover better ways to produce their finalproducts, resulting in an increase of labor productivity. In case thelatter is successful, the firm copies the productive technologies of otherfirms’ production, which may also lead to increased productivity. Thus,both processes refer to improved ways to produce the same product.

Other macroeconomic ABMs locate the CA process in both theonsumption and the capital good sector: Ciarli et al. (2018) considerwo types of CA processes: first, capital good firms invest into R&D toevelop higher quality versions of their capital good, which are morettractive to final good firms since they feature higher labor productiv-ty. Then, final good firms invest into R&D based on anticipated payoffsnd, if successful, produce new products (or, in the jargon of the paper,oods in higher quality). Thus, the model features a kind of CA similaro that in Grossman and Helpman (1991a,b) with regard to product

innovation.6

5 A ‘family’ is made up by several versions of the same underlying model.ince all instances of the respective families implement CA similarly, we focusn the most recent instance or the one that pays most attention to the topicf CA.

6 This also illustrates how the explicit event protocols in ABMs allow formore explicit and diverse representation of CA processes: in Ciarli et al.

2018), for example, firms decide on their R&D expenses based on anticipatedayoffs, whereas in Dosi et al. (2019) it is a fixed share of past profits. Thisntroduces more degrees of freedom, but also allows for a more fine-grained

lignment of the model specification with the existing empirical literature.

Research Policy 50 (2021) 104258C. Gräbner et al.

eagptOw

Table 2Implementation of the CA process in selected macroeconomic agent-based models, where it is usually discussed under the label ‘innovation’. The column ‘determinants’ only liststhe ultimate drivers of CA. We do not mention ‘policies’ when they impact CA only via fostering R&D investments.

Building blocks of CA in macroeconomic models II: agent-based models (ABMs)

Model Locus of CA Effect of CA Description of CA process Determinants of CADosi et al. (2019, representative for the‘Keynes-meets-Schumpeter’ models),Caiani et al. (2019, representative forthe ‘Benchmark‘ models)

Consumption good firms Increased labor productivity

Firms invest into R&D to increase theprobability for successful innovation orimitation which, in turn, may increaseproductivity

R&D expenses,spillovers

Ciarli et al. (2018)Capital good firms

Capital goods with higher laborproductivity

Firms invest into R&D based on anticipatedpayoffs; this increases trials for innovationactivity per time step

R&D expensesConsumption good firms

Higher product quality

Dawid et al. (2019), Hötte (2020,representative for the EURACE@Unibimodels)

Employees

Better exploitation of highproductivity productiontechnologies in the employingfirm

Workers have exogenously given generalskills that determine speed oflearning-by-doing, which increases specificskills (i.e. tacit knowledge that workers cantake with them to other firms)

Experience

Capital good firmsPotentially higher productivity ofcapital goods, if matched withskilled workers

Firms engage in R&D; success randomly addshigher quality vintage, which might be soldto consumer good firms

R&D expenses

Dawid et al. (2019), Hötte (2020,representative for the EURACE@Unibimodels)

Employees

Better exploitation of highproductivity productiontechnologies in the employingfirm

Workers have exogenously given generalskills that determine speed oflearning-by-doing, which increases specificskills (i.e. tacit knowledge that workers cantake with them to other firms)

Experience

Capital good firmsPotentially higher productivity ofcapital goods, if matched withskilled workers

Firms engage in R&D; success randomly addshigher quality vintage, which might be soldto consumer good firms

R&D expenses

Rengs and Scholz-Wäckerle (2020) Consumption good firms

Reduction of carbon intensity ofproduction

Firms decide to invest into R&D directed atgreen technology reducing carbon intensity, orat gray technology reducing labor intensity ofproduction; for both technologies, spilloversacross firms exist

Spillovers, R&DReduction of labor intensity ofproduction

Hötte (2020), who extends the Eurace@Unibi-eco model (Dawidt al., 2019), models the CA process both on the level of employeesnd firms: the firm side is similar to the models discussed above: capitalood firms invest into R&D to increase the probability to invent moreroductive capital goods, which can then be sold more attractively tohe final good firms that produce a homogeneous consumption good.n the level of individuals, employees improve their abilities to workith particular capital goods on the job via a learning-by-doing process

through which workers acquire technology-specific skills. This has im-portant implications: by representing the specificity of employees’ skillsand the mechanism of learning-by-doing, Hötte (2020) is able to studythe co-evolutionary character of technological change: the diffusion oftechnologies depends on the adoption decisions of the users and thesolution of an underlying coordination problem, which is why lock-in effects on inferior, but already widely adopted technologies, mightoccur.

Finally, Rengs and Scholz-Wäckerle (2020) distinguish between ca-pabilities necessary for two aspects of production: firms can reduce theamount of CO2 emissions associated with production, or they can in-crease their labor productivity. When deciding on which competenciesto improve upon, firms compare themselves with their competitors andfollow the strategy of those firms that are making the highest profits.Then, based on past profits, they decide on the amount of funds to beallocated to the improvement of their capabilities. The effect of theseinvestments also depends on spillover effects from other firms: the morefirms focus on one particular way of improving their products, the moreeffective their investments become.

In addition to the CA processes just described, many macroeconomicABMs also study the effect of institutions and government activity onthe processes of CA. For instance, Dawid et al. (2018) study the effect ofinnovation policy on the speed of CA in less developed countries. Here,innovation policy takes the form of (directed or undirected) subsidiesfor the R&D investments of firms, thus serving as a second level

5

determinant of CA operating through the channel of R&D investment

as explained above. Another example is provided by Dosi et al. (2018),where the authors study the effect of an active government investinginto R&D activities on its own. Via knowledge-spillovers, these R&Dactivities then affect the rest of the economy as well. Again, in thiscase the fundamental channels of CA are R&D activities and knowledgespillovers, yet it is clearly acknowledged that since the government canengage in these activities itself, its behavior directly affects CA in theoverall economy. Finally, as evidenced by Caiani et al. (2019), differentinstitutions of wage determination may also affect CA. They showthat coordinated wage can, in the long run, lead to a concentrationof firms and, via the channel of ‘Schumpeter Mark II’, to higher andmore targeted investments into R&D activities. As in the two examplesabove, the institutions determining wages here affect CA via the morefundamental channel of R&D investments. Thus, while it is importantto keep in mind that many current ABMs do investigate the role ofinstitutions and policy makers for CA, they do so via reference to themore fundamental determinants of CA explicated before, i.e. mainlyR&D investment.

3.4. Taking stock

This brief overview of the status quo highlights a number of is-sues: first, the methodological (and epistemological) starting point ofmacroeconomic models matters: EGMs are built on the neoclassicaltheory of the firm. This theory is, in turn, built on entirely differentpremises than most of the organization and management literature,which is mainly concerned with how CA actually takes place. Thismakes it difficult to consider newer insights from this literature withinthe EGM framework. Strong advocates of EGMs would probably notconsider this to be a problem since they are usually convinced that theneoclassical theory of the firm, while being descriptively inaccurate,is a useful foundation for macroeconomic theorizing, especially sinceit allows the construction of analytically tractable models within thetheoretical framework of classical microeconomic theory. The fact that

Research Policy 50 (2021) 104258C. Gräbner et al.

a

such theory is unable to make sense of the considerable heterogeneityof firms in terms of capabilities and competitiveness is not considered adrawback, since the main task of EGMs is, according to this view, not torepresent every part of the economy accurately, but to produce insightsabout the dynamics of the economy as a whole. This is, at least fromour view, problematic since (1) the accumulation of capabilities is atthe heart of technological change, which itself is essential to understandmacroeconomic dynamics, (2) it seems somewhat unsatisfying to buildmodels of an economic system that accommodate central insights onlyon the macro level, but are by virtue of their methodological startingpoint unable to validate their (intermediate) ‘results’ on the micro andmeso level against facts identified by expert disciplines for these levels.Such an approach makes it basically impossible to build structurallyvalid models. And (3) when these models are used to design policies,the latter usually have an impact not only on the macro level, butalso on firms and their employees. Fully neglecting or misrepresentingthese levels may, then, come with considerable costs. This is why suchmodels should at least be complemented with models that also take intoaccount the micro and meso level.

Second, the last argument also applies to Keynesian models: despitebeing methodologically more open than EGMs, the focus of thesemodels is usually on the macroeconomic level alone. This allowed forthe identification of some important macroeconomic determinants ofCA, yet it makes it necessary to complement them with models of themicro and meso level. This is in fact what happened with the emergenceof the newest generation of agent-based macroeconomic models, whichtook the important macroeconomic insights from Keynesian modelsand integrated them with an evolutionary consideration of what ishappening on the micro and meso level of the economy. They areconsistent with a systemist approach with its layered ontology and theconsideration of top-down and bottom-up effects (see also Gräbner andKapeller, 2015, and Section 2.2), they allow for profit-seeking insteadof profit-maximizing firms, and are in principle flexible enough toinclude a variety of institutional and behavioral factors driving CA.

The newer class of ABMs could, therefore, be a useful vehicleto bridge macroeconomics with insights from the organization andmanagement literature. However, in their flexibility, they bear thedanger of too many ad hoc assumptions and the potential of orientingthese models on the empirical evidence generated by the empiricalliterature has certainly not been exhausted: up till now, the mostprominent determinant for CA is R&D investment (and, to a lesserextent, knowledge spillovers), and the main effect is an increase in firmproductivity. Relatively little attention has been given to the role ofCA for the invention of radically new products (see also Gräbner andHornykewycz, 2021). The remaining part of the paper tries to providemacroeconomic modelers with an overview about existing insights onCA and hopefully facilitate the consideration of these more fine-grainedresults in future macroeconomic models.

4. Theory and empirics beyond macroeconomics

This section comprises an explication of the survey strategy (Sec-tion 4.1) as well as a presentation of the results for the empirical(Section 4.2) and theoretical (Section 4.3) literature.

4.1. Methods and data

The empirical literature on ‘capabilities’ as defined in Section 2 issubstantial, and was reviewed via a two-fold search strategy, compris-ing both a formal and an informal screening process (see Fig. 1). Duringthe formal screening process, papers were selected in three stages —baseline, impact and content screening. The baseline screening consistedof a bibliographic search in the Web of Science (WoS) database usingthe search strings described in Table 3. At this stage, the search focusedon papers published in SSCI-ranked journals between 2006 and 2020(searched 28 November 2020). In order to exclude purely technical

6

c

and/or engineering papers, the search was restricted to journals thatare listed in at least one of the WoS categories ‘Business’, ‘Economics’or ‘Management’. After excluding retracted publications, proceedingspapers, book chapters, editorial material and papers not written inEnglish, the baseline screening resulted in a selection of 4.034 papers.

A two-stage impact screening was then used to identify the mostinfluential papers. First, only papers that were published in a journalwith a 5-Year Impact Factor (2019) higher or equal to 5, or that belongto the top 15 percent of most cited papers from their publication yearwere considered further, resulting in a sub-sample of 1.862 papers.Second, papers with less citations per publication year than the medianvalue were excluded. This resulted in a set of 860 papers.

The final step consisted of a two-fold qualitative content analysisin terms of topical relevance. First, we searched for notions of CA – assuggested by our definition in Section 2 – in the titles and abstracts.Here, our focus was on mechanisms underlying CA; papers mainlyconcerned with the implications of CA were excluded. Moreover, wefocused on the manifestation of CA in the manufacturing of products,meaning that papers that mainly deal with the provision of services,or that exclusively deal with certain aspects of process innovation,were excluded. This resulted in a sub-selection of 211 papers. In thefinal stage, their full-texts were qualitatively evaluated according totheir degree of topical relevance and their empirical strategy. Withregard to the latter, we prioritized empirical large-scale analyses (multi-country/sectoral/firm) and multiple-case designs over single-scale orsingle-case designs. Only for areas for which such more general studieswere rarely or not available, single-case designs were kept in thesample. This final screening stage resulted in a selection of 68 papers.

This formal screening process was complemented by an informalone, by which we added papers to our sample that did not match ourformal baseline search strings, despite being of high topical relevance.More precisely, this includes papers that (1) were recommended byexperts from theory and practice that we consulted during the literaturesearch or that (2) were referenced multiple times by the other studies inour literature selection (‘snowball sampling’).7 A list of all the 22 papersadded during the informal screening process is provided in Appendix B.

In sum, the final set of papers consisted of 90 representative em-pirical studies summarized in Tables 4 and 5. The literature review onrepresentative theoretical models of CA processes (Table 6) was basedon informal search, since the overall literature is much smaller andeasier to handle directly.

4.2. Empirical contributions

4.2.1. Empirical results on the aggregated levelDeterminants of CA on the aggregated level can be clustered ac-

cording to three main topoi (of which the boundaries are not alwaysclear-cut): (1) the cumulative and path-dependent nature of CA, (2) theinstitutional environment, and (3) the relevance of government activity.The results are summarized in Table 4.

Cumulative and path dependent nature of CA. New capabilities are accu-mulated step-wise and new capabilities tend to be similar to existingones. This ‘Principle of Relatedness’ (Hidalgo et al., 2018, hereafterPoR) has been examined on the basis of different operationalizationsthat can be considered as different but related attempts to empiricallytest the somewhat abstract principle:

• Skill relatedness (e.g. Neffke et al., 2017) as measured by inter-industry labor flows: capabilities develop along similar humancapital or skill requirements.

7 Evidence suggests that this kind of search strategy often appears to bemore viable approach in identifying the relevant literature compared to

onventional database searching (Pawson, 2006).

Research Policy 50 (2021) 104258C. Gräbner et al.

cpaCewatt

Fig. 1. The methodological strategy for the literature review.

Table 3Search strings and combinations of them used for the bibliographic search. The search covers the timespan 2006–2020 and was conducted on November 282020. Description of the WoS field tags: AK=Author Keywords (as listed by the authors), KP=Keywords Plus® (index terms automatically generated by WoS fromthe titles of cited references), TS=Topic (includes Title, Abstract, Author Keywords and Keywords Plus®).

Bibliographic search strategy: Web of Science (SSCI)

Set Results Search strings and combinations

#6 4.034 (#2 OR #3) AND #1 AND #4 Refined by: Web of Science categories: (Management OR Business OR Economics) AND [excluding]document types: (book chapter OR retracted publication OR proceedings paper OR editorial material) AND languages: (english)

#5 5.523 (#2 OR #3) AND #1 AND #4

#4 38.706 (KP=(evolutio* OR dynamic* OR upgrad* OR change OR develop* OR transit*) OR AK=(evolutio* OR dynamic* OR upgrad* ORchange OR develop* OR transit*)) AND (KP=(techn* OR innovation* OR capabilit*) OR AK=(techn* OR innovation* OR capabilit*)) ORAK=((evolutio* OR dynamic* OR upgrad* OR change OR develop*) AND (techn* OR innovation* OR capabilit*))

#3 8.312 TS=((‘‘product innovation*’’) OR (‘‘product* capabilit*’’) OR (product* NEAR/1 complex*) OR (product* NEAR/1 sophisticat*) OR(‘‘product* space’’) OR (product* NEAR/1 diversificat*) OR (‘‘new product*’’ NEAR/1 introduction*) OR (‘‘new product*’’ NEAR/1development*) OR (radical NEAR/1 innovat*))

#2 19.975 TS=((learn* OR capabilit* OR complex* OR progress* OR deepen* OR knowledge) NEAR/1 (techn* OR innovat* OR accumulat*))

#1 1.038.020 TS=(countr* OR industr* OR nation* OR region* OR firm* OR compan* OR organi?ation*)

• Technological relatedness (e.g. Neffke et al., 2011; Balland et al.,2019; Acemoglu et al., 2016; Petralia et al., 2017) as measured byplant-level product portfolios, country-level production basketsand/or patent data: capabilities develop along regional paths ofproduction.

• Export relatedness (e.g. Bahar et al., 2014) as measured by thecomposition of export baskets: a country’s capabilities convergeto the capabilities of neighboring countries, i.e. there are infor-mation spillovers between close regions.

Hidalgo et al. (2007) have formalized this general idea on theountry level within the framework of the product space, where ex-orted products are used as a proxy for the technological capabilitiesvailable within a country. Although there is no elaborated theory ofA underlying the concept of the product space, it provides furthervidence on the country level that CA happens in a cumulative fashion,ith new capabilities being similar to those that have already beenccumulated. Hence, CA depends on local experimentation and therebyhe exploration of similar, or the re-combination of existing capabili-

7

ies (see also Hidalgo and Hausmann, 2009; Vermeulen and Pyka, 2014,

p. 10575). The framework thus bears much similarity to evolutionarytheories of learning, including the idea of technological capabilities asa solution for particular technological challenges (e.g. Silverberg andVerspagen, 2005).

The path dependency of CA implies that previous capabilities deter-mine the development (path) of future ones. For instance, Lee and Yoon(2015) show how CA depends on a country’s focus of its knowledgebase: countries that have already invested resources in R&D and in-digenous technological capabilities may pursue a different strategy forfurther CA (e.g. ‘make’ new products) than countries that have primar-ily focused on improving their production capabilities (e.g. ‘buy’ newproducts). Coccia and Watts (2020) propose a theory of technologicalparasitism to explain development trajectories of different technologies.They argue, that a ‘host’ technology – defined as a ‘‘complex systemdesigned to be a platform system’’ (Coccia and Watts, 2020, pp. 6–

7) – evolves more rapidly when it is associated with many ‘parasitic’

Research Policy 50 (2021) 104258C. Gräbner et al.

olfpteeieLmafcc

Iuafieiao

d

at

Table 4Central empirical findings on the determinants for CA on the aggregated level.

Empirical findings on the aggregated level

Findings on the CA process Sources

Cumulative and path dependent nature of CA

Relatedness: CA is related to preexisting capabilities Neffke et al. (2017), Balland et al. (2019), Petralia et al. (2017), Neffke et al.(2011), Acemoglu et al. (2016), Bahar et al. (2014)

CA is determined by a country’s knowledge base focus Lee and Yoon (2015)

CA is determined by the degree of inter-relatedness betweentechnologies

Coccia and Watts (2020)

CA is hampered by structures causing institutional ‘lock-in’ Brem et al. (2016), De Noni et al. (2017), Radosevic and Yoruk (2018)

Institutional environment

CA is linked with intellectual property rights (ambiguous) Kim et al. (2012), Ang (2011, 2010), Fernandez Donoso (2014), Dosi et al. (2006)

CA may be supported by protectionism Figueiredo (2008)

CA is linked with economic integration (ambiguous) Fu et al. (2011), Castellacci and Natera (2013), Ascani et al. (2020), Filipescu et al.(2013), Molina-Domene and Pietrobelli (2012), García et al. (2013), Wu (2012), Xieand Li (2018)

CA is linked with labor market institutions (ambiguous) Hoxha and Kleinknecht (2020), Cetrulo et al. (2019), Filippetti and Guy (2020)

CA may be supported by financial development but is negativelyassociated with financial liberalization (policies)

Ang (2010, 2011), Wu et al. (2016)

CA is linked with socio-political and socio-economicdevelopment: inequality, education, health and publicinfrastructure

Castellacci and Natera (2013), Gao and Zheng (2020), Dutrénit et al. (2019)

Government activity

CA is fostered by public R&D investment Mazzucato and Semieniuk (2017), Castellacci and Natera (2013), Stojčić et al.(2020), Greco et al. (2017)

CA is fostered by public ownership, governance and procurement Mazzucato and Semieniuk (2017), Stojčić et al. (2020), Salerno et al. (2015), Yuet al. (2015), Castelnovo et al. (2018)

CA is fostered by active industrial policies: training programs,coordination, mediation and collaboration

Lee and Yoon (2015), Mazzucato and Semieniuk (2017), Kwak and Yoon (2020),Lebdioui (2019), Mehri (2015), Figueiredo (2008), Perez-Aleman and Alves (2017)

technologies, i.e. related technologies that interact and adapt with thehost.8

The path dependency of CA can also explain why countries, regionsr industries fail to sustain CA: it bears the danger of a suboptimalock-in to an inferior capability set. Brem et al. (2016), for instance,ind that while standardization (measured by an industry’s dominantroduct design) may shape the path for fostering process innova-ion, it may also aggravate radical product innovation. And De Nonit al. (2017) show that a diversified regional knowledge base is nec-ssary to reduce the risk of spatial lock-ins caused by (excessive)ntra-regional collaboration. This might also be one reason for thexistence of the ‘middle-income trap’ (Radosevic and Yoruk, 2018;ee and Malerba, 2018): In comparison to high-income countries,iddle-income economies may not benefit from increased technology

nd knowledge exchange with the global economy, since they mayind themselves in a ‘vicious specialization’ in activities related toheap labor and poor productivity, due to their lock-in on inferiorapabilities (Gräbner and Hafele, 2020).

nstitutional environment. Numerous studies stress the relevance of reg-latory measures and other institutions for CA at the macro, mesond micro level. Determinants include the design of property rights,inance, trade policy, labor market legislation and the overall (public)nfrastructure, all of which were subsumed under the topos ‘institutionalnvironment’. While almost all studies agree on the relevance of thenstitutional environment, disagreements about the relative strengthnd the direction of the effects, as well as the underlying mechanismsf particular institutions, remain.

One determinant of CA that has attracted much attention is theesign of policies that foster economic integration. Various studies find

8 The authors illustrate their argument by providing different historical ex-mples such as the evolution of aircrafts (host technology with many parasiticechnologies) or bicylces (host technology with few parasitic technologies).

8

positive spillover effects caused by trade liberalization and/or FDI-flows in developing (Castellacci and Natera, 2013; Molina-Domeneand Pietrobelli, 2012), and high-income countries (Ascani et al., 2020;Filipescu et al., 2013). However, there is no consensus on whether tradeliberalization is generally beneficial to CA. Some studies even point tonegative effects of economic integration on CA. García et al. (2013),for instance, find that inward FDI flows are negatively associated withex-post innovation performance of Spanish manufacturing firms. Onepossible explanation for their finding is that foreign investors may shiftinnovation activities into their home country, another reference to the‘vicious specialization’ argument made above.

Several studies suggest that the effect of economic integration onCA is contingent upon various context factors. Wu (2012), for instance,finds that market competition might expose a firm to high-risks wheninvolved in technological collaboration with competing firms (e.g op-portunistic behavior), thereby dampening the effect of collaboration onproduct innovation. This negative effect might, however, be offset inhigh-technology contexts where the high level of uncertainties makescollaboration between competitors mandatory. Xie and Li (2018) arguethat less openness may prove beneficial to emerging-market exporterssince these firms can use an information arbitrage to innovate moreactively in their home region than their non-exporting counterparts.Moreover, they argue that learning-by-exporting is only effective whenhome and export destination have a similar level of institutional de-velopment. Castellacci and Natera (2013), on the other hand, arguethat for the development of an economy’s innovation system, interna-tional trade is important for middle-income countries — yet that foradvanced and less developed countries other determinants are muchmore decisive.

Overall, the literature seems to suggest that the effect of opennessdepends on a country’s development stage and the maturity of the firmsin the respective regions (see also, e.g., Radosevic and Yoruk, 2018; Fuet al., 2011). For instance, in an extensive case study on the ‘Industrial

Pole’ of Manaus in Brazil, Figueiredo (2008) finds that protectionist

Research Policy 50 (2021) 104258C. Gräbner et al.

dCdon

asRtc

policies triggered various knowledge flows from outside firms intoManaus. These flows have led to an accumulation of initial firm-leveltechnological capabilities, which later served as a precondition forsuccessful liberalization reforms in the 1990s and that further boostedCA. Yet this boost due to liberalization would not have been successfulwithout earlier protectionist measures — an interpretation that alignswell with the work of economic historians on the determinants of catch-up processes after the Industrial Revolution (e.g. Allen, 2011; Chang,2003).

However, there are also some contradictions regarding the insti-tutional environment which cannot be solved easily by adopting adynamic perspective. For instance, although it is widely agreed uponthat financial development generally benefits CA via inducing R&Dspending,9 financial liberalization policies tend to impede the CA pro-cess — particularly in developing countries (e.g. Ang, 2010). Onereason for this may lie in the reallocation of human capital to thefinancial sector, resulting in a lack of CA in the technology sector (Ang,2011).

The organization of intellectual property rights (IPR) provides a sim-ilar case: While several empirical studies suggest a generally positiveeffect of IPR institutions on CA (e.g. Ang, 2010, 2011; Kim et al.,2012), Fernandez Donoso (2014) finds an inverted U-shaped relation-ship between patents and technological complexity. Moreover, therealso is literature that questions the positive effect of IPR at all, arguingthat institutions of open access are essential for collective learning and,thereby, for CA (Dosi et al., 2006; see also Marengo et al., 2012;Heinrich, 2013).

Finally, the case of labor market institutions further highlights thepotential ambiguity of the role of the institutional environment inCA. The removal of rigidities on the labor market has, for instance,been found to create incentives for the creation of low-paid and low-skilled jobs — at the expense of well-paid jobs with a higher innovativepotential. Such liberalization policies also tend to discourage workertraining and the accumulation of worker experience. Both of thesefactors harm an economy’s capacity for CA (Hoxha and Kleinknecht,2020; Cetrulo et al., 2019). These findings, however, are inconsistentwith Filippetti and Guy (2020) who find that both lower levels ofemployment protection and high levels of unemployment protectionmay be conducive to CA. They argue that unemployment protectionactively promotes knowledge diversity and a broader range of learningwhile employment protection may rather lead to the opposite.

Results are less inconsistent when it comes to public infrastructure.Various studies find that a wide spectrum of different infrastructureindicators ranging from the degree of electrification to the quality ofinstitutions and governance have been found to be essential for success-ful CA, yet mainly as an ‘absorptive capacity variable’ (see below) thatmoderates the positive effect R&D investments, the accumulation ofhuman capital and collaboration efforts among firms (see Section 4.2.2below), especially in middle-income countries (Castellacci and Natera,2013; Dutrénit et al., 2019; Gao and Zheng, 2020).

Government activity. Numerous studies highlight the relevance of di-rect government action, with the entrepreneurial state as describedby Mazzucato (2014) being a recent example.10 Channels through

9 This is also supported by evidence on the ambiguous role of informalebt for entrepreneurial finance especially in emerging economies such ashina (Wu et al., 2016): In the absence of formal access to finance, informalebt might improve a ventures’ innovation performance while excessive levelsf (expensive) informal debt may lead to high financial pressure resulting inegative effects for product innovation.10 There is not always a clear line to draw between active governmentction and regulatory policy, e.g. when it comes to incentive-based policiesuch as tax incentives for firms, investing a share of their revenues into&D (Figueiredo, 2008). There is also considerable evidence on the rela-

ionship between public subsidies for R&D and the propensity of firms to

9

ollaborate with external organizations (Greco et al., 2017).

which more direct government action enhances CA include public R&Dinvestment (Stojčić et al., 2020; Greco et al., 2017; Mazzucato andSemieniuk, 2017; Castellacci and Natera, 2013), public ownership andgovernance of key sectors and enterprises (Yu et al., 2015; Mazzucatoand Semieniuk, 2017) but also public procurement policies (Castelnovoet al., 2018; Salerno et al., 2015.

The possibilities for constructive government action have also beenexplored in numerous case studies. Mazzucato and Semieniuk (2017)show for the case of greener technologies that by being involved in R&Dspending along the entire innovation chain (e.g. from basic research tofinal product development) public organizations may not only absorbhigh risk levels and uncertainty often related to exploring radicalproduct innovation. They may also use well-crafted ‘mission-oriented’agencies that are actively involved in shaping and creating new marketconditions necessary for new product development. Figueiredo (2008)illustrates this in the specific context of Manaus, where governmentorganizations acted as active development agencies in shaping capa-bilities of local firms by forcing them to obtain quality standards orto introduce new management and organization techniques beyondwhat had been demanded on the market. Further examples are pro-vided by Kwak and Yoon (2020), Lee and Yoon (2015), Perez-Alemanand Alves (2017), Lebdioui (2019) and Mehri (2015) who explorea diverse repertoire of active government policies aimed to facili-tate coordination, mediation and collaboration between relevant actors(e.g. domestic and foreign firms, universities, research institutions)for triggering or supporting CA processes. Often, the government caneffectively support the mechanisms operating on the firm level, asdescribed in Section 4.2.2.

4.2.2. Empirical results on the firm levelResearch on CA on the firm level has highlighted a broader spectrum

of determinants than on the aggregate level, with numerous firm-levelstudies suggesting a complex interplay of several distinct determinantsand channels of CA. More than in the case of the aggregated level, andpartly because of the greater methodological pluralism in the manage-ment sciences, one finds an increasing number of in-depth case studiesthat follow a qualitative approach. This focus on detailed case-basedexposition in the business literature aggravates the identification ofprocesses on the firm level that should be considered in macroeconomicmodels: many of the more recent papers provide ‘too much’ detailabout how CA operates on the firm level than could reasonably beaccommodated within a macroeconomic model, or they refer to casesthat are too specific — e.g. by only referring to firms that are ownedby families since at least three generations. Thus, without denying theirvaluable contribution, we do not discuss such contributions; rather, wetry to distill the insights from the literature at a level of detail that isstill potentially digestible for macroeconomic modelers.

Moreover, for the sake of clarity the determinants will be clusteredas follows (see Table 5): first, one may broadly distinguish betweeninternal and external sources of CA.11 Some determinants, however, donot fit neatly into this distinction: absorptive capacities and the relat-edness of new capabilities to old ones both are important in the contextof both internal and external CA processes. Therefore, in Table 5 theyare reported separately. Moreover, since the literature treats some sub-categories of internal and external CA processes in particular depth,these are reported distinctly in Table 5 as well.

Absorptive capacities. The concept of absorptive capacities (ACAP) wascoined by Cohen and Levinthal (1990, p. 128), who defined it as afirm’s ability ‘‘to recognize the value of new, external information,assimilate it, and apply it to commercial ends’’. It is one of the keytopics of empirical research on the firm level and its role as an under-lying mechanism of CA has been confirmed by numerous studies (e.g.

11 These topoi were inspired by the framework of Bell and Figueiredo (2012,pp. 24–25).

Research Policy 50 (2021) 104258C. Gräbner et al.

m

lrtpa

Table 5Central empirical findings on the determinants for and functioning of CA on the firm level. Although the referenced papers identified the factors listed here as a maindeterminant for CA, most of these factors are not discussed in isolation.

Empirical findings on the firm level

Findings on the CA process Sources

Absorptive capacities

CA is positively moderated by ACAP Love and Roper (2015), Kobarg et al. (2018), Kafouros et al. (2020), Forés and Camisón(2016), Song et al. (2018), Vlačić et al. (2019), Zou et al. (2018)

Relatedness

Acquiring related knowledge is relatively easier Phene et al. (2006), Kavusan et al. (2016)

CA is moderated by relatedness of the knowledge base and thebusiness model

Enkel and Heil (2014)

External learning, esp. alliances

CA is enhanced by alliances Lasagni (2012), Enkel and Heil (2014), Park et al. (2014), Love and Roper (2015), Todoet al. (2016), Frankort (2016), Kavusan et al. (2016), Chuang and Hobday (2013)

CA is moderated by the kind and diversity of alliance partners(non-linear)

de Leeuw et al. (2014), Kobarg et al. (2019), Haus-Reve et al. (2019), Edmondson andHarvey (2018)

CA is linked with the number and intensity of alliances Greco et al. (2016)

CA is fostered by maintaining alliances Nieto and Santamaría (2007), Love et al. (2014)

External learning

CA is facilitated by supplier and customer consideration Foss et al. (2011), Coviello and Joseph (2012), Jean et al. (2014), Chang and Taylor (2016)

CA depends on spatial characteristics Laursen et al. (2011), Kafouros et al. (2015)

CA operates through employee mobility Edler et al. (2011), Herstad et al. (2015), Schaefer (2020)

Internal learning

CA is linked with the diversity of management teams, employeesor ownership

Bogers et al. (2018), Schubert and Tavassoli (2019), Chen et al. (2014)

CA is linked with management, leadership, emotional capabilities,intra-firm knowledge-sharing technologies and culture

Akgün et al. (2007), Tellis et al. (2009), Felekoglu and Moultrie (2014), Büschgens et al.(2013), Caridi-Zahavi et al. (2016), Schneckenberg et al. (2015), Chang et al. (2015)

CA is fostered by R&D spending (but relatively less relevant forlow and medium tech firms)

Nakata and Im (2010), Love and Roper (2015), Chen et al. (2016), Heij et al. (2020),Santamaría et al. (2009)

CA is accomplished by learning by doing Dosi et al. (2017)

CA is linked with firm size and age (non-linear) Vaona and Pianta (2008), Leiblein and Madsen (2009), García-Quevedo et al. (2014), Zouet al. (2018)

Song et al., 2018; Vlačić et al., 2019).12 While many recent studieshave departed from the original definition, most of them agree that(1) they have a moderating effect on CA by positively influencing theimpact of other determinants (such as R&D investment or alliances) and(2) are important for the firm to absorb external knowledge (see alsoSong et al., 2018). Thus, ACAP do not immediately help firms to createadditional revenues, but rather to acquire new knowledge and capa-bilities more quickly (Zou et al., 2018), which suggests they are bestunderstood as a kind of dynamic capability of the firm. The literatureon what determines ACAP itself is less exhaustive, yet suggests at leastthe relevance of the human capital of the employees and, especially,the level of codified experience, cooperative routines and culture onthe organizational level (e.g. Engelman et al., 2017).

Relatedness. Firms acquire new capabilities more quickly if they aresimilar to those already accumulated — the PoR plays an importantrole not only on the aggregated (see Section 4.2.1), but also on thefirm level. It is also discussed under the topic of ‘proximity’ (especiallyby authors in the field of economic geography) and is understood asa highly multidimensional issue: Boschma (2005), for instance, dis-tinguishes between cognitive, social, institutional, organizational andphysical proximity,13 and for each dimension, the literature usually

12 Castellacci and Natera (2013) and Criscuolo and Narula (2008) introduceodels that translate the idea on an aggregated level.13 Cognitive proximity is high when the actors involved have a simi-

ar knowledge base, social proximity when actors are connected via socialelations such as friendship or trust, institutional proximity when two organiza-ions are embedded into similar institutional arrangements, and organizationalroximity when they share similar sets of structural relations. Finally, two

10

ctors are geographically proximate if their physical distance is low.

finds that CA is easier the higher the level of relatedness. In thissense, relatedness can – similar to ACAP above – be thought of as ameta-determinant that moderates other channels of CA such as the for-mation of alliances or the use of learning technologies discussed below.Frankort (2016), for instance, finds that when firms form alliances theylearn more effectively from each other if they are active in relatedtechnology domains.

This, however, does not mean that firms should always strive toaccumulate new capabilities that are most related to their currentcapability stock. Rather, the optimal level of relatedness depends onthe strategy of the firm: Here, the literature distinguishes explorativeand exploitative search (or innovation) activities of firms: in the firstcase, firms try to expand their existing capabilities, in the second casethey try to find entirely new ones – defined as those capabilities that arenot (very) related to the existing stock of capabilities. Both processesare important – but according to the PoR, the former is more accessiblethan the latter. In all, the relevance of the PoR finds considerableempirical support not only on the aggregated, but also on the firmlevel (see also Hidalgo et al., 2018).

Alliances. The analysis of how firms accumulate new capabilities byforming alliances with each other has a long tradition in the empiricalliterature. Most papers agree that one way for a firm to accumulatemore capabilities is to form alliances with other firms and, thereby,benefit from spillovers between them.

The overall effect of alliances on CA depends on a number of mod-erating variables and contextual factors such as the respective dynamiccapabilities and the experience of the firm in forming and maintainingalliances (e.g. Nieto and Santamaría, 2007; Love et al., 2014). A moreextensively discussed factor is the diversity of the alliance partners:diversity may reinforce the positive effect of alliances (see de Leeuw

Research Policy 50 (2021) 104258C. Gräbner et al.

a

ia(tinnRwd

–ecefmisfiiLfea

4

o

et al., 2014; Edmondson and Harvey, 2018). The effect is not linear,however: some argue that the positive effect of diversity is U-shaped,i.e. the positive effect increases only up to a certain threshold (e.g. deLeeuw et al., 2014). Others stress that it is not so much the number ofpartners, but that the knowledge base and the technological domain ofthe partners must be sufficiently similar to that of the host firm (see‘relatedness’ above; e.g. Frankort, 2016), or that the precise optimallevel of similarity depends on the innovation model pursued by thefirms (e.g. radical innovation capabilities require more diversity thanincremental ones, see Enkel and Heil, 2014). Finally, it also comes tothe kind of partners: Haus-Reve et al. (2019), using data from Norway,point out that while alliances with partners from the supply chain andpartners from science have individually positive effect, their joint effectis negative — they are substitutes rather than complements. Thus, thediversity of alliances is an important, yet intricate moderator for thepositive effect of alliances on CA.

Aside from these moderators, there are a number of related conceptsfrom the alliance literature that deserve mentioning: first, an importantconceptual distinction is that between the breadth and depth of thelliances, where breadth refers to the quantity and depth to the intensity

of alliances. This is especially relevant since Greco et al. (2016) suggestthat the just mentioned diminishing marginal returns of alliances arerelevant only for the breadth, but not necessarily the depth of alliances.Second, firms that form alliances usually are both cooperating and com-peting with each other simultaneously — they are in a relationship of‘coopetition’. It seems that such relationships benefit both partners mostif the degree of competition among them is moderate. Finally, severalauthors point out that while the inflow of new knowledge as such isbeneficial, the firm must also set up the routines required to transformthe new knowledge into new capabilities properly and allocate scarcemanagerial resources — in short, building and maintaining alliancesalso comes with costs (e.g. Greco et al., 2016; Audretsch and Belitski,2020). These costs might in some instances offset the potential benefitsfrom the knowledge inflow, the precise threshold as such depends onthe kind, goal, and context of the alliance.

In all, alliances help firms in the process of CA, yet there are certainpreconditions for them to be useful, the most important of which referto the concepts of ACAP and relatedness introduced above. More re-cently, research on ‘open innovation’ (Chesbrough, 2003) complementsthe literature on alliances: open innovation refers to an institutionalsetting in which firms use ideas both from inside and outside tocome up with innovations of various sorts. This can more generally beunderstood as the deliberate provision and exploitation of spilloversacross the whole population of firms.

External learning. A major group of firm-level determinants of CArefers to the influence of external learning mechanisms on CA (Belland Figueiredo, 2012). Depending on factors such as the institutionalenvironment, sectoral specificities and the firm type, firms interact withsuppliers, customers, competitors, research institutions and other socialnetworks and are thus able to access external resources. The literaturesuggests various channels through which external learning occurs.

One such channel concerns the ability of firms to improve upontheir capabilities by setting up routines that allow them to gatherinformation from their suppliers and customers (e.g. Foss et al., 2011;Coviello and Joseph, 2012; Jean et al., 2014; Chang and Taylor, 2016).Another channel simply refers to the spatial environment of the firm:Laursen et al. (2011), for instance, finds that firms located in a re-gion characterized by a high level of social capital find it easier tobuild up innovative capabilities and to increase the effectiveness ofthe own R&D efforts. Kafouros et al. (2015) find that region-specificinstitutional idiosyncrasies such as differences in IPR-enforcement, thelevel of openness and the quality of universities and research insti-tutes significantly impact the effect of academic collaboration on firminnovation performance. These insights also serve as a bridge to the

11

determinants on the aggregated level (see Section 4.2.1), since the f

institutional environment is often a major determinant for CA processesof the individual firm, illustrating again the need to take top-downeffects seriously (see Section 2.2). This environment governs activitiessuch as mergers or the hiring of new workers, which are often a keysource of CA within the firm (on the role of worker mobility see, e.g.,Edler et al., 2011; Herstad et al., 2015; Schaefer, 2020).

Internal learning processes. Especially in the management literature,one can find an enormous body of very detailed literature on internalmechanisms underlying CA on the firm level. Here, we focus onlyon the small share of insights that at least to some extent could beintegrated into macroeconomic models. This excludes viable insightsparticularly on leadership practices, corporate culture and governanceroutines (e.g. Tellis et al., 2009), the moderating and mediating roleof management (e.g. Felekoglu and Moultrie, 2014; Salerno et al.,2015), or the use of intra-firm knowledge sharing technologies (e.g.Schneckenberg et al., 2015). While we do not discuss these factors indetail, it is worth noting that firms usually are highly heterogeneous inthese respects.

One important topic of internal learning processes concerns thediversity of the employees and the management team (e.g. Bogers et al.,2018; Schubert and Tavassoli, 2019). Diversity may refer to educationalbackground, work histories or other dimensions. While the results areoften more nuanced – Bogers et al. (2018), for instance, find a positiveeffect only for diversity in terms of educational backgrounds — one mayroughly summarize that at least a certain degree of diversity is benefi-cial to CA whenever a firm has developed the routines to integrate thedistinct perspectives from different individuals, organizational learningaccelerates. Furthermore, diversity of ownership types (e.g. domesticvs. foreign ownership) appears to be an important resource channelfor CA, in particular in the context of an inadequate institutionalenvironment (Chen et al., 2014).

Another important set of internal learning processes is the accu-mulation of experience. In this context, Dosi et al. (2017) find robustevidence for learning by doing as a driver of CA, but mostly for industriesn developed economies. Experience is often proxied via firm size andge. The effect of age and size on CA is, however, intricate: Zou et al.2018), for instance, do not find empirical support for the idea thathe growth of a firm comes with an increase of its innovation capabil-ties, as often assumed in Schumpeterian studies. They conjecture thatew innovation processes such as crowd-sourcing and open innovationowadays may be transforming the advantages or disadvantages of size.egarding firm age they note that age is often negatively correlatedith the level of the ACAP of a firm – most likely because of the higheregree of formalization – and thereby not automatically good for CA.

Finally, and not surprisingly, many studies find a – mostly positivelink between R&D spending and CA (e.g. Nakata and Im, 2010; Chen

t al., 2016; Heij et al., 2020). However, there are some importantaveats: first, there is huge variation across firms with regard to theffectiveness of R&D spending. While there is no definite explanationor this variation, it seems to relate to internal factors, such as the firmanagement practice (e.g. Heij et al., 2020) or whether a firm is operat-

ng in a high- or low-tech sector (e.g. Santamaría et al., 2009). Second,ome authors even find adverse effects of R&D on CA: Li et al. (2018),or instance, analyze U.S.-based small and medium enterprises (SMEs)n the biopharmaceutical sector and find that the in-house R&D stocks initially conducive for the accumulation of intellectual resources.ater however, R&D only played a minor (or even negative role) forurther CA. The authors argue that in a dynamic and complex globalnvironment, other factors such as strong alliances might, overall, playmore important role for successful CA.

.2.3. Taking stockThe empirical literature reviewed in Section 4.2 highlights a variety

f factors contributing to CA on both the aggregate as well as on theirm level. Noticeably, many papers analyze the determinants of CA in a

Research Policy 50 (2021) 104258C. Gräbner et al.

m((

tetaotoappoikatnasr

twstcaewtecigmp

very context-specific setting (e.g. specific industries, specific firm typesor countries, in particular catching-up vs. developed ones), and mostpapers discuss several determinants simultaneously. Not surprisingly,this suggests that CA is a highly complex process, with importantdeterminants on both the aggregated and the firm level. Therefore,a systemist approach as suggested in Section 2.2 proves indeed to beadequate. Before we turn to concrete implications for modeling thesemechanisms, we complement our review of the empirical literature bytaking a closer look on specialized theoretical models of CA processes,since these models might provide a useful source of inspiration for theoperationalization of CA.

4.3. Specialized models of CA

As already discussed in Section 3, modeling CA in macroeconomicscomes with certain trade-offs regarding the models’ complexity. Modelsthat are not macroeconomic by nature and that focus exclusively onthe mechanisms underlying CA do not face these trade-offs. One maysuppose that these models have already formalized a richer set of mech-anisms than macroeconomic models, and may serve as a useful sourceof inspiration for the latter. Therefore, this section presents mecha-nisms that have received particular attention in this more specializedliterature (see the summary in Table 6).14

The fact that R&D activity fosters CA is taken to be self-evident inmany models and the corresponding processes will not be discussedin more detail; the basic mechanisms are similar to those discussedfor macroeconomics models (see Section 3). Noteworthy are, however,models that enrich this mechanism in an additional dimension, suchas Marengo et al. (2012), who study the effect of patent systemson CA. In their ABM, firms invest into (innovative and imitative)R&D to discover new products, which are assembled from differentcomponents. The overall performance of a product depends on thequality of the single components, as well as how well they fit together.The space containing all products is called a ‘product space’, andits complexity depends on the strengths of the interdependencies ofthe components when determining the overall product performance.Marengo et al. (2012) then study the effects of different degrees of IPRprotection: the stronger the patent system, the more finely-grained canfirms protect their innovations. In the case of maximum IPR protectionthey can prevent other firms from using any single component theyhave discovered so far, hampering knowledge diffusion considerably.The model indicates that too strong IPR protection hampers overallinnovation and welfare when the product space is more complex sincepatents work as obstacles in the product space and the invention of newproducts gets aggravated. The key underlying mechanism here is thatof external learning and spillovers across firms.

Another determinant underlying CA and often modeled in the spe-cialized literature is the organization of human capital. Here, the skillsand the knowledge embodied in each individual worker are the maindriver of CA and, therefore, firms have to organize their human cap-ital strategically in order to achieve CA. In a general equilibriummodel Dasgupta (2012) studies CA as a learning process which isspurred by the mobility of workers from countries and firms with highercapabilities. By moving to other firms, their knowledge is passed onto the latter firm’s workers, thus leading to CA in firms as well asentire regions (as the knowledge is passed on further). Thus, CA is hereunderstood as a social process, propelled by the interaction of peoplewith diverse knowledge.

Savin and Egbetokun (2016) introduce a model of CA in which theystudy the role of alliances, ACAP and R&D. They situate firms on a two-dimensional metric space which they call ‘knowledge space’ and where

14 We focus on models that study the interplay of several mechanisms. Forodels focused on one particular form of CA see, e.g., Vermeulen and Pyka

2014), Tur and Azagra-Caro (2018) (alliances), Santos Arteaga et al. (2019)internal learning) or Zhuang et al. (2011) (external spillovers).

12

the distance between two firms provides information on how similartheir production technologies are. A firm’s ‘proximity radius’, whichcan be extended via R&D investment, determines the knowledge thefirm has potential access to. In accordance with the empirical literature(see above) the authors assume that there is an inverted U-shapedrelation between technological distance and CA, i.e. knowledge fromtechnologically close and distant firms is less useful than that of firmswith medium proximity. However, to actually absorb the knowledgeof other firms in the first place, firms must either engage in alliances(leading to voluntary spillovers) or invest into their ACAP (leading toinvoluntary spillovers). Involuntary spillovers are purely a function ofACAP of a firm. Voluntary spillovers emerge from alliances, which areformed profit maximizingly, based on the expected R&D investment ofthe potential partner as well as the disadvantages in competitivenessthat arise from the spillovers that go toward the other firm. A partner-ship will only be formed if the alliance is mutually beneficial. Basedon the alliance decision, a firm will split up its (exogenously given)R&D investment, spending as much as needed to benefit from alliancesas well as involuntary spillovers. The rest will go to the expansion ofinternal knowledge. The model indicates that what cooperation strategyis best in terms of competitiveness depends on the current stage ofa firm and whether it is currently more dependent on voluntary orinvoluntary spillovers.

Pyka et al. (2019) model the interrelation of three different mecha-nisms of CA: R&D investment, alliances and learning-by-doing, as well ashe potential role of policies in fostering CA processes. In the model,ach firm is endowed with knowledge units that can then be combinedo invent new products that will then be sold on a market. CA processesre modeled as the acquisition of new knowledge units. There are threeptions: first, firms can invest into R&D. With a given probability,his creates a new knowledge unit that differs from the existing unitsnly within a pre-specified range, as suggested by the PoR discussedbove. The second option is to copy a knowledge unit from cooperationartners with whom the firm has formed an alliance. This copies onlyarts of the knowledge unit, however: aspects related to the experiencef the firm of applying this knowledge are not copied. Finally, firmsmprove upon their knowledge by learning by doing if they apply theirnowledge to produce products. At the same time, they can also forgetbout knowledge that remains unused in practice. The authors thenest various policy scenarios in which firms are motivated to engage inew alliances or invest more into R&D. They find that CA is the mostccessible when firms diversify their channels for CA and if they areupported by nuanced policy programs that provide firms to engage inegional collaboration.

Landini and Malerba (2017) study the catch-up dynamics betweenechnologically lagging latecomers and an advanced incumbent countryithin a globalized industry. Their major research interest refers to the

et of public policies that is best suited to achieve convergence amonghe two countries. Firms in both countries produce goods, which theyan sell on the domestic and foreign market. Prices depend on qualitynd quality depends on the techniques used by, and the capabilitiesxistent in a firm.15 Firms invest a fixed share of their profits into R&D,hich increases their chance to extract new and better techniques from

he technological space. There is also a country specific informationffect, which depends on the level of technological knowledge in theountry and can be thought of as general spillovers that facilitate thedentification of better techniques. The current technological spaceets expanded by chance (a ‘radical change’), in which case firmsight adopt the enlarged space if they have enough capabilities toerceive the enlargement. However, they might decide against it if too

15 This distinction does not fit neatly into our broad definition of capabilitiesabove. Rather, both techniques used and capabilities in the sense of Landiniand Malerba (2017) would count as capabilities in the sense of the presentpaper.

Research Policy 50 (2021) 104258C. Gräbner et al.

Table 6Implementation of the CA process in selected specialized models.

Specialized models of CA

Source Determinants Description of the CA process Model type

Marengo et al. (2012) Patenting and externalspillovers, R&D

By investing into (innovative and imitative) R&D, firms assemble products fromdifferent components, which can be patented such that other firms cannot producethem for some time. Central result: if the product space is complex, IPRs are anobstacle, rather than an incentive, to innovation.

ABM

Dasgupta (2012) Spillovers (Human Capital) International worker mobility leads to knowledge spillovers among workers; CA isrelated to a reduction in inequality and an increase in individual income.

General equilibrium model

Wersching (2010) R&D Investment,Spillovers, ACAP

Aiming for a competitive advantage, firms invest into R&D in order to accumulateinternal knowledge and into ACAP in order to gain from knowledge spillovers.

ABM

Savin and Egbetokun (2016) Alliances, ACAP, R&DInvestment

Through alliances and ACAP, firms (that aim for a competitive advantage) acquirevoluntary and involuntary spillovers, respectively; R&D investment can lead toincremental innovation.

ABM

Pyka et al. (2019) Alliances, Learning byDoing, R&D Investment

Firms collaborate on the intra-regional and interregional level to acquire implicit andexplicit knowledge, respectively. Also, they invest into R&D to expand internalknowledge. Knowledge that is used is intensified, leading to CA (via a radical orincremental increase in product complexity), whereas other knowledge is graduallyforgotten.

ABM

Caiani (2017) R&D Investment, Firm-Size,ACAP

Aiming for a competitive advantage, firms move on a technology space. In order toreach certain points, they invest in imitative and innovative R&D. The probability ofsuccessful imitation increases with size of the target industry.

ABM

Criscuolo and Narula (2008) R&D, FDI, ACAP,Intra-industry Spillovers

CA on the national level is determined by domestic R&D investment, FDI andinternational spillovers. The speed of CA (i.e. individual and national knowledgeaccumulation) is determined by ACAP.

Differential equations

Landini and Malerba (2017) ACAP, spillovers, R&D,national policies

Firms invest into R&D to extract new techniques from a technology space; theextraction gets facilitated by spillovers from domestic competitors. Sometimes thedomestic technological frontier expands and firms can use new techniques if theirACAP are high enough. Moreover, firms accumulate new capabilities according to apre-specified logistic function. National policies can accelerate this process.

ABM

many specific capabilities for the old space have been accumulated.Thus, lock-ins in inferior technologies are possible. Firms’ capabilitiesevolve according to a logistic learning equation, which is set as aparameter. Learning can, however, be accelerated by national policies.Aside from such policies that foster CA directly, i.e. speed up theaccumulation of new capabilities or the initial level of capabilities forfirms, other policies are also helpful, but depend on the technologicalenvironment. If radical technological change is relevant, facilitatingthe entry of domestic firms is highly complementary to other CA-enhancing policies, yet it remains ineffective if radical change is absent.Protectionism on the other hand is found to be helpful only in theabsence of radical technological change and only in combination withother CA-supporting policies.

Caiani (2017) discusses an ABM in which CA occurs through inno-vative and imitative R&D with the firm size determining the successof innovation and the extent of ACAP. Firms move on a network-liketechnological space where each node that is farther from the startingpoint is associated with a reduction of production costs via CA. In afirst step, each firm chooses a target node. In order to reach that node,firms can either innovate or imitate, whereby costs of imitating arelower than those of innovating. The probability of successful innovationincreases with the amount invested into innovative R&D. Since largerfirms invest more, they have an advantage. In order to imitate, a firmcollects a sample of firms. If any of them has the target node in theirportfolio, imitation will be successful. The number of firms that can besampled depends – just as before – on the amount of R&D investment,giving larger firms an advantage. However, success also depends onthe size and structure of the industry, with a larger industry offeringa higher probability of success. The sample size represents the sizeof the region within which firms to imitate can be sampled and can,therefore, be interpreted as its ACAP. In the first instance, the sampleis selected randomly. However, if imitation was successful, the imitatedfirm will be a preferred choice in the next period. Thus, industrieshave a different branch structure and they develop differently as firmsmove along, varying in size (i.e. number of firms). Both of these

13

factors influence the benefits that can be expected from innovation and

imitation such that it might be easier or harder for firms to becomelarge and for large firms to stay successful.

A similar approach is put forward by Wersching (2010). Here, CAcan occur both with respect to processes and products. Firms move ona circular technological space where they take a new spot betweentwo occupied spots if they innovate incrementally and where theytake a spot on an enlarged circle that is relatively farther away fromthe occupied spots if they innovate radically. Firms can invest intointernal R&D or ACAP in order to accumulate capabilities. Here againACAP enable firms to acquire spillovers from more distant points in thetechnology space. However, it is assumed that the relationship betweendistance and productivity of new knowledge is inverted U-shaped.

Criscuolo and Narula (2008) take a different approach and proposea differential equations model in which they generalize the originaltheory of ACAP to the national level. Thus, in their model nationalCA depends on domestic and foreign R&D investment, inward FDI andinternational spillovers, all moderated via national ACAP. The latterdepends on R&D expenditures, an institutional variable representingthe ease of knowledge diffusion, and the technology gap to mostadvanced countries. This way, the model replicates the stylized factthat individual and national knowledge accumulation is fastest duringan intermediate stage of catching-up. This way it also provides a niceformal operationalization of the ambiguous concepts of ACAP.

5. Discussion

5.1. Synthesis of the review

The results of the review are summarized in Table 7. It providesan overview on whether and how empirical findings on CA have, sofar, been considered in specialized and comprehensive macroeconomicmodels.

The first lesson is that the clearest alignment of the empiricaland theoretical literature is with regard to the positive effect of R&D

expenses on CA. This positive effect has not only been documented

Research Policy 50 (2021) 104258C. Gräbner et al.

Ceolti

5

ii

Satfw

5

dtatCi(ihfmHecfmlWne

nciaiab2ambgomDRrga

taoTofMtslastsAtD

5

oewpfp

empirically, it has also been considered in both specialized and macroe-conomic models various times (being part of all but two models in-troduced in this paper). But of course, the empirical literature studiesR&D activities on a much more fine-grained level than is acknowledgedin the models, and the potentially non-linear and context-dependentrole of R&D is considered only in some of the specialized modelsby distinguishing between strategies for exploitative and explorativesearch.

The second, maybe more surprising, insight is that there is no clearindication on whether specialized or comprehensive macro models tendto implement empirical findings more accurately. Noticeably, the mech-anisms that are related to government activity and institutions havereceived more attention in macroeconomic models; most likely sincethe former include a government by definition, whereas specializedmodels – in taking a narrower focus – usually do not. At the sametime, it is easier to integrate more complex processes such as externallearning mechanisms or ACAP into a model that has its main focus on

A, rather than into an already complex macroeconomic model (for anxception see Hötte, 2020). Also, the empirically highly relevant topicf the relatedness of CA has received more attention in the specializediterature. In these dimensions, there is certainly potential in takinghe existing specialized models and to think about how they could bentegrated into a more comprehensive macroeconomic framework.

.2. Modeling implications

We now discuss implications for advancing macroeconomic model-ng of CA. The first implication is a rather general call for caution whent comes to the contingency and context-dependency of CA. They repre-

sent a fundamental challenge for modelers and should, therefore, beconsidered carefully when interpreting models. Two further implicationsmay then serve as a more positive guide for macroeconomic modeling:first, the need for adopting a systemist view on CA as discussed inection 2 that allows one to take seriously the primacy of institutionsnd the micro-macro links affecting CA, and second, to take the firm ashe main repository for capabilities and to take seriously a number ofirm-level mechanisms that emerge from our review; all of these pointsill be discussed one by one.

.2.1. Contingency and context-dependency of CANumerous studies in our review stress the contingency and context-

ependency of CA processes, i.e. they highlight the relevance of factorshat are specific to particular spaces and points in time (e.g. Kwaknd Yoon, 2020; Kafouros et al., 2015; Chuang and Hobday, 2013),he stage of development (e.g. Kim et al., 2012; Dutrénit et al., 2019;astellacci and Natera, 2013) or the institutional environment of an

nnovating firm (e.g. Salerno et al., 2015). In particular, singular eventse.g. socio-political events, financial crises) often have an importantmpact on CA. An example is how the Asian financial crisis of 1997as led Japanese firms to enter cooperation agreements with Taiwaneseirms, a strategy that has ultimately led to the establishment of nu-erous and highly successful transnational companies (Chuang andobday, 2013). While having received considerable attention in thempirical literature, modeling contingency is difficult (and maybe evenounter-productive) per se, since models are mostly used as a way toormalize general mechanisms. It would, however, be desirable if theechanisms modeled allowed for the emergence of distinct and non-

inear development paths that are suggested by the empirical literature.e conclude that by keeping the relevance of contingency in mind, one

eeds to take seriously the limits of modeling when it comes to thexplanation of CA processes.

14

5.2.2. The primacy of institutions and the micro-macro links of CAThe reviewed literature further shows that institutions – here en-

compassing both the regulatory environment as well as governmentactivity – are crucial determinants of CA processes. This result ishighlighted both via case studies (e.g. Figueiredo, 2008; Lee and Yoon,2015; Mazzucato and Semieniuk, 2017; Mehri, 2015) and quantitativeanalyzes (e.g. Yu et al., 2015; Castellacci and Natera, 2013; Greco et al.,2017). Many of these studies stress the fact that implementing specificgovernment policies preceded an ongoing and successful process ofCA.16 Thus, in order to integrate CA more adequately into macroeco-omic models, the role of institutions and government action needsareful attention. Considering this insight against the models surveyedn Section 3, one may conjecture that there is a chance for modelerscross various schools to harvest some low-hanging fruits: in EGMs, fornstance, institutional aspects of regulation or direct government actionre often considered merely in the context of barriers to technology,eing reduced to ‘‘Pareto-improving policy interventions’’ (Acemoglu,009, p.483) such as subsidies to research, subsidies to capital inputs,nti-trust and patent policies. Keynesian models, on the other hand,ay not adequately reflect the effects on the micro and meso level

y considering the role of government institutions merely on the ag-regate level (e.g. labor market institutions). In ABMs, various formsf industrial policy (e.g. Dawid et al., 2019), implications of laborarket regulations (e.g. Caiani et al., 2019) or state-led research (e.g.osi et al., 2018) have been considered, yet only via their effect on&D investment. This leaves room for further enhancement, e.g. withegard to the investigation of the implications of public ownership andovernance, the government as a facilitator of alliances among relevantctors or the role of private property and open-access norms.

In addition to the need for revising the role of government institu-ions, the literature also shows a close interconnection of aggregatednd firm level determinants (especially in the long-standing literaturen NIS discussed in Section 2 Lundvall, 2007; Lee and Malerba, 2018).he institutional setting of a NIS does not only comprise a wide rangef actors and institutions beyond the immediate environment of theirm but also their well-crafted interactions with each other (Lee andalerba, 2018; Dosi and Nelson, 2018). Moreover, and analogous to

he dynamic capabilities view of the firm, the NIS of a country isubject to change depending on the stage of development (e.g. Castel-acci and Natera, 2013, see also Lee and Kim, 2009) and may differcross national contexts (Lee and Malerba, 2018). This implies that anyatisfactory analysis of CA – particularly in a modeling context – mustake into account determinants on both the micro and the macro levelimultaneously. Such a challenging task seems to be a natural topic forBMs and there only are few examples for other non-ABMs that take

his mutual interdependency of micro, meso and macro seriously (e.g.osi et al., 2019).

.2.3. The firm as the main locus of CAWe now come to implications for the firm level. These involve inter-

rganizational factors and processes working through various forms ofxternal learning as well as processes by which resources are drawn fromithin the firm — such as internal R&D or ACAP-building. All theserocesses were shown to give rise to considerable heterogeneity acrossirms, suggesting to give the representation of such heterogeneity highriority in the construction of models.

16 Figueiredo (2008) is representative for this when stating that ‘‘it isimportant to recognize the relevance of different government policies that havebeen implemented over the past 40 years, especially since the 1990s in thecontext of Manaus. In the absence of such government policies, all these firmsand some supporting organizations would probably not even have been there

in the first place.’’ (p. 84).

Research Policy 50 (2021) 104258C. Gräbner et al.

Table 7Mechanisms of CA in the sampled literature. (Red and ‘0’ stands for nonexistent or rare, orange and ∼ for still under-explored, and yellow and ‘+’ for extensiveimplementation.).

Synthesis of the review findings

Mechanism Level Macro models Specialized models

Relatedness and path dependency (cumulative accumulation of capabilities) Aggregated, firm 0 ∼

Institutional environment (regulations) Aggregated ∼ 0

Institutional environment (other) Aggregated ∼ 0

Government activity (industrial policy) Aggregated + 0

Government activity (other) Aggregated ∼ 0

R&D investment Aggregated, firm + +

Absorptive capacities Firm 0 ∼

External learning Firm 0 +

Internal learning (management) Firm 0 0

Internal learning (other) Firm 0 ∼

aG -ovM

D

c

Absorptive capacities. The consideration of ACAP as a determinant forhow fast firms accumulate further capabilities and how much theybenefit from other determinants, such as alliances, seems to be astraightforward and promising way to make contact with the empir-ical literature. This is not to deny that the latter also offers moredetailed elaborations into the functioning of ACAP which, however,seem to be difficult to transfer into a macroeconomic modeling context.For the concrete operationalization, existing theoretical models suchas Savin and Egbetokun (2016) or Gräbner and Hornykewycz (2021),already provide the necessary routines, at least for ABMs, that could beimplemented more or less directly within a macro ABM.

Relatedness. CA approaches should be operationalized in such a waythat they take place in a cumulative way and such that it is easierto acquire skills that is more related with a firm’s preexisting set ofcapabilities. In the specialized ABM literature, the use of technologyspaces that can be searched by firms (based on, e.g., Caiani, 2017;Landini and Malerba, 2017) seems to be useful. However, this is notthe only possibility, two alternatives being the operationalization offirm size on an NK landscape in the spirit of Kauffman (1993) (e.g.Ganco, 2017) or using a percolation approach in the spirit of Silver-berg and Verspagen (2005). Another straightforward addition could beto distinguish between explorative and exploitative search activities offirms (based on, e.g., Marengo et al., 2012; Gräbner and Hornykewycz,2021).

Alliances. Many studies are concerned with the capabilities required toexploit the knowledge that is available in environments characterizedby open innovation. In macroeconomic models this might be consideredby adding a parameter that controls for the public-good character of in-novations, or by allowing firms to form R&D alliances. The latter can beachieved by using existing routines from specialized models (e.g. fromPyka et al., 2019; Savin and Egbetokun, 2016) more or less directly.

Internal learning. The empirical results with regard to internal learningof firms suggest that modeling firms as simply accumulating capabilitiesvia aging, growth or ongoing activity might result in an inadequateconsideration of important intra-firm processes, despite these factorsbeing prominent in the current literature. Moreover, while most ofthe findings are too fine-grained for macroeconomic modeling, theyhighlight the importance of considering firm heterogeneity.

When it comes to the concrete mechanisms, it might be preferableto operationalize the mechanisms identified so far first in specializedmodels only and, thereby, develop routines that can be tested andthen integrated into macroeconomic models at a more mature stage.While the internal processes might be too detailed for macroeconomicmodels for now, at least the resulting differences in performance couldbe allowed for.

15

i

6. Summary and conclusion

This paper comprised a comparative survey of CA-related literaturein order to identify possibilities to better align the macroeconomicmodeling literature with recent insights from the specialized literature.The empirical literature on CA is manifold and shows that CA is notthe outcome of a single mechanism but that the determinants arehighly diverse. Considering this diversity in macroeconomic models isfar from trivial and there is an important methodological dimension tothis challenge: given that the applied literature, especially in the fieldof business and management studies, shows that the main repositoryof capabilities is the firm, and given the relevance of both top-downand bottom-up mechanisms for the dynamics of CA, macroeconomicmodeling frameworks consistent with a systemist approach are themost promising device for a truly comprehensive integration of CA intomacroeconomic models. Currently, it is clearly the framework of agentbased modeling, which, despite being incompatible with many keyassumptions of the economic mainstream, seems the most promisingfor achieving a closer integration of empirical findings on CA intomacroeconomics.

Nevertheless, although offering promising potential, ABMs still facea trade-off between providing a detailed representation of particu-lar mechanisms and keeping the comprehensive representation of acomplete economy manageable. Still, given the high relevance of CAprocesses both on the micro and macro level, we believe that workingon a closer integration of these processes into macroeconomic modelsis a fruitful avenue for future research. With the present paper wehopefully have facilitated this endeavor.

CRediT authorship contribution statement

Matthias Aistleitner: Investigation, Methodology, Software, Visu-lization, Writing - original draft, Writing - review & editing. Claudiusräbner: Conceptualization, Funding acquisition, Investigation, Methodlogy, Software, Supervision, Writing - original draft, Writing - re-iew & editing. Anna Hornykewycz: Conceptualization, Investigation,ethodology, Writing - original draft, Writing - review & editing.

eclaration of competing interest

The authors declare that they have no known competing finan-ial interests or personal relationships that could have appeared tonfluence the work reported in this paper.

Research Policy 50 (2021) 104258C. Gräbner et al.

Table B.8Papers added via the informal search process. Note that theoretical models were only added via informal search since thefield of this literature is not so broad that a formal search was necessary and we aimed to consider most representative andintuitive models rather than only the most influential ones. The surplus of papers on the aggregated level indicates that inthe applied literature, discussions of firm level determinants are currently the most influential.Reference Reason for consideration

Acemoglu et al. (2016) Aggregated/RelatednessAng (2010) Aggregated/Institutional EnvironmentAng (2011) Aggregated/Institutional EnvironmentBahar et al. (2014) Aggregated/RelatednessCaiani (2017) Theoretical model of CACastelnovo et al. (2018) Aggregated /Government ActivityCetrulo et al. (2019) Aggregated/Institutional EnvironmentChuang and Hobday (2013) Firm/AlliancesCriscuolo and Narula (2008) Theoretical model of CADasgupta (2012) Theoretical model of CADosi et al. (2006) Aggregated/Institutional EnvironmentDosi et al. (2017) Firm/Internal LearningDutrénit et al. (2019) Aggregated/Institutional EnvironmentFernandez Donoso (2014) Aggregated/Institutional EnvironmentFigueiredo (2008) Aggregated/Institutional EnvironmentFilippetti and Guy (2020) Aggregated/Institutional EnvironmentHoxha and Kleinknecht (2020) Aggregated/Institutional Environment, Government ActivityLandini and Malerba (2017) Theoretical model of CALebdioui (2019) Aggregated/Government ActivityMarengo et al. (2012) Theoretical model of CAMehri (2015) Aggregated/Government ActivityMolina-Domene and Pietrobelli (2012) Aggregated/Institutional EnvironmentNeffke et al. (2011) Aggregated/RelatednessNeffke et al. (2017) Aggregated/RelatednessPerez-Aleman and Alves (2017) Aggregated/Government ActivityPyka et al. (2019) Theoretical model of CARadosevic and Yoruk (2018) Aggregated/RelatednessSavin and Egbetokun (2016) Theoretical model of CAWersching (2010) Theoretical model of CAYu et al. (2015) Aggregated/Government Activity

Kcy

oopom

Acknowledgments

We are grateful to Manuel Scholz-Wäckerle, Muhamed Kudic,Torsten Heinrich, Jakob Kapeller, Steffen Müller and the other dis-ciplinary experts that gave us feedback on the manuscript and ourliterature search strategy. We thank the editor and two anonymousreferees for their meticulous reading of the manuscript and theircritical, constructive and detailed comments, which were extremelyvaluable during the revision process. We also acknowledge the feed-back given at seminars and conferences where this work has beenpresented, particularly the EAEPE Annual Conference, the RebuildingMacroeconomics Globalization Hub Workshop and the ICAE ResearchSeminar. We are particularly grateful to Kerstin Hötte for her extensivefeedback on the manuscript. All remaining errors are our own.

Appendix A. More details on CA mechanisms in endogenousgrowth models

Here, some of the central mechanisms underlying CA in endogenousgrowth models are summarized. Many of the referenced papers aresomewhat dated, yet conceptually they still provide the methodologicalbase for the standard treatment of CA in EGM. Thereby, they are stillrelevant for EGMs today. In the following, we focus on the mechanismsunderlying CA. Therefore, important contributions that characterize theempirical implications of these mechanisms, such as the skill biasednessof technological change, are not discussed explicitly. They can bethought of as consequences of the mechanisms mentioned here.

In one of the first contributions to the EGM literature, Rivera-Batizand Romer (1991) subsume CA under a very general definition oftechnology which, aside from technological capabilities, comprises ma-chinery, process knowledge and the quality of physical inputs. Here, CAtakes place as an immediate consequence of firms investing into R&D ormachines and it results in a reduction of the marginal cost of production

16

for the firm — the process of production becomes more efficient and

CA is conceptualized as a kind of process innovation. A similar treatmentdates back to Romer (1990), for whom the carriers of capabilities areskilled workers who benefit from the knowledge accumulated by oldergenerations. In effect, CA, which again gets triggered by R&D invest-ment, involves the process of inter-generational knowledge spillovers,and takes place in a cumulative way, where current generations buildupon and expand the stock of technological capabilities accumulatedby previous generations. Weitzman (1998) introduces a variant of thissetup by treating labor as a fixed factor in the production function, butthe level of knowledge is treated as endogenous (i.e. 𝑌 = 𝐹 (𝐾,𝐴)).

nowledge expands as a consequence of R&D investments and the re-ombination of the existing set of ideas, thereby providing a different,et complementary view on how CA improves productivity.

A slightly different perspective is offered by the models in the spiritf Grossman and Helpman (1991a,b), where CA is, again, the resultf investment into R&D, but where it leads to the production of newroduct varieties. Thus, for firms, CA pays off not by the reductionf production costs but because of temporary monopoly profits on thearket. This mechanism is more similar to product innovation, although

it is only about the development of new varieties of existing prod-ucts, rather than the invention of brand new products.17 Temporarymonopoly rents are also the main implication of CA in the modelsbuilding upon Aghion and Howitt (1992), where R&D investmentslead to the production of better products, which then eliminate theirpredecessors. Because of the resulting competition between incumbentand entrant firms, as well as the resemblance of a process of creativedestruction, these models are often labeled as ‘Schumpeterian models’.Here, CA is something firms must strive for in order to survive in themarket.

17 This distinction has relatively little impact on the mathematical structureof the models, yet it represents very different aspects of the real worldinnovation and CA process, which is why we consider it important to highlightthis difference.

Research Policy 50 (2021) 104258C. Gräbner et al.

erhcdao

A

sebafpd

R

A

A

A

A

A

A

A

A

A

AA

A

A

A

A

B

B

B

B

B

In much of the newer literature, the main interest of researchershas shifted to the explanation of uneven CA on the global level.Starting with the contributions of Howitt (2000) and Acemoglu et al.(2007), the investigation of barriers to CA, such as incomplete contracts,xtractive institutions or social conflicts, is now an active area ofesearch. While these models are less concerned with the question ofow CA takes place, their treatment of barriers to CA allows to deriveonjectures on these processes as well. The same applies to modelsealing with the mismatch between skills and technology (e.g. Acemoglund Zilibotti, 2001; Acemoglu, 2002b), which also consider processesf learning-by-doing.

ppendix B. Papers added during the informal literature search

The papers mentioned in Table B.8 were not added via the formalearch process described in Section 4.1 of the main paper. They wereither recommended by experts from theory and practice, or wererought to our attention since they were cited regularly by the liter-ture identified via the formal search process (‘snowball sampling’). Inact, the latter search strategy often appears to be a more viable ap-roach in identifying the relevant literature compared to conventionalatabase searching (Pawson, 2006).

eferences

cemoglu, D., 2002a. Directed technical change. Rev. Econom. Stud. 69, 781–809.http://dx.doi.org/10.1111/1467-937X.00226.

cemoglu, D., 2002b. Technical change, inequality, and the labor market. J. Econ. Lit.40, 7–72. http://dx.doi.org/10.1257/0022051026976.

cemoglu, D., 2009. Introduction to Modern Economic Growth. Princeton UniversityPress, Princeton.

cemoglu, D., Akcigit, U., Kerr, W.R., 2016. Innovation network. Proc. Natl. Acad. Sci.113, 11483–11488. http://dx.doi.org/10.1073/pnas.1613559113.

cemoglu, D., Antràs, P., Helpman, E., 2007. Contracts and technology adoption. Amer.Econ. Rev. 97, 916–943. http://dx.doi.org/10.1257/aer.97.3.916.

cemoglu, D., Zilibotti, F., 2001. Productivity differences. Q. J. Econ. 116, 563–606.http://dx.doi.org/10.1162/00335530151144104.

ghion, P., Howitt, P., 1992. A model of growth through creative destruction.Econometrica 60, 323–351. http://dx.doi.org/10.3386/w3223.

haronson, B.S., Schilling, M.A., 2016. Mapping the technological landscape: Measuringtechnology distance, technological footprints, and technology evolution. Res. Policy45, 81–96. http://dx.doi.org/10.1016/j.respol.2015.08.001.

kgün, A.E., Keskin, H., Byrne, J.C., Aren, S., 2007. Emotional and learning capabilityand their impact on product innovativeness and firm performance. Technovation27, 501–513. http://dx.doi.org/10.1016/j.technovation.2007.03.001.

llen, R.C., 2011. Global Economic History. Oxford University Press, Oxford.ng, J.B., 2010. Financial reforms, patent protection, and knowledge accumulation in

India. World Dev. 38, 1070–1081. http://dx.doi.org/10.1016/j.worlddev.2009.12.011.

ng, J.B., 2011. Financial development, liberalization and technological deepening. Eur.Econ. Rev. 55, 688–701. http://dx.doi.org/10.1016/j.euroecorev.2010.09.004.

scani, A., Balland, P.A., Morrison, A., 2020. Heterogeneous foreign direct investmentand local innovation in Italian Provinces. Struct. Change Econ. Dyn. 53, 388–401.http://dx.doi.org/10.1016/j.strueco.2019.06.004.

tkinson, A.B., Stiglitz, J.E., 1969. A new view of technological change. Econ. J. 79,573–578. http://dx.doi.org/10.2307/2230384.

udretsch, D.B., Belitski, M., 2020. The limits to collaboration across four of the mostinnovative UK industries. Br. J. Manag. 31, 830–855. http://dx.doi.org/10.1111/1467-8551.12353.

ahar, D., Hausmann, R., Hidalgo, C., 2014. Neighbors and the evolution of the com-parative advantage of nations: Evidence of international knowledge diffusion? J.Int. Econ. 92, 111–123. http://dx.doi.org/10.1016/j.jinteco.2013.11.001.

alland, P., Boschma, R., Crespo, J., Rigby, D.L., 2019. Smart specialization policyin the European Union: relatedness, knowledge complexity and regional diver-sification. Reg. Stud. 53, 1252–1268. http://dx.doi.org/10.1080/00343404.2018.1437900.

arney, J., 1991. Firm resources and sustained competitive advantage. J. Manag. 17,99–120. http://dx.doi.org/10.1177/014920639101700108.

ell, M., Figueiredo, P.N., 2012. Innovation capability building and learning mech-anisms in latecomer firms: recent empirical contributions and implications forresearch. Can. J. Dev. Stud. 33, 14–40. http://dx.doi.org/10.1080/02255189.2012.677168.

lecker, R.A., Setterfield, M., 2019. Heterodox Macroeconomics: Models of Demand,Distribution and Growth. Edward Elgar, Cheltenham, UK.

17

Bogers, M., Foss, N.J., Lyngsie, J., 2018. The human side of open innovation: Therole of employee diversity in firm-level openness. Res. Policy 47, 218–231. http://dx.doi.org/10.1016/j.respol.2017.10.012.

Boschma, R., 2005. Proximity and innovation: A critical assessment. Reg. Stud. 39,61–74. http://dx.doi.org/10.1080/0034340052000320887.

Brem, A., Nylund, P.A., Schuster, G., 2016. Innovation and de facto standardization:The influence of dominant design on innovative performance, radical innovation,and process innovation. Technovation 50–51, 79–88. http://dx.doi.org/10.1016/j.technovation.2015.11.002.

Bunge, M., 2004. How does it work? The search for explanatory mechanisms. Phil. Soc.Sci. 34, 182–210. http://dx.doi.org/10.1177/0048393103262550.

Büschgens, T., Bausch, A., Balkin, D.B., 2013. Organizational culture and innovation:A meta-analytic review. J. Prod. Innov. Manage. 30, 763–781. http://dx.doi.org/10.1111/jpim.12021.

Caiani, A., 2017. Innovation dynamics and industry structure under different techno-logical spaces. Ital. Econ. J. 3, 307–341. http://dx.doi.org/10.1007/s40797-017-0049-z.

Caiani, A., Catullo, E., Gallegati, M., 2019. The effects of alternative wage regimes in amonetary union: A multi-country agent based-stock flow consistent model. J. Econ.Behav. Organ. 162, 389–416. http://dx.doi.org/10.1016/j.jebo.2018.12.023.

Caridi-Zahavi, O., Carmeli, A., Arazy, O., 2016. The influence of CEOs’ visionaryinnovation leadership on the performance of high-technology ventures: The me-diating roles of connectivity and knowledge integration. J. Prod. Innov. Manage.33, 356–376. http://dx.doi.org/10.1111/jpim.12275.

Castellacci, F., Natera, J.M., 2013. The dynamics of national innovation systems: Apanel cointegration analysis of the coevolution between innovative capability andabsorptive capacity. Res. Policy 42, 579–594. http://dx.doi.org/10.1016/j.respol.2012.10.006.

Castelnovo, P., Florio, M., Forte, S., Rossi, L., Sirtori, E., 2018. The economic impact oftechnological procurement for large-scale research infrastructures: Evidence fromthe Large Hadron Collider at CERN. Res. Policy 47, 1853–1867. http://dx.doi.org/10.1016/j.respol.2018.06.018.

Cetrulo, A., Cirillo, V., Guarascio, D., 2019. Weaker jobs, weaker innovation. Exploringthe effects of temporary employment on new products. Appl. Econ. 51, 6350–6375.http://dx.doi.org/10.1080/00036846.2019.1619015.

Chang, H., 2003. Kicking away the Ladder: Development Strategy in HistoricalPerspective. Anthem Press, London.

Chang, J., Bai, X., Li, J.J., 2015. The influence of leadership on product and processinnovations in China: The contingent role of knowledge acquisition capability. Ind.Mark. Manag. 50, 18–29. http://dx.doi.org/10.1016/j.indmarman.2015.04.014.

Chang, W., Taylor, S.A., 2016. The effectiveness of customer participation in newproduct development: A meta-analysis. J. Mark. 80, 47–64. http://dx.doi.org/10.1509/jm.14.0057.

Chen, V.Z., Li, J., Shapiro, D.M., Zhang, X., 2014. Ownership structure and innovation:An emerging market perspective. Asia Pac. J. Manag. 31, 1–24. http://dx.doi.org/10.1007/s10490-013-9357-5.

Chen, Y., Vanhaverbeke, W., Du, J., 2016. The interaction between internal R & Dand different types of external knowledge sourcing: an empirical study of Chineseinnovative firms. R & D Manag. 46, 1006–1023. http://dx.doi.org/10.1111/radm.12162.

Chesbrough, H., 2003. Open Innovation. Harvard Business School Press, Boston, MA.Chuang, Y., Hobday, M., 2013. Technological upgrading in Taiwan’s TFT-LCD indus-

try: signs of a deeper absorptive capacity? Technol. Anal. Strateg. Manag. 25,1045–1066. http://dx.doi.org/10.1080/09537325.2013.832748.

Ciarli, T., Lorentz, A., Valente, M., Savona, M., 2018. Structural changes and growthregimes. J. Evol. Econ. 29, 119–176. http://dx.doi.org/10.1007/s00191-018-0574-4.

Coccia, M., Watts, J., 2020. A theory of the evolution of technology: Technologicalparasitism and the implications for innovation magement. J. Eng. Technol. Manag.55, 101552. http://dx.doi.org/10.1016/j.jengtecman.2019.11.003.

Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: A new perspective on learningand innovation. Adm. Sci. Q. 35, 128–152. http://dx.doi.org/10.2307/2393553.

Costinot, A., Rodriguez-Clare, A., Werning, I., 2020. Micro to macro: Optimal tradepolicy with firm heterogeneity. Econometrica 88, 2739–2776. http://dx.doi.org/10.3982/ecta14763.

Coviello, N.E., Joseph, R.M., 2012. Creating major innovations with customers: Insightsfrom small and Young technology firms. J. Mark. 76, 87–104. http://dx.doi.org/10.1509/jm.10.0418.

Criscuolo, P., Narula, R., 2008. A novel approach to national technological accumula-tion and absorptive capacity: aggregating Cohen and Levinthal. Eur. J. Dev. Res.20, 56–73. http://dx.doi.org/10.1080/09578810701853181.

Dasgupta, K., 2012. Learning and knowledge diffusion in a global economy. J. Int.Econ. 87, 323–336. http://dx.doi.org/10.1016/j.jinteco.2011.11.012.

Dawid, H., Delli Gatti, D., 2018. Agent-based macroeconomics. In: Hommes, C.,LeBaron, B. (Eds.), Handbook of Computational Economics. Springer, pp. 63–156.

Dawid, H., Harting, P., Neugart, M., 2018. Cohesion policy and inequality dynamics:Insights from a heterogeneous agents macroeconomic model. J. Econ. Behav. Organ.150, 220–255. http://dx.doi.org/10.1016/j.jebo.2018.03.015.

Dawid, H., Harting, P., van der Hoog, S., Neugart, M., 2019. Macroeconomics withheterogeneous agent models: fostering transparency, reproducibility and replication.J. Evol. Econ. 29, 467–538. http://dx.doi.org/10.1007/s00191-018-0594-0.

Research Policy 50 (2021) 104258C. Gräbner et al.

de Leeuw, T., Lokshin, B., Duysters, G., 2014. Returns to alliance portfolio diversity:The relative effects of partner diversity on firm’s innovative performance andproductivity. J. Bus. Res. 67, 1839–1849. http://dx.doi.org/10.1016/j.jbusres.2013.12.005.

De Noni, I., Ganzaroli, A., Orsi, L., 2017. The impact of intra- and inter-regionalknowledge collaboration and technological variety on the knowledge productivityof European regions. Technol. Forecast. Soc. Change 117, 108–118. http://dx.doi.org/10.1016/j.techfore.2017.01.003.

Dosi, G., Fagiolo, G., Roventini, A., 2010. Schumpeter meeting Keynes: A policy-friendlymodel of endogenous growth and business cycles. J. Econom. Dynam. Control 34,1748–1767. http://dx.doi.org/10.1016/j.jedc.2010.06.018.

Dosi, G., Grazzi, M., Mathew, N., 2017. The cost-quantity relations and the diversepatterns of learning by doing: Evidence from India. Res. Policy 46, 1873–1886.http://dx.doi.org/10.1016/j.respol.2017.09.005.

Dosi, G., Lamperti, F., Mazzucato, M., Napoletano, M., Roventini, A., 2018. TheEntrepreneurial State at Work: an Agent Based Exploration. IsIGrowth WorkingPaper 41. URL: http://www.isigrowth.eu/wp-content/uploads/2018/07/working_paper_2018_41.pdf.

Dosi, G., Marengo, L., Pasquali, C., 2006. How much should society fuel the greed ofinnovators? Res. Policy 35, 1110–1121. http://dx.doi.org/10.1016/j.respol.2006.09.003.

Dosi, G., Nelson, R.R., 2018. Technological advance as an evolutionary process.In: Nelson, R.R., Dosi, G., Helfat, C.E., Pyka, A., Saviotti, P.P., Lee, K.,Dopfer, K., Malerba, F., Winter, S.G. (Eds.), Modern Evolutionary Economics:An Overview. Cambridge University Press, pp. 35–84. http://dx.doi.org/10.1017/9781108661928.002.

Dosi, G., Roventini, A., 2019. More is different. and complex! The case for agent-basedmacroeconomics. J. Evol. Econ. 29, 1–37. http://dx.doi.org/10.1007/s00191-019-00609-y.

Dosi, G., Roventini, A., Russo, E., 2019. Endogenous growth and global divergencein a multi-country agent-based model. J. Econom. Dynam. Control 101, 101–129.http://dx.doi.org/10.1016/j.jedc.2019.02.005.

Dutrénit, G., Natera, J.M., Puchet Anyul, M., Vera-Cruz, A.O., 2019. Developmentprofiles and accumulation of technological capabilities in Latin America. Technol.Forecast. Soc. Change 145, 396–412. http://dx.doi.org/10.1016/j.techfore.2018.03.026.

Edler, J., Fier, H., Grimpe, C., 2011. International scientist mobility and the locus ofknowledge and technology transfer. Res. Policy 40, 791–805. http://dx.doi.org/10.1016/j.respol.2011.03.003.

Edmondson, A.C., Harvey, J.F., 2018. Cross-boundary teaming for innovation: Integrat-ing research on teams and knowledge in organizations. Human Resour. Manag.Rev. 28, 347–360. http://dx.doi.org/10.1016/j.hrmr.2017.03.002.

Engelman, R.M., Fracasso, E.M., Schmidt, S., Zen, A.C., 2017. Intellectual capital,absorptive capacity and product innovation. Manag. Decis. 55, 474–490. http://dx.doi.org/10.1108/MD-05-2016-0315.

Enkel, E., Heil, S., 2014. Preparing for distant collaboration: Antecedents to potentialabsorptive capacity in cross-industry innovation. Technovation 34, 242–260. http://dx.doi.org/10.1016/j.technovation.2014.01.010.

Felekoglu, B., Moultrie, J., 2014. Top management involvement in new productdevelopment: A review and synthesis. J. Prod. Innov. Manage. 31, 159–175.http://dx.doi.org/10.1111/jpim.12086.

Fernandez Donoso, J., 2014. Do complex inventions need less international patentprotection? Econom. Lett. 125, 278–281. http://dx.doi.org/10.1016/j.econlet.2014.09.026.

Figueiredo, P.N., 2008. Government policies and sources of latecomer firms’ capabilitybuilding: A learning story from Brazil. Oxford Development Studies 36, 59–88.http://dx.doi.org/10.1080/13600810701848177.

Filipescu, D.A., Prashantham, S., Rialp, A., Rialp, J., 2013. Technological innovationand exports: Unpacking their reciprocal causality. J. Int. Mark. 21, 23–38. http://dx.doi.org/10.1509/jim.12.0099.

Filippetti, A., Guy, F., 2020. Labor market regulation, the diversity of knowledgeand skill, and national innovation performance. Res. Policy 49, 103867. http://dx.doi.org/10.1016/j.respol.2019.103867.

Foley, D., Michl, T., Tavani, D., 2019. Growth and Distribution, second ed. HarvardUniversity Press, Cambridge, MA, London, UK.

Forés, B., Camisón, C., 2016. Does incremental and radical innovation performance de-pend on different types of knowledge accumulation capabilities and organizationalsize? J. Bus. Res. 69, 831–848. http://dx.doi.org/10.1016/j.jbusres.2015.07.006.

Foss, N.J., Laursen, K., Pedersen, T., 2011. Linking customer interaction and innovation:The mediating role of new organizational practices. Organ. Sci. 22, 980–999.http://dx.doi.org/10.1287/orsc.1100.0584.

Frankort, H.T., 2016. When does knowledge acquisition in R & D alliances increasenew product development? The moderating roles of technological relatedness andproduct-market competition. Res. Policy 45, 291–302. http://dx.doi.org/10.1016/j.respol.2015.10.007.

Freeman, C., 1995. The ‘National System of Innovation’ in historical perspective. Camb.J. Econ. 19, 5–24. http://dx.doi.org/10.1093/oxfordjournals.cje.a035309.

Fu, X., Pietrobelli, C., Soete, L., 2011. The role of foreign technology and indigenousinnovation in the emerging economies: Technological change and catching-up.World Dev. 39, 1204–1212. http://dx.doi.org/10.1016/j.worlddev.2010.05.009.

18

Gabaix, X., Landier, A., 2008. Why has CEO pay increased so much? Q. J. Econ. 123,49–100. http://dx.doi.org/10.1162/qjec.2008.123.1.49.

Ganco, M., 2017. NK model as a representation of innovative search. Res. Policy 46,1783–1800. http://dx.doi.org/10.1016/j.respol.2017.08.009.

Gao, Y., Zheng, J., 2020. The impact of high-speed rail on innovation: An empiricaltest of the companion innovation hypothesis of transportation improvement withChina’s manufacturing firms. World Dev. 127, 104838. http://dx.doi.org/10.1016/j.worlddev.2019.104838.

García, F., Jin, B., Salomon, R., 2013. Does inward foreign direct investment improvethe innovative performance of local firms? Res. Policy 42, 231–244. http://dx.doi.org/10.1016/j.respol.2012.06.005.

García-Quevedo, J., Pellegrino, G., Vivarelli, M., 2014. R & D drivers and age: Areyoung firms different? Res. Policy 43, 1544–1556. http://dx.doi.org/10.1016/j.respol.2014.04.003.

Gräbner, C., Hafele, J., 2020. The emergence of core–periphery structuresin the European Union: a complexity perspective. ZOE Discussion Pa-pers. URL: https://zoe-institut.de/wp-content/uploads/2020/09/zoe-dp6-graebner-hafele-core--periphery.pdf.

Gräbner, C., Heimberger, P., Kapeller, J., Schütz, B., 2020. Is the Eurozone disintegrat-ing? Macroeconomic divergence, structural polarisation, trade and fragility. Camb.J. Econ. 49, 647–669. http://dx.doi.org/10.1093/cje/bez059.

Gräbner, C., Hornykewycz, A., 2021. Capability accumulation and product innovation:an agent-based perspective. J. Evol. Econ. (forthcoming).

Gräbner, C., Kapeller, J., 2015. New perspectives on institutionalist pattern modeling:Systemism, complexity, and agent-based modeling. J. Econ. Issues 49, 433–440.http://dx.doi.org/10.1080/00213624.2015.1042765.

Greco, M., Grimaldi, M., Cricelli, L., 2016. An analysis of the open innovation effecton firm performance. Eur. Manag. J. 34, 501–516. http://dx.doi.org/10.1016/j.emj.2016.02.008.

Greco, M., Grimaldi, M., Cricelli, L., 2017. Hitting the nail on the head: Exploringthe relationship between public subsidies and open innovation efficiency. Technol.Forecast. Soc. Change 118, 213–225. http://dx.doi.org/10.1016/j.techfore.2017.02.022.

Grossman, G.M., Helpman, E., 1991a. Innovation and Growth in the Global Economy.MIT Press, Cambridge, MA.

Grossman, G.M., Helpman, E., 1991b. Quality ladders in the theory of growth. Rev.Econom. Stud. 68, 43–61. http://dx.doi.org/10.2307/2298044.

Haus-Reve, S., Fitjar, R.D., Rodríguez-Pose, A., 2019. Does combining different typesof collaboration always benefit firms? Collaboration, complementarity and productinnovation in Norway. Res. Policy 48, 1476–1486. http://dx.doi.org/10.1016/j.respol.2019.02.008.

Heij, C.V., Volberda, H.W., Bosch, F.A.J.V.d., Hollen, R.M.A., 2020. How to leveragethe impact of R & D on product innovation? The moderating effect of managementinnovation. R & D Manag. 50, 277–294. http://dx.doi.org/10.1111/radm.12396.

Heinrich, T., 2013. Technological Change and Network Effects in Growth Regimes.Routledge, London and New York.

Helfat, C.E., Peteraf, M.A., 2003. The dynamic resource-based view: capabilitylifecycles. Strateg. Manag. J. 24, 997–1010. http://dx.doi.org/10.1002/smj.332.

Herstad, S.J., Sandven, T., Ebersberger, B., 2015. Recruitment, knowledge integrationand modes of innovation. Res. Policy 44, 138–153. http://dx.doi.org/10.1016/j.respol.2014.06.007.

Hidalgo, C., Balland, P., Boschma, R., Delgado, M., Feldman, M., Frenken, K.,Glaeser, E., He, C., Kogler, D.F., Morrison, A., Neffke, F., Rigby, D., Stern, S.,Zheng, S., Zhu, S., 2018. The principle of relatedness. In: Morales, A.J., Ger-shenson, C., Braha, D., Minai, A.A., Bar-Yam, Y. (Eds.), Unifying Themes inComplex Systems, Vol. IX. Springer International Publishing, Cham, pp. 451–457.http://dx.doi.org/10.1007/978-3-319-96661-8_46.

Hidalgo, C., Hausmann, R., 2009. The building blocks of economic complexity. Proc.Natl. Acad. Sci. 106, 10570–10575. http://dx.doi.org/10.1073/pnas.0900943106.

Hidalgo, C., Klinger, B., Barabási, A.L., Hausmann, R., 2007. The product spaceconditions the development of nations. Science 317, 482–487. http://dx.doi.org/10.1126/science.1144581.

Hötte, K., 2020. How to accelerate green technology diffusion? Directed technologicalchange in the presence of coevolving absorptive capacity. Energy Econ. 85, 104565.http://dx.doi.org/10.1016/j.eneco.2019.104565.

Howitt, P., 2000. Endogenous growth and cross-country income differences. Amer.Econ. Rev. 90, 829–846. http://dx.doi.org/10.1257/aer.90.4.829.

Hoxha, S., Kleinknecht, A., 2020. When labour market rigidities are useful forinnovation. Evidence from german IAB firm-level data. Res. Policy 49, 104066.http://dx.doi.org/10.1016/j.respol.2020.104066.

Jean, R.J.B., Sinkovics, R.R., Hiebaum, T.P., 2014. The effects of supplier involvementand knowledge protection on product innovation in customer–supplier relation-ships: A study of global automotive suppliers in China. J. Prod. Innov. Manage.31, 98–113. http://dx.doi.org/10.1111/jpim.12082.

Kafouros, M., Love, J.H., Ganotakis, P., Konara, P., 2020. Experience in R & Dcollaborations, innovative performance and the moderating effect of differentdimensions of absorptive capacity. Technol. Forecast. Soc. Change 150, 119757.http://dx.doi.org/10.1016/j.techfore.2019.119757.

Kafouros, M., Wang, C., Piperopoulos, P., Zhang, M., 2015. Academic collaborationsand firm innovation performance in China: The role of region-specific institutions.Res. Policy 44, 803–817. http://dx.doi.org/10.1016/j.respol.2014.11.002.

Research Policy 50 (2021) 104258C. Gräbner et al.

Kaplan, G., Violante, G.L., 2018. Microeconomic heterogeneity and macroeconomicshocks. J. Econ. Perspect. 32, 167–194. http://dx.doi.org/10.1257/jep.32.3.167.

Kauffman, S.A., 1993. The Origins of Order: Self-Organization and Selection inEvolution. Oxford University Press.

Kavusan, K., Noorderhaven, N.G., Duysters, G.M., 2016. Knowledge acquisition andcomplementary specialization in alliances: The impact of technological overlap andalliance experience. Res. Policy 45, 2153–2165. http://dx.doi.org/10.1016/j.respol.2016.09.013.

Khan, A., Thomas, J.K., 2008. Idiosyncratic shocks and the role of nonconvexitiesin plant and aggregate investment dynamics. Econometrica 76, 395–436. http://dx.doi.org/10.1111/j.1468-0262.2008.00837.x.

Kim, Y.K., Lee, K., Park, W.G., Choo, K., 2012. Appropriate intellectual propertyprotection and economic growth in countries at different levels of development.Res. Policy 41, 358–375. http://dx.doi.org/10.1016/j.respol.2011.09.003.

Kobarg, S., Stumpf-Wollersheim, J., Welpe, I.M., 2018. University-industry collabora-tions and product innovation performance: the moderating effects of absorptivecapacity and innovation competencies. J. Technol. Transf. 43, 1696–1724. http://dx.doi.org/10.1007/s10961-017-9583-y.

Kobarg, S., Stumpf-Wollersheim, J., Welpe, I.M., 2019. More is not always better:Effects of collaboration breadth and depth on radical and incremental innovationperformance at the project level. Res. Policy 48, 1–10. http://dx.doi.org/10.1016/j.respol.2018.07.014.

Kwak, K., Yoon, H., 2020. Unpacking transnational industry legitimacy dynamics,windows of opportunity, and latecomers’ catch-up in complex product systems.Res. Policy 49, 103954. http://dx.doi.org/10.1016/j.respol.2020.103954.

Landini, F., Malerba, F., 2017. Public policy and catching up by developing countriesin global industries: a simulation model. Camb. J. Econ. 41, 927–960. http://dx.doi.org/10.1093/cje/bex017.

Lasagni, A., 2012. How can external relationships enhance innovation in SMEs? Newevidence for Europe. J. Small Bus. Manag. 50, 310–339. http://dx.doi.org/10.1111/j.1540-627X.2012.00355.x.

Laursen, K., Masciarelli, F., Prencipe, A., 2011. Regions matter: How localized socialcapital affects innovation and external knowledge acquisition. Organ. Sci. 23,177–193. http://dx.doi.org/10.1287/orsc.1110.0650.

Lebdioui, A., 2019. Chile’s export diversification since 1960: A free market miracle ormirage? Dev. Change 50, 1624–1663. http://dx.doi.org/10.1111/dech.12545.

Lee, K., Kim, B.Y., 2009. Both institutions and policies matter but differently fordifferent income groups of countries: Determinants of long-run economic growthrevisited. World Dev. 37, 533–549. http://dx.doi.org/10.1016/j.worlddev.2008.07.004.

Lee, K., Malerba, F., 2018. Economic catch-up by latecomers as an evolutionaryprocess. In: Nelson, R.R., Dosi, G., Helfat, C.E., Pyka, A., Saviotti, P.P., Lee, K.,Dopfer, K., Malerba, F., Winter, S.G. (Eds.), Modern Evolutionary Economics: AnOverview. Cambridge University Press, pp. 172–207. http://dx.doi.org/10.1017/9781108661928.006.

Lee, J.J., Yoon, H., 2015. A comparative study of technological learning and orga-nizational capability development in complex products systems: Distinctive pathsof three latecomers in military aircraft industry. Res. Policy 44, 1296–1313. http://dx.doi.org/10.1016/j.respol.2015.03.007.

Leiblein, M.J., Madsen, T.L., 2009. Unbundling competitive heterogeneity: incentivestructures and capability influences on technological innovation. Strateg. Manag.J. 30, 711–735. http://dx.doi.org/10.1002/smj.746.

Li, L., Li, D., Goerzen, A., Shi, W., 2018. What and how do SMEs gain by goinginternational? A longitudinal investigation of financial and intellectual resourcegrowth. J. World Bus. 53, 817–834. http://dx.doi.org/10.1016/j.jwb.2018.07.001.

Love, J.H., Roper, S., 2015. SME innovation, exporting and growth: A reviewof existing evidence. Int. Small Bus. J. 33, 28–48. http://dx.doi.org/10.1177/0266242614550190.

Love, J.H., Roper, S., Vahter, P., 2014. Learning from openness: The dynamics ofbreadth in external innovation linkages. Strateg. Manag. J. 35, 1703–1716. http://dx.doi.org/10.1002/smj.2170.

Lundvall, B.A. (Ed.), 1995. National Systems of Innovation: Towards a Theory ofInnovation and Interactive Learning, paperback ed. Pinter, London.

Lundvall, B.A., 2007. National innovation systems–analytical concept and developmenttool. Ind. Innov. 14, 95–119. http://dx.doi.org/10.1080/13662710601130863.

Marengo, L., Pasquali, C., Valente, M., Dosi, G., 2012. Appropriability, patents, andrates of innovation in complex products industries. Econ. Innov. New Technol. 21,753–773. http://dx.doi.org/10.1080/10438599.2011.644666.

Mazzucato, M., 2014. The Entrepreneurial State: Debunking Public Vs. Private SectorMyths, revised ed. In: Anthem Frontiers of Global Political Economy, Anthem Press,London, New York.

Mazzucato, M., Semieniuk, G., 2017. Public financing of innovation: new questions.Oxford Rev. Econ. Policy 33, 24–48. http://dx.doi.org/10.1093/oxrep/grw036.

Mehri, D.B., 2015. The role of engineering consultancies as network-centred actorsto develop indigenous, technical capacity: the case of Iran’s automotive industry.Socio-Econ. Rev. 13, 747–769. http://dx.doi.org/10.1093/ser/mwu041.

Melitz, M.J., 2003. The impact of trade on intra-industry reallocations and aggregateindustry productivity. Econometrica 71, 1695–1725. http://dx.doi.org/10.1111/1468-0262.00467.

19

Molina-Domene, M.A., Pietrobelli, C., 2012. Drivers of technological capabilities indeveloping countries: An econometric analysis of Argentina, Brazil and Chile.Struct. Change Econ. Dyn. 23, 504–515. http://dx.doi.org/10.1016/j.strueco.2011.11.003.

Nakata, C., Im, S., 2010. Spurring cross-functional integration for higher new productperformance: A group effectiveness perspective. J. Prod. Innov. Manage. 27,554–571. http://dx.doi.org/10.1111/j.1540-5885.2010.00735.x.

Neffke, F., Henning, M., Boschma, R., 2011. How do regions diversify over time?Industry relatedness and the development of new growth paths in regions. Econ.Geogr. 87, 237–265. http://dx.doi.org/10.1111/j.1944-8287.2011.01121.x.

Neffke, F., Otto, A., Weyh, A., 2017. Inter-industry labor flows. J. Econ. Behav. Organ.142, 275–292. http://dx.doi.org/10.1016/j.jebo.2017.07.003.

Nelson, R.R. (Ed.), 1993. National Innovation Systems: A Comparative Analysis. OxfordUniversity Press, New York.

Nelson, R.R., 2008. What enables rapid economic progress: What are the neededinstitutions? Res. Policy 37, 1–11. http://dx.doi.org/10.1016/j.respol.2007.10.008.

Nelson, R.R., Winter, S., 1982. An Evolutionary Theory of Economic Change. BelknapPress of Harvard Univ. Press, Cambridge, MA.

Nieto, M.J., Santamaría, L., 2007. The importance of diverse collaborative networks forthe novelty of product innovation. Technovation 27, 367–377. http://dx.doi.org/10.1016/j.technovation.2006.10.001.

Ottonello, P., Winberry, T., 2020. Financial heterogeneity and the investment chan-nel of monetary policy. Econometrica 88, 2473–2502. http://dx.doi.org/10.3982/ecta15949.

Park, B.J.R., Srivastava, M.K., Gnyawali, D.R., 2014. Walking the tight rope ofcoopetition: Impact of competition and cooperation intensities and balance onfirm innovation performance. Ind. Mark. Manag. 43, 210–221. http://dx.doi.org/10.1016/j.indmarman.2013.11.003.

Pawson, R., 2006. Evidence-Based Policy: A Realist Perspective. SAGE, London;Thousand Oaks, Calif.

Perez-Aleman, P., Alves, F.C., 2017. Reinventing industrial policy at the frontier:catalysing learning and innovation in Brazil. Camb. J. Regions Econ. Soc. 10,151–171. http://dx.doi.org/10.1093/cjres/rsw038.

Petralia, S., Balland, P.A., Morrison, A., 2017. Climbing the ladder of technologicaldevelopment. Res. Policy 46, 956–969. http://dx.doi.org/10.1016/j.respol.2017.03.012.

Phene, A., Fladmoe-Lindquist, K., Marsh, L., 2006. Breakthrough innovations in the U.S.biotechnology industry: the effects of technological space and geographic origin.Strateg. Manag. J. 27, 369–388. http://dx.doi.org/10.1002/smj.522.

Pyka, A., Kudic, M., Müller, M., 2019. Systemic interventions in regional innovationsystems: entrepreneurship, knowledge accumulation and regional innovation. Reg.Stud. 53, 1321–1332. http://dx.doi.org/10.1080/00343404.2019.1566702.

Radosevic, S., Yoruk, E., 2018. Technology upgrading of middle income economies:A new approach and results. Technol. Forecast. Soc. Change 129, 56–75. http://dx.doi.org/10.1016/j.techfore.2017.12.002.

Rengs, B., Scholz-Wäckerle, J., 2020. Evolutionary macroeconomic assessment ofemployment and innovation impacts of climate policy packages. J. Econ. Behav.Organ. 169, 332–368. http://dx.doi.org/10.1016/j.jebo.2019.11.025.

Ribeiro, R.S., McCombie, J.S., Lima, G.T., 2016. Exchange rate, income distribution andtechnical change in a balance-of-payments constrained growth model. Rev. Polit.Econ. 28, 545–565. http://dx.doi.org/10.1080/09538259.2016.1205819.

Rivera-Batiz, L.A., Romer, P.M., 1991. Economic integration and endogenous growth.Q. J. Econ. 106, 531–555. http://dx.doi.org/10.2307/2937946.

Romer, P.M., 1990. Endogenous technological change. J. Polit. Econ. 98, S71–S102.http://dx.doi.org/10.1086/261725.

Salerno, M.S., Gomes, L.A.d.V., Silva, D.O.d., Bagno, R.B., Freitas, S.L.T.U., 2015.Innovation processes: Which process for which project? Technovation 35, 59–70.http://dx.doi.org/10.1016/j.technovation.2014.07.012.

Santamaría, L., Nieto, M.J., Barge-Gil, A., 2009. Beyond formal R & D: Taking advantageof other sources of innovation in low- and medium-technology industries. Res.Policy 38, 507–517. http://dx.doi.org/10.1016/j.respol.2008.10.004.

Santos Arteaga, F.J., Tavana, M., Di Caprio, D., Toloo, M., 2019. A dynamic multi-stage slacks-based measure data envelopment analysis model with knowledgeaccumulation and technological evolution. European J. Oper. Res. 278, 448–462.http://dx.doi.org/10.1016/j.ejor.2018.09.008.

Savin, I., Egbetokun, A., 2016. Emergence of innovation networks from R & Dcooperation with endogenous absorptive capacity. J. Econom. Dynam. Control 64,82–103. http://dx.doi.org/10.1016/j.jedc.2015.12.005.

Schaefer, K.J., 2020. Catching up by hiring: The case of Huawei. J. Int. Bus. Stud. 51,1500–1515. http://dx.doi.org/10.1057/s41267-019-00299-5.

Schneckenberg, D., Truong, Y., Mazloomi, H., 2015. Microfoundations of innovativecapabilities: The leverage of collaborative technologies on organizational learningand knowledge management in a multinational corporation. Technol. Forecast. Soc.Change 100, 356–368. http://dx.doi.org/10.1016/j.techfore.2015.08.008.

Schubert, T., Tavassoli, S., 2019. Product innovation and educational diversity in topand middle management teams. Acad. Manag. J. 63, 272–294. http://dx.doi.org/10.5465/amj.2017.0741.

Setterfield, M., 2002. A model of Kaldorian traverse: cumulative causation, structuralchange and evolutionary hysteresis. In: Setterfield, M. (Ed.), The Economics ofDemand-Led Growth. Edward Elgar, Cheltenham, UK, pp. 215–233.

Research Policy 50 (2021) 104258C. Gräbner et al.

Silverberg, G., Verspagen, B., 2005. A percolation model of innovation in complextechnology spaces. J. Econom. Dynam. Control 29, 225–244. http://dx.doi.org/10.1016/j.jedc.2003.05.005.

Song, Y., Gnyawali, D.R., Srivastava, M.K., Asgari, E., 2018. In search of pre-cision in absorptive capacity research: A synthesis of the literature andconsolidation of findings. J. Manag. 44, 2343–2374. http://dx.doi.org/10.1177/0149206318773861.

Stojčić, N., Srhoj, S., Coad, A., 2020. Innovation procurement as capability-building:Evaluating innovation policies in eight Central and Eastern European countries. Eur.Econ. Rev. 121, 103330. http://dx.doi.org/10.1016/j.euroecorev.2019.103330.

Sutton, J., 2012. Competing in Capabilities. Oxford University Press, Oxford, UK.Syverson, C., 2011. What determines productivity? J. Econ. Lit. 49, 326–365. http:

//dx.doi.org/10.1257/jel.49.2.326.Teece, D.J., 2017. A capability theory of the firm: an economics and (Strategic)

management perspective. New Zealand Econ. Pap. 53, 1–43. http://dx.doi.org/10.1080/00779954.2017.1371208.

Teece, D.J., Pisano, G., Shuen, A., 1997. Dynamic capabilities and strategic man-agement. Strateg. Manag. J. 18, 509–533. http://dx.doi.org/10.1002/(SICI)1097-0266(199708)18:7<509::AID-SMJ882>3.0.CO;2-Z.

Tellis, G.J., Prabhu, J.C., Chandy, R.K., 2009. Radical innovation across nations: Thepreeminence of corporate culture. J. Mark. 73, 3–23. http://dx.doi.org/10.1509/jmkg.73.1.003.

Todo, Y., Matous, P., Inoue, H., 2016. The strength of long ties and the weakness ofstrong ties: Knowledge diffusion through supply chain networks. Res. Policy 45,1890–1906. http://dx.doi.org/10.1016/j.respol.2016.06.008.

Tur, E.M., Azagra-Caro, J.M., 2018. The coevolution of endogenous knowledge networksand knowledge creation. J. Econ. Behav. Organ. 145, 424–434. http://dx.doi.org/10.1016/j.jebo.2017.11.023.

Vaona, A., Pianta, M., 2008. Firm size and innovation in European manufacturing.Small Bus. Econ. 30, 283–299. http://dx.doi.org/10.1007/s11187-006-9043-9.

20

Vermeulen, B., Pyka, A., 2014. Technological progress and effects of (Supra) regionalinnovation and production collaboration. An agent-based model simulation study.In: 2014 IEEE Conference on Computational Intelligence for Financial Engineering& Economics. CIFEr, IEEE, London, UK, pp. 357–364. http://dx.doi.org/10.1109/CIFEr.2014.6924095.

Vlačić, E., Dabić, M., Daim, T., 2019. Exploring the impact of the level of absorptivecapacity in technology development firms. Technol. Forecast. Soc. Change 138,166–177. http://dx.doi.org/10.1016/j.techfore.2018.08.018.

Weitzman, M.L., 1998. Recombinant growth. Q. J. Econ. 113, 331–360. http://dx.doi.org/10.1162/003355398555595.

Wersching, K., 2010. Schumpeterian competition, technological regimes and learningthrough knowledge spillover. J. Econ. Behav. Organ. 75, 482–493. http://dx.doi.org/10.1016/j.jebo.2010.05.005.

Wu, J., 2012. Technological collaboration in product innovation: The role of marketcompetition and sectoral technological intensity. Res. Policy 41, 489–496. http://dx.doi.org/10.1016/j.respol.2011.09.001.

Wu, J., Si, S., Wu, X., 2016. Entrepreneurial finance and innovation: Informal debt asan empirical case. Strateg. Entrepreneurship J. 10, 257–273. http://dx.doi.org/10.1002/sej.1214.

Xie, Z., Li, J., 2018. Exporting and innovating among emerging market firms: Themoderating role of institutional development. J. Int. Bus. Stud. 49, 222–245.http://dx.doi.org/10.1057/s41267-017-0118-4.

Yu, X., Dosi, G., Lei, J., Nuvolari, A., 2015. Institutional change and productivitygrowth in China’s manufacturing: the microeconomics of knowledge accumulationand creative restructuring. Ind. Corp. Change 24, 565–602. http://dx.doi.org/10.1093/icc/dtv011.

Zhuang, E., Chen, G., Feng, G., 2011. A network model of knowledge accumulationthrough diffusion and upgrade. Physica A 390, 2582–2592. http://dx.doi.org/10.1016/j.physa.2011.02.043.

Zou, T., Ertug, G., George, G., 2018. The capacity to innovate: a meta-analysis ofabsorptive capacity. Innovation 20, 87–121. http://dx.doi.org/10.1080/14479338.2018.1428105.