Shaping Innovation: Can
Industrial Policies Boost
Patent Applications?
Sandra Baquie, Yueling Huang, Florence Jaumotte, Jaden Kim,
Rafael Machado Parente, and Samuel Pienknagura
WP/26/47
IMF Working Papers describe research in
progress by the author(s) and are published to
elicit comments and to encourage debate.
The views expressed in IMF Working Papers are
those of the author(s) and do not necessarily
represent the views of the IMF, its Executive Board,
or IMF management.
2026
MAR
INTERNATIONAL MONETARY FUND 2
© 2026 International Monetary Fund WP/26/47
IMF Working Paper
Research Department
Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Prepared by Sandra Baquié, Yueling Huang
Florence Jaumotte, Jaden Kim, Rafael Machado Parente, and Samuel Pienknagura
Authorized for distribution by Antonio Spilimbergo
March 2026
IMF Working Papers describe research in progress by the author(s) and are published to elicit
comments and to encourage debate. The views expressed in IMF Working Papers are those of the
author(s) and do not necessarily represent the views of the IMF, its Executive Board, or IMF management.
ABSTRACT: This paper presents a global empirical analysis of how industrial policies (IPs) affect patent
applications, with an instrumental-variable strategy that addresses selection in policy targeting by leveraging
retaliatory dynamics. On average, IPs do not increase domestic patent applications over a four-year period,
except when they target sectors with potential distortions or externalities, such as infant industries or low-carbon
technologies. However, IPs temporarily boost foreign patent filings within the same timeframe, consistent with
strategic front-loading by foreign inventors seeking to secure technology protection, and perhaps market access,
in the IP-targeted sector. This link between foreign patent applications and IPs is stronger for export-oriented
policies compared to domestic subsidies, for IPs targeting innovation-central sectors, and in emerging markets
and developing economies.
RECOMMENDED CITATION: Baquie, Sandra, Yueling Huang, Florence Jaumotte, Jaden Kim, Rafael
Machado Parente, and Samuel Pienknagura. 2026. “Shaping Innovation: Can Industrial Policies Boost Patent
Applications?”, IMF Working Paper No. 2026/47.
JEL Classification Numbers: L52, O25, O31, O33, L14, Q55
Keywords:
Industrial Policies, Innovation, Patents, Networks, Low-carbon
technology
Author’s E-Mail Address:
sbaquie@, yhuang5@, fjaumotte@, jkim6@,
rmachadoparente@, spienknagura@
INTERNATIONAL MONETARY FUND 3
WORKING PAPERS
Shaping Innovation: Can Industrial
Policies Boost Patent
Applications?
Prepared by Sandra Baquie, Yueling Huang, Florence Jaumotte, Jaden
Kim, Rafael Machado Parente, and Samuel Pienknagura*
* The authors are grateful to Reka Juhász and Nathan Lane for kindly sharing the data in Juhasz et al. (2025). They also thank Pierre-
Olivier Gourinchas and Antonio Spilimbergo for their excellent feedback and guidance throughout the project, and Rui Mano and
IDB’s EconNet seminar participants for their valuable comments. The views expressed in this paper are those of the authors and
should not be attributed to the International Monetary Fund, its Executive Board, or its management. All errors are our own.
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Contents
1 Introduction 5
2 Data and Summary Statistics 9
Industrial Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Patents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Innovation network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Infant industries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Methods 14
Local projections-IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Heterogeneity and targeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4 Results 17
Protectionist industrial policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Liberalizing industrial policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Industrial policies by country income group . . . . . . . . . . . . . . . . . . . . . . . 20
Targeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Conclusion 27
A Additional Data Description 33
B Additional Results 36
C Robustness checks 42
INTERNATIONAL MONETARY FUND 4
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
1 Introduction
Industrial policies (IPs)–defined as targeted interventions to change the structure of economic ac-
tivity (Juhász et al., 2024)–have taken center stage in the global policy agenda. After a period of
decline during the liberalization wave of the 1990s, IPs have seen a resurgence since 2017 in both
advanced and emerging economies (Evenett et al., 2024). In addition to competitiveness objectives,
the recent wave of IPs began as countries grappled with heightened geopolitical tensions, calls for
self-reliance in industries critical for national security (for example, semiconductors), increased
vulnerabilities in global value chains, and a need to accelerate the green transition.
The recent resurgence of IPs has reignited policy debates on their effectiveness, including their
potential to foster innovation, one of the main engines of productivity growth. In theory, IPs can
boost innovation by supporting investments in R&D, overcoming market failures, and mitigating
coordination problems. Yet, IPs may also reallocate resources in a non-welfare-enhancing way by
supporting targeted firms at the expense of more productive or innovative ones, exert negative
externalities across sectors, or distort countries’ relative competitiveness. In practice, the literature
suggests a mixed track record for IPs (Aghion et al., 2015; Harrison, 2024; Lane, .). Two
examples are often alluded to when presenting arguments against and in favor of IPs: import-
substitution IPs in Latin America and export-oriented IPs in East Asia. The former was linked to
limited long-run productivity growth (Rodrigues, 2010), whereas the latter likely supported rapid
export-led growth and productivity improvements (Krueger, 1997; Rodrik, 2009; Hu and Jefferson,
2009; Cherif and Hassanov, 2019; Branstetter et al., 2005). These contrasting cases illustrate
the considerable heterogeneity in the effectiveness of IPs documented in the literature. They also
suggest that, although IPs hold potential, their effectiveness may depend on a large set of conditions,
including appropriate targeting and design, institutional quality, implementation conditions, and
cross-sectoral and cross-country spillovers (Aghion et al., 2015; Harrison, 2024; Juhász and Lane,
2024b,a; OECD, 2025; Baquié et al., 2025; IMF, 2025; Hodge et al., 2024).
We contribute to this debate by providing, to the best of our knowledge, the first global cross-
sectoral empirical analysis of the impact of IPs on innovation. Leveraging the novel classification
of IPs proposed by Juhász et al. (2025), who identify IPs within the Global Trade Alert (GTA)
database (Evenett and Fritz, 2020), and the sector- and country-level patent data from the INPACT-
S dataset (LaBelle et al., 2024), we assemble a unique dataset containing information on patents
and IPs for 177 countries, 31 manufacturing sectors, from 2009 to 2019. This global data enables
us to shed light on the effectiveness of IPs in shaping innovation.
Our methodology employs the local projection approach to estimate the dynamic response
of patent filings to the implementation of IPs at various horizons. To address the endogeneity
of IP targeting–the possibility that policymakers implement IPs in sectors already experiencing
an innovation boom–we pursue an instrumental variable (IV) strategy inspired by the shift-share
literature. Specifically, we construct a novel instrumental variable based on changes in IPs in other
sectors and politically distant countries. This approach allows for causal interpretation of the
INTERNATIONAL MONETARY FUND 5
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
local projection coefficients and addresses identification concerns in the earlier empirical literature
(Juhász and Lane, 2024b).
Our results, in line with the earlier theoretical or country- or sector-specific empirical literature,
point to a nuanced relationship between IPs and innovation. First, both protectionist and liber-
alizing IPs–mainly lifting import barriers–have no significant effect on domestic patenting within
the considered four-year time span, suggesting that, on average, new innovation by domestic firms
takes more than four years to materialize and that targeting promising sector is essential to unleash
innovation. In addition, protectionist IPs are, on average, associated with a temporary increase
in patent applications from foreign inventors: one additional policy leads to a percent increase
in foreign patent filings within the first two years, but the effect dissipates thereafter. This quick
and short-lived boost in foreign patent applications suggests that foreign inventors may expedite
the filing of innovations already in the pipeline in response to protectionist policies, leading to a
front-loading of patent applications abroad rather than the generation of new innovations in the
considered time frame. Such behavior likely reflects efforts to secure intellectual property protection
or access to the market targeted by the policy. By contrast, liberalizing policies appear to generate
larger and more persistent effects on foreign patent applications. This rise in patenting may result
from firms gaining access to cheaper or higher-quality imported inputs that foster innovation or
from intensified product market competition following import liberalization that prompts firms to
patent more (Goldberg et al., 2009; Amiti and Konings, 2007; Bloom et al., 2011).
Beyond average effects, we find that the link between IPs and foreign patenting varies across
IP instruments and country income groups. We show that export incentives and domestic subsi-
dies are the primary drivers of the short-term increase in foreign patenting following protectionist
IPs. Export-oriented policies yield slightly higher effects, although the results are less precise. In
terms of country characteristics, the average effect of protectionist IPs appears to be driven by
emerging markets and developing economies (EMDEs), where protectionist IPs boost received for-
eign patent applications. Conversely, liberalizing IPs in advanced economies (AEs)–mainly lifting
import barriers–are associated with increased cross-border patenting by domestic inventors, with
AE-based inventors disseminating innovations worldwide.
Our results also highlight the role of targeting to spur domestic innovation by focusing on promis-
ing cases highlighted in the literature: infant industries, and in particular, low-carbon technologies
(LCTs) (Melitz, 2005; Aghion et al., 2015; Criscuolo et al., 2019; Mazzucato, 2011; Aghion et al.,
2016). The rationale for targeting infant– young, leveraged, and frontier– industries is that market
failures, such as knowledge spillovers, coordination externalities, or capital market imperfections,
may prevent them from reaching efficient scale (IMF, 2025). As such, IPs targeting those nascent
sectors are expected to be more effective since they address those market failures. Indeed, when
it comes to domestic patenting, our results suggest that IPs directed toward infant industries are
associated with increases in domestic patenting, consistent with evidence that targeted support can
relax financing constraints and accelerate innovation in early-stage, high-potential sectors (Aghion
et al., 2015; Criscuolo et al., 2019; Mazzucato, 2011). The rationale for IPs targeting low-carbon
INTERNATIONAL MONETARY FUND 6
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
technologies is also strong, as they are often in infant or upstream and distorted sectors, and they
also involve solving coordination failures and emission externalities (Huang et al., 2025; Baquié
et al., 2025). When it comes to innovation, climate-related IPs significantly increase domestic
patenting over the considered four-year horizon, likely by encouraging new entrants (Aghion et al.,
2016).
Finally, we investigate IPs’ impacts along the innovation network (Liu and Ma, 2021). To
do so, we construct innovation networks using patent citations across sectors. This allows us
to measure a sector’s innovation centrality. Consistent with the baseline findings, IPs do not
affect domestic innovation within four years, irrespective of whether they target innovation-central
sectors or not. However, IPs targeting innovation-central sectors–those with high centrality in
the country’s knowledge network–generate larger increases in foreign patenting, likely reflecting
patent optimization as foreign firms secure access and participation in frontier sectors that are
both supported by government policy and underpin a potential wide range of innovations. These
findings are consistent with the recent literature highlighting the role of learning by doing and
innovation networks (Liu and Ma, 2021; Aghion et al., 2025; Barwick et al., 2025).
While our analysis provides evidence on whether IPs are effective at boosting patent applications
in the targeted sectors, it remains a positive (descriptive and data-driven) assessment rather than
a normative evaluation of whether these policies are desirable overall. Indeed, innovation is only
one dimension of the multiple potential policy impacts of industrial policies. A full assessment
would require examining broader consequences, including effects on productivity, resource allocation
within and across sectors, and fiscal costs. Such an assessment would also need to consider potential
spillovers to non-targeted sectors, general equilibrium effects, cross-country effects, and whether
alternative policy instruments could achieve similar gains at a lower cost. Still, one may think that
industrial policies having a positive effect on the targeted sector, for instance, on its innovation,
may be a first indication that they may have an overall positive effect.
Related Literature: Our paper contributes to three strands of literature. First, we contribute to
the literature on the economic effects of IPs by examining its contribution to one of the primary
drivers of growth: innovation. Theoretically, markets may under-invest in innovation due to exter-
nalities, market failures, and coordination problems, thereby creating room for strategic government
interventions (Rodrik, 2009; Aghion et al., 2015). In practice, skepticism emerged from concerns
about resource misallocation, poor targeting, rent-seeking, and capture (Krueger, 1990; Pack and
Saggi, 2006). The early empirical literature on the effects of industrial policy offers mixed insights,
reflecting methodological challenges, identification issues, and heterogeneity in effects (Juhász and
Lane, 2024b). Nevertheless, successful case studies from East Asian economies, such as China,
Korea, and Taiwan Province of China, highlighted the potential for IPs to contribute to industrial
transformation, including through domestic innovation and knowledge transfers (World Bank, ed,
1993; Weiss, 2005; Branstetter et al., 2005; Chang, 2006; Hu and Jefferson, 2009; Aghion et al.,
2015).
INTERNATIONAL MONETARY FUND 7
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
We contribute to this literature by offering, to the best of our knowledge, the first global
and cross-sectoral empirical analysis of IPs and innovation. Prior research in this area has been
theoretical (Liu and Ma, 2021) or focused on specific countries, sectors, or policy instruments
(Barwick et al., 2024, 2025). Our analysis draws on the machine-learning-based classification of
IPs in Juhász et al. (2025), which identifies relevant interventions within the Global Trade Alert
(GTA) database (Evenett and Fritz, 2020), and combines it with sector- and country-level patenting
data to study innovation outcomes globally. Other recent studies have leveraged the GTA data to
examine the effects of IPs on other outcomes or in a country setting (Huang et al., 2025; Machado-
Parente et al., 2025; Baquié et al., 2025; Barwick et al., 2024). We also propose a new way to
address a limitation of the earlier empirical work–the endogeneity of policy targeting–by applying
an instrumental variable approach inspired by the shift-share literature.
Second, our analysis contributes to the literature on the targeting of industrial policies. Specif-
ically, we empirically examine two sectors with promising rationales for IP interventions according
to the literature: infant industries and low-carbon technologies (LCTs). Using ORBIS data, we
identify sectors–within specific countries–where firms are disproportionately young, leveraged, and
operating at the technological frontier. Prior research suggests that targeting such sectors could
enhance the effectiveness of IPs by relaxing financing constraints and facilitating early-stage in-
novation (Aghion et al., 2015; Criscuolo et al., 2019; Mazzucato, 2011; Melitz, 2005; IMF, 2025).
Our empirical results support this view, showing that protectionist IPs targeting infant industries
significantly boost domestic patenting activity. We then focus on a particular subset of infant
industries–LCTs– which theoretical and country studies identify as relevant targets for IPs due to
novelty-related coordination failures and emissions externalities (Aghion et al., 2016; Barwick et al.,
2025). Consistent with this rationale for effectiveness, our findings indicate that climate-motivated
IPs are particularly effective at spurring domestic innovation, notably through extensive-margin
effects.
Third, this paper contributes to the literature on cross-sectoral spillovers and innovation net-
works. We build on Liu and Ma’s model, showing that a social planner seeking to maximize long-run
growth would allocate R&D resources toward industries with high innovation centrality–that is, in-
dustries playing a central role in the economy’s innovation network–to take advantage of knowledge
spillovers for future growth (Liu and Ma, 2021). Garcia-Macia and Sollaci’s also suggest that tar-
geting sectors with high knowledge spillovers is essential for IPs to have positive welfare impacts
(Garcia-Macia and Sollaci, 2024). In this paper, we empirically find that, on average, IPs lead to a
temporary increase in foreign patenting that is larger in innovation-central sectors, likely as foreign
firms optimize their patenting to secure access to frontier sectors supported by government policy
and in which they anticipate a potential wide range of resulting innovations.
The rest of the paper is organized as follows. Section 2 describes the data and construction
of the variables. Section 3 presents the empirical strategy. Section 4 shows the main findings and
associated robustness exercises, and Section 5 concludes. Additional information, data descriptions,
and robustness results are provided in the Appendix.
INTERNATIONAL MONETARY FUND 8
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
2 Data and Summary Statistics
To assess whether industrial policies (IPs) stimulate innovation, we construct a novel panel dataset
at the country-year-sector level by merging information on IPs with patenting activity across 177
countries and 31 manufacturing sectors (ISIC 2-digit), covering the period 2009–2019. This Section
describes the construction and composition of the dataset.
Industrial Policies
Our measure of IP builds on the dataset developed by Juhász et al., who apply a machine learning
classifier to policy descriptions recorded in the Global Trade Alert (GTA) database between 2009
and 2022. The GTA, established in 2008, systematically compiles governments’ policy measures
and announcements that may disadvantage foreign commercial actors (Evenett and Fritz, 2020).
Juhász et al. define IPs as goal-oriented state interventions intended to shift the composition of
economic activity. This database contains allows for the quantification of the number of IPs for each
product (HS1992 product codes and CPC industry codes), in each country and year. To ensure
consistency over time, we implement a reporting-lag adjustment, as recommended by Juhász et al..
This adjustment retains only policies recorded in the same calendar year as their announcement,
thereby reducing artificial inflation in early-year policy counts resulting from continuous updates in
the GTA database. With this classification of IPs, we build a count of IPs in place in each country,
sector, and year. We use the World Integrated Trade Solutions’ (WITS) concordance table to link
the products targeted by IPs (HS1992 product and CPC industry codes) with their respective sector
(ISIC Rev. 2) (WITS, 2025). Non-reported policy counts are treated as zeros if the country has
never reported an IP in that sector since 2009.
To capture policy heterogeneity, we use GTA’s red-amber-green classification of discriminatory
intent. Our analysis focuses on protectionist IPs (“red” measures that “almost certainly discrimi-
nate against foreign commercial interests”), such as export subsidies, and liberalizing IPs (“green”
measures that “liberalize towards foreign commercial interests”), such as the removal of export bans
(Evenett and Fritz, 2020). Ambiguous (“amber”) measures are excluded from the analysis, but they
represent only 5 percent of IPs in GTA, compared to 63 percent for protectionist IPs and 32 per-
cent for liberalizing ones. Examples of policies by category are provided in Table . Figure 1A
shows that the use of IPs has increased sharply since 2017 across both AEs and EMDEs, consistent
with trends documented by Baquié et al. (2025), Machado-Parente et al. (2025), and Huang et al.
(2025)1.
We also classify IPs by policy instrument, based on GTA instrument categories aggregated
into five groups aligned with the UN MAST classification for non-tariff measures: (i) trade barri-
ers (export/import restrictions), (ii) domestic subsidies, (iii) export incentives, (iv) local content
requirements, and (v) other instruments (., public procurement or FDI measures), following
1Figure 1 leverages the data used for the regression analysis, as such IPs applying to several countries or sectors
are double counted. Although this counting method helps to capture IPs importance, it can lead to differences in IP
counts when compared to other papers that use a different sectoral definition.
INTERNATIONAL MONETARY FUND 9
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Goldberg et al. (2024); Evenett and Fritz (2020); Evenett et al. (2024). Figure 1B shows that, over
2009-2019, protectionist IPs were predominantly composed of subsidies (55%) and export incen-
tives (35%), while liberalizing IPs are primarily import liberalization measures (84%). Other IP
instruments have been less implemented historically, which reduces the available variation in our
dataset and the precision of the results for those instruments.
Figure 1: Distribution of the change in total active protectionist and liberalizing industrial policies,
by country income group and year (left) and by instrument (right).
A. Change in total active country-sector IPs over time,
in AEs and EMDEs
B. Change in active IPs between 2009 and
2019, by instrument
Notes: Panel A shows the year-to-year change in active country-sector IPs in AEs and EMDEs. IPs are counted as
they are in the analysis, represented by the first difference in the stock of IP policies. An IP applying to two sectors
will be counted twice. Similarly, IPs applying to several countries will be counted several times. Double counting
policies in this way enables to capture their spread. Panel B shows the composition of the change in total active IPs
between 2009 and 2019. In both panels, the purple color represents protectionist IPs, and the blue color represents
liberalizing IPs.
Sources: GTA (2022), Juhász et al. (2025), and authors’ calculations.
To study the targeting of IPs supporting LCTs, we follow a three-step classification of these
policies, combining information about targeted products and IPs’ description (Huang et al., 2025).
First, we rely on an extensive literature review to identify products (Harmonized System (HS) codes)
corresponding to LCTs (Pigato et al., 2020; OECD/Eurostat, 1999; Rosenow and Mealy, 2024;
Kowalski and Legendre, 2023; Goldschlag et al., 2020; Hasna et al., 2023; Mealy and Teytelboym,
2022). We code the related products as LCTs in the product-country-year dataset of Juhász
et al. (2025). Then, as we aggregate this dataset to the sector level, we define climate-related
IPs as those targeting sectors where at least 70 percent of products are LCTs. Second, we use
Huang et al. (2025)’s large language model (LLM) to detect motives related to climate mitigation
or environmental concerns in GTA policy descriptions (Evenett and Fritz, 2020; Evenett et al.,
2024). Third, in cases where the HS-based and LLM-based methods disagree, we conduct manual
validation to ensure accurate classification. This methodology is described in more depth in Huang
INTERNATIONAL MONETARY FUND 10
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
et al. (2025).
Several data limitations are worth noting. First, although GTA enables a cross-country analy-
sis, heterogeneity in countries’ reporting standards may bias IP counts. Second, the data begins in
2009, undercounting IP stocks in countries with pre-existing IP frameworks. These two concerns
are partially addressed by our methodology, which considers the effects of changes in the number of
active IPs rather than the stock level and includes fixed effects for country and year. Third, GTA
captures only policies affecting foreign commercial interests, omitting purely domestic or subna-
tional interventions, which may have potentially important implications in strongly decentralized
systems (Goldberg et al., 2024). We address this by excluding countries with a potentially large
number of subnational IPs in a robustness check (see Section ). Fourth, the dataset captures
the presence, not the intensity, of policies. While this limitation prevents us from discussing policy
magnitudes, recent analysis from the New Industrial Policy Observatory shows a positive correla-
tion between the IP count and the total value of subsidies in 2023 (Evenett et al., 2024), providing
some reassurance on the representativeness of our IP measure.
Patents
We gauge innovation activity by using patent applications across countries and sectors, building
on the INPACT-S dataset developed by LaBelle et al. (2024). The authors leverage the PATSTAT
database to quantify domestic patent applications and flows of patent applications from one country
to another in manufacturing sectors (ISIC Rev. 2) from 1980 to 2019 (LaBelle et al., 2024). To
ensure accurate attribution of patents to their country of origin, LaBelle et al. apply a fractional
counting method assigning patent applications to origin countries based on both the residence of
the applicant (which can be the firm owning the patent) and of the inventors (the individual or
team developing the technology). Regarding patent destination, when a patent is submitted to
a regional patent authority (such as the European Patent Office), the INPACT-S dataset uses a
weighted-dispersion method to distribute applications across individual member states, improving
spatial precision in measuring patent receipt. Moreover, INPACT-S includes all patent filings,
regardless of family size, capturing the complete set of innovations. As it covers all patents within
a family, this dataset enables us to trace the complete trajectory of an innovation rather than solely
the initial breakthrough. At the same time, because subsequent filings often reflect incremental
refinements, this approach makes it more difficult to isolate the original inventive step.
In our empirical analysis, we aggregate this information into three different country-sector-
year measures: domestic inventions–with the same country of invention and application–, patents
received from abroad, and patents submitted abroad. The resulting dataset spans 1980–2019,
covering 31 manufacturing industries and 212 countries. These indicators are then merged with
our IP dataset at the country-year-sector level, resulting in data that covers 180 countries, 31
manufacturing sectors, and the period from 2009 to 2019. As shown in Figure , China, the
USA, Japan, Korea, and Germany have the highest yearly number of patent applications over the
considered period. The three sectors receiving the most patent applications are chemistry, medical
INTERNATIONAL MONETARY FUND 11
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
and precision equipment, and computing machinery (Figure ).
Even if patent applications are widely used to measure innovation in the economics literature,
there are limitations to keep in mind when analyzing the results. First, differences in patent-
ing behaviors across countries and industries—some systematically patent-intensive, others rarely
patenting—could affect the comparability across sectors and countries. We partially address this
concern in our analysis by including sector and country fixed effects, as well as testing the robust-
ness of the results to the exclusion of the main patenting economies. Second, the count of patent
applications may not capture the quality of patents (Hall et al., 2001). Finally, as we will see in
the rest of the paper, patent applications can sometimes stem from strategic behaviors rather than
genuine inventions (Blind et al., 2006).
Innovation network
To describe the propagation of innovations in the economy, we leverage INPACT-S’s patent citation
data to build a directed innovation network at the sector level. Indeed, Liu and Ma show that
sectors with higher innovation centrality have a greater influence on future knowledge creation
across the economy. As a result, in their model, the innovation centrality vector coincides with the
growth-maximizing R&D allocation along a balanced growth path.
Formally, the innovation network is a matrix Ω, where the (i, j)-th element, wi,j , is the share of
sector i’s citations of sector j. In other words, wi,j represents how sector i’s innovation production
benefits from sector j’s existing knowledge stock. In this case, j is upstream of i, and i downstream
of j. We calculate innovation networks for each country c and year t and get wcts→s′ , the share
of patent citations in sector s made to sector s′. We calculate for each country c and year t, the
network’s innovation centrality, also known as eigenvector centrality and defined as the dominant
eigenvector of the Ω matrix. This measure captures the relative importance of each sector in
propagating innovation spillovers. We will use it to shed light on the importance of IP targeting.
The average innovation network across countries and years is represented in Figure 2. Each node
represents a sector (ISIC Rev 2, 2 digits). The citing sector is the arrow’s source, and the cited
sector is the head of the arrow. Citation-weighted edges capture the extent to which innovation in
one sector contributes to another sector as measured by wcts→s′ . The size of the dots represents the
sector’s innovation centrality. According to this representation, the most innovative central sector,
on average, is the chemical industry. It is widely cited in the following sectors: mining, food and
beverages, petroleum, textile, and publishing.
Infant industries
We follow the definition of infant industries proposed by IMF (2025), identifying them within each
country as sectors whose average firm age is below the national average and whose average firm
leverage and distance to the global productivity frontier are above their respective country averages.
The measures of firm age, leverage, and distance to the frontier are derived by aggregating firm-
level information from the Bureau van Dijk (BvD) Orbis global database. This large cross-country
INTERNATIONAL MONETARY FUND 12
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Figure 2: Average innovation network.
Notes: Each node represents a sector (ISIC Rev. 2, 2 digits), and citation-weighted edges capture the extent to
which innovation in one sector contributes to another sector. The size of the dots represents the sector’s innovation
centrality. The citing sector is the arrow’s source, and the cited sector is the head of the arrow. Only links with a
citation percentage higher than 5% are represented on the graph for readability.
Source: LaBelle et al. (2024).
firm-level database combines data from multiple sources—in particular, publicly available national
company registries—and harmonizes them into a consistent international format. The dataset used
in our analysis is derived from Orbis by applying the cleaning procedures described in Gopinath et
al. (2017) and IMF (2025) and selecting the following variables: firm age, total assets, operating
revenue (gross output), tangible and intangible fixed assets, material costs, liabilities, earnings
before interest and taxes, and cash flow. All variables are expressed in constant 2010 . dollars,
and sectoral total factor productivity, which underlies the measure of distance to the productivity
frontier, is calculated following Hsieh and Klenow (2009) and IMF (2025).
INTERNATIONAL MONETARY FUND 13
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
3 Methods
This Section describes our empirical strategy for estimating the effect of industrial policies (IPs) on
patenting activity. We begin by presenting our main regression specification, using the combined
local projection method with instrumental variable (IV) strategy to address endogeneity concerns.
Then, we present methodological extensions to explore heterogeneity, assess the importance of
targeting, and examine potential externalities along the innovation network.
Local projections-IV
Our baseline analysis builds on the local projection method developed by Jordà (2005) to estimate
the dynamic effect of IPs at time t on patent filings over multiple horizons h. The corresponding
regression equation is:
∆h log
(
PatentStockc,s,t+h + 10
−6)
= βh∆IPStockc,s,t + θ1 log
(
PatentStockc,s,t−1 + 10
−6)+ θ2IPStockc,s,t−1
+θ3Xc,s,t + λc,s + λc,t + λs,t + εc,s,t
(1)
where the dependent variable, ∆h log
(
PatentStockc,s,t+h + 10
−6), is the change over h years in the
log of the patent stock for country c and sector s, with a small constant added to retain observations
with zero values. ∆IPStockc,s,t is the change in either protectionist or liberalizing IP stock between
t and t-1 for country c and sector s. In the absence of IP removals, this corresponds to the count
of newly introduced policies. The coefficient of interest, βh, measures the association between one
additional active IP and the growth rate of the patent stock over h years. Following the local
projections method, we control for the stock of patents and IPs at t − 1. In addition, Xc,s,t are
controls including changes and past stocks of other IPs and non-IP trade policies. λc,s, λc,t, λs,t
are country-sector, country-year, and sector-year fixed effects that respectively control for global
shocks to different sectors (., sectoral trends), growth shocks in different countries, and sectoral
differences across countries. Standard errors are clustered at the country level. This conservative
level of clustering accounts for the correlation of IPs within-country and across sectors, as an IP
targeting products in several sectors is counted as one for each one of them.
Although the local projection specification includes an extensive set of controls and fixed effects
to capture potential omitted variables behind the decision of countries to implement IPs and past
dynamics in IPs, we may still not be able to interpret OLS estimates of βh causally. Indeed,
policymakers may target sectors based on their innovation trends, introducing a reverse causality
bias. For instance, if governments target sectors with increasing innovation activity, OLS estimates
of βh would be upward biased. The pre-trend estimates of our OLS estimations suggest that such
targeting may indeed be happening.
To tackle this issue, we combine the local projection approach with an instrumental variable
INTERNATIONAL MONETARY FUND 14
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
strategy, following Jordà and Taylor (2016). Our instrument for the change in protectionist IPs is
constructed from the change in protectionist IPs in other sectors and politically distant countries:
∑
c′ ̸=c
s′ ̸=s
(
PolDistc,c′,t∑
c′′ ̸=c PolDistc,c′′,t
)
∆ProtectionistIPStockc′,s′,t. (2)
Political distance is measured following Bailey et al. (2017), who use voting alignment in the
United Nations General Assembly for a given pair of countries and year to gauge political proxim-
ity/distance. The relevance of the instrument comes from the fact that countries adopt retaliatory
IPs in response to protectionist IPs in politically distant countries. We check for relevance by
reporting the first-stage KP rank Wald F-statistics throughout the paper and confirm they exceed
the conventional threshold of 10. In addition, the first stage coefficients are presented in Figure
. As expected, the instrument performs less well for IPs not driven by retaliation—such as those
targeting infant industries or with climate motives—so we estimate their effects using OLS 2
The exogeneity of the instrument relies on the assumption that IPs in other sectors (s′ ̸= s) and
politically distant countries (c′ ̸= c) are not related to innovation trends in sector s and country
c. Considering other sectors supports the exogeneity assumption, as IPs in unrelated sectors are
unlikely to generate patenting responses in the sector of interest, particularly when the country
implementing the IP is politically distant. A potential threat arises if the change in IPs in politically
distant countries is a response to the change in IP in country c. Since this is more likely for major
trading nations, we test the robustness of our results to excluding the top three global traders from
the sample (Section ).
For liberalizing IPs, we construct a similar instrument based on the change in active liberalizing
IPs in other countries and other sectors. However, to improve the strength of the first stage, we
add a second instrument: the change in active liberalizing IPs in other sectors and in the country’s
main trade partners (those accounting for over 90 percent of trade). The corresponding formula is:
∑
c′ ̸=c
s′ ̸=s
(
1(Main Trade Partner)c,c′,t∑
c′′ ̸=c 1(Main Trade Partner)c,c′′,t
)
∆LiberalizingIPStockc′,s′,t (3)
While these instruments are relevant, the exogeneity condition is slightly weaker, as liberalizing
IPs from trade partners could indirectly affect innovation in country c, even in another sector s,
if there are second-order cross-sectoral spillovers via trade, prices, or innovation. However, the
Hansen J-statistics (Figure ) and robustness checks (Section ) are reassuring.
Although the IV strategy enables a causal interpretation of the estimated coefficient, it identifies
a local average treatment effect (LATE), focusing on IPs influenced by the instrument, in our case,
retaliatory policies. This raises questions about external validity, as the targeting of retaliatory
2Since they apply to novel industries, the exogeneity of IP targeting with respect to innovation is more likely than
in the average case. The pre-trends are insignificant, supporting a causal interpretation of the effect.
INTERNATIONAL MONETARY FUND 15
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
IPs may differ from more strategic interventions, preventing a generalization of the results to non-
retaliatory IPs. This trade-off between identification and external validity is explained by Juhász
and Lane (2024b). Nonetheless, if retaliatory IPs are on average less well-targeted and, in turn,
less effective, our estimates would be a lower bound of the true average effects of all IPs–retaliatory
and non-retaliatory–(Juhász and Lane, 2024b).
Heterogeneity and targeting
To study heterogeneity, we extend Equation 1 in three ways. First, we split the sample by income
level—AEs and EMDEs—to assess differences across income groups. Second, we estimate separate
effects by policy instrument (., export incentives, subsidies), replacing the overall IP count with
instrument-specific counts and adjusting controls accordingly to account for all other policies3.
We still use all IPs in other sectors and countries as our instrumental variable, since retaliatory
responses do not necessarily rely on the same policy tools as the initial intervention. As shown in
Section , restricting the analysis to identical IP instruments, leaves the results unchanged.
Third, we examine the role of targeting by focusing on innovation-central sectors—defined as
those with eigenvector centrality above the 80th percentile in the country’s innovation network.
Indeed, theory suggests that protectionist IPs targeting innovation-central industries may have a
larger effect on innovation outcomes than those targeting peripheral ones4 (Liu and Ma, 2021;
Garcia-Macia and Sollaci, 2024). We test this hypothesis by including interaction terms between
IPs and a dummy for innovation-central sectors, hence comparing the effects of protectionist policies
on received patent applications in innovation central and non-central sectors. The same instruments
are used, but controls are expanded to account for all other policies (., IPs in non-innovation
central sectors, IPs in upstream and downstream sectors, other IPs, and other trade policies).
Fourth, we investigate IPs supporting the development of infant industries. Our measure of
infant industry is derived from firm-level Orbis data, as described in Section . The Orbis dataset
provides uneven coverage of forms across countries, sectors, and years; as a result, 31 percent of
the country–sector–year observations are missing for the infant industry measure. Missing values
are overall evenly distributed across sectors, though relatively niche industries–such as recycling,
tobacco manufacturing, and uranium mining–account for a disproportionate share. Data gaps are
also more pronounced in developing economies, which may bias the sample toward advanced coun-
tries, although emerging and developing economies remain relatively well represented. In addition,
coverage is somewhat thinner in earlier years (2009–2010), limiting the accuracy of regressions for
large horizons. Hence, the corresponding regression results are only presented up to horizon 3.
Despite these limitations, the resulting dataset identifies infant industries for 13,820 observations.
This allows us to test whether the effects of IPs on innovation are stronger in infant industries
by including an interaction term between IPs and a dummy variable indicating infant-industry
3We control for other (liberalizing or protectionist) policies involving this instrument, other IPs, and other non-
trade policies.
4Firms may optimize their patenting to secure access to these frontier sectors supported by government policy and
for which they anticipate a potential wide range of resulting innovations.
INTERNATIONAL MONETARY FUND 16
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
sectors, thereby comparing the impact of protectionist policies on patent applications across in-
fant and non-infant industries. In addition to the controls used in the baseline regression, we also
control for IPs in non-infant industry sectors. As policies targeting infant industries are most of-
ten designed for strategic purposes rather than as responses to foreign actions, our instrumental
variable approach has limited power in this context with low first-stage F statistics. Therefore, we
base this analysis on OLS estimates. Nonetheless, since such policies focus on promoting emerging
sectors, their introduction is plausibly unrelated to prior innovation patterns, and the timing of
their implementation is likely exogenous to innovation levels, supporting a causal interpretation of
the results5.
Finally, we dig deeper into one infant industry: low-carbon technologies 6. As mentioned in
the data section, we are able to classify industrial policies targeting LCTs, therefore keeping the
same number of observations and power as in the baseline regression. However, these policies
are typically strategic and not retaliatory, weakening the power of our IV approach. Therefore,
like with the infant industry regressions, we rely on OLS estimates for this analysis. Still, as with
infant industries, since climate-related IPs often aim to develop novel sectors, their targeting may be
uncorrelated with past innovation trends, and their timing may be exogenous relative to innovation,
which again, supports a causal interpretation7..
4 Results
This Section presents our findings on the effects of industrial policies (IPs) on patent activity.
We start by examining the effects of protectionist and liberalizing IPs on patent applications,
distinguishing between foreign and domestic inventors. We then explore how results vary across
policy instruments, income groups, and targeting strategies, before analyzing potential externalities
along the innovation network and assessing the robustness of our findings.
Protectionist industrial policies
On average, protectionist industrial policies do not have a significant effect on patent applications
over the considered four-year horizon, underscoring the importance of careful targeting to unlock
potential benefits. Estimates in Figure 3A are insignificant across all horizons, indicating a limited
response of domestic inventors to protectionist industrial policies in the first four years of the policy
intervention. Figure in the Annex also shows that patent applications submitted abroad my
domestic inventors does not change, in line with the null effect on domestic innovation. This lack
of response suggests that domestic inventors are likely to have already patented innovations in the
pipeline in their home country prior to policy implementation and therefore cannot accelerate filings
5This is supported by insignificant pre-trends in Figure 7
6In our dataset, sectors associated with LCTs–if 40% of their products are LCTs or if LCTs represent more than
10 percent of its products, when defining low-carbon technology products as in Huang et al. (2025) are associated
with percentage point more infant industry observations.
7This is supported by insignificant pre-trends in Figure 8
INTERNATIONAL MONETARY FUND 17
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
to benefit from the intervention. It is important to note, however, that these results do not rule
out possible effects over longer periods, as innovation cycles can last decades (Alston et al., 2023;
Wang et al., 2024).
Protectionist industrial policies are also associated with a short-lived increase in patent ap-
plication from foreign inventors. As illustrated in Figure 3B, the introduction of an additional
protectionist industrial policy is, on average, associated with a percent increase in the num-
ber of foreign patent applications in the targeted sector. However, this effect dissipates rapidly
and becomes statistically insignificant after the second year. This reversal suggests that foreign
inventors may expedite filing patent applications for innovations already in the pipeline in response
to policy implementation. In other words, IPs seem to encourage front-loading applications for
existing innovations rather than stimulating new ones. A plausible rationale is that firms seek to
secure market access or strengthen intellectual property protection in the targeted sector. Unlike
domestic inventors, their innovations non-patented abroad are “low-hanging fruits”. As such, in the
considered horizons, industrial policies likely fasten the pace of patent applications, but without
fostering foreign innovation. That being said, the average effects documented in Figure 3 also mask
crucial heterogeneity related to policy design and implementation. Section will show that even
in the considered 4-year horizon IPs targeting promising sectors can foster innovation, including by
domestic inventors.
Figure 3: Protectionist industrial policies and patent applications from domestic (left) and foreign
(right) inventors.
A. Domestic inventors
−5
0
5
10
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50
K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
B. Foreign inventors
−4
−2
0
2
4
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50
K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
Notes: The upper part of the Panels presents the main results. The y-axes represent the percentage change in patent
applications received from abroad in percent in Panel B. In Panel A, the y-axis represents the percentage change in
patent applications received from domestic inventors in percent. In both panels, the x-axis represents the considered
horizon in years. Year 0 is the year of IP implementation. The percentage changes are estimated following Equation
1 with 100× (exp(βh)− 1). Standard errors are clustered at the country level and the regression includes as well as
sector-year, country-year and country-sector fixed effects. The dashed lines represent 90 percent confidence intervals.
The histograms in the bottom parts of the Figure represent the first stage KP rank Wald F-statistic for the considered
horizon. The black line delineates F=10.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
Turning to heterogeneity across instruments, we find that subsidies and export incentives drive
the short-term average increase in foreign patent applications following the implementation of
INTERNATIONAL MONETARY FUND 18
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
protectionist industrial policies, with export-oriented instruments potentially yielding slightly larger
gains. Applying the local projection-IV specification to specific policy instruments, Figure 4A shows
that an additional protectionist subsidy is associated with a 2 percent increase in received foreign
patent applications in the first year. However, like the average effect, the effects of subsidies
are short-lived. Export incentives are associated with a slightly higher and more sustained effect
(Figure 4B). Note that since export incentives are less frequently implemented than subsidies, there
is less variation in the IP variable. Consequently, the results are more imprecise (and statistically
insignificant) and the first stage is weaker than with subsidies. Still, Appendix Figure confirms
the robustness of the findings under alternative clustering assumptions, with stronger statistical
significance and instrument relevance when standard errors are clustered at the country-sector
level.
These findings on the effectiveness of industrial policy tools are in agreement with the broader
literature on the role of industrial policies in supporting East Asia’s export-led growth model
(Cherif and Hassanov, 2019; Choi and Levchenko, 2025). According to this strand of the literature,
export-oriented IPs are particularly effective in helping firms overcome domestic market limitations,
facilitating the accumulation of technological capabilities and scale economies that would be diffi-
cult to attain through domestic markets alone (Reed, 2024). In line with this mechanism, recent
empirical evidence also points to more sustained effects of export-oriented IPs on firm performance
and trade competitiveness relative to subsidies (Machado-Parente et al., 2025; Huang et al., 2025).
Figure 4: Protectionist subsidies and export incentives and patent applications received from foreign
inventors.
A. Subsidies
−10
−5
0
5
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50
K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
B. Export incentives
−10
−5
0
5
10
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50
K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
Notes: The upper part of the Panels presents the main results. Panel A estimates the effect of one additional
protectionist subsidy and Panel B presents the effect of one additional export incentive. In both Panels, the y-axis
represents the percentage change in patent applications received from abroad in percent and the x-axis represents the
considered horizon in years. Year 0 is the year of IP implementation. The percentage changes are estimated following
Equation 1 with 100× (exp(βh)− 1). Standard errors are clustered at the country level and the regression includes
as well as sector-year, country-year and country-sector fixed effects. The dashed lines represent 90 percent confidence
intervals. The histograms in the bottom parts of the Figure represent the first stage KP rank Wald F-statistic for
the considered horizon. The black line delineates F=10.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
INTERNATIONAL MONETARY FUND 19
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Liberalizing industrial policies
Protectionist industrial policies are not always the most effective policy tool to boost innovation.
While well-targeted protectionist measures can yield short-term gains, liberalizing policies–mainly
lifting import barriers–appear to generate larger and more persistent effects. As shown in Figure
5B, lifting one additional policy is associated with a 4 percent increase in received foreign patent
applications after four years. This positive effect may reflect the removal of growth-constraining
policies, or enhanced trade-driven collaboration and investment opportunities through increased
competition and openness (Aghion et al., 2015; Harrison, 2024). By contrast, there is no evidence
of a significant impact of liberalizing IPs on patenting by domestic inventors within the same time
frame (Figure 5A), consistent with the literature indicating that innovation cycles typically span a
decade or more (Alston et al., 2023; Wang et al., 2024). Similarly to the results for protectionist
policies, these findings suggest that liberalization facilitates cross-border patenting rather than
domestic invention in the short to medium term.
The strong response of foreign patenting to liberalizing industrial policies–mainly lifting import
barriers–underscores the importance of trade flows in fostering long-term innovation. Indeed, as
noted by Harrison (2024), effective industrial policy design should satisfy four criteria: (i) Correct
a market failure, (ii) Consult with the private sector, (iii) Enhance competition, (iv) Conclude (for
instance, by including a sunset clause to progressively spur firms towards innovation and growth). In
this context, lifting import barriers may enhance medium-term cross-border patenting, potentially
reflecting technological transfers, by fulfilling criteria (iii), enhancing competition, and potentially
(iv) if they effectively end former protectionist policies, such as import barriers. These results
reinforce the broader point that industrial policy effectiveness depends on a wide range of factors,
including their design, the timing of their implementation and removal, but also their targeting–an
issue explored further in Section .
Industrial policies by country income group
The effect of industrial policies on patenting differ across AEs and EMDEs. Running the above
specification separately for AEs and EMDEs sheds light on this differential effect, despite reduced
statistical power and weaker first-stage relevance due to smaller sample sizes. Results show that
protectionist industrial policies have no significant impact on patent applications by domestic and
foreign inventors in AEs over the observed horizon (Figure in the Appendix). As shown in
Figure 1A, this may partly be due to the fact that, in AEs, most protectionist industrial policies
have been implemented after 2017, which prevents us from analyzing their impact beyond two years.
However, protectionist industrial policies are associated with an increase in received foreign patent
applications in EMDEs (Figure 6A), consistent with the interpretation that they accelerate the
transfer of existing technologies rather than foster novel innovation. Nevertheless, even if industrial
policies play a more catalytic role in EMDEs, AEs have been the destination of 74 percent of all
patents submitted by foreign inventors between 1990 and 2019, potentially due to the integration
INTERNATIONAL MONETARY FUND 20
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Figure 5: Liberalizing industrial policies and patent applications from domestic (left) and foreign
(right) inventors.
A. Foreign inventors
−15
−10
−5
0
5
10
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
B. Domestic inventors
−2
0
2
4
6
8
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
Notes: The upper part of the Panels presents the main results. In Panel A, the y-axis represents the percentage
change in patent applications received from abroad in percent. In Panel B, the y-axis represents the percentage
change in patent applications received from domestic inventors in percent. In both panels, the x-axis represents the
considered horizon in years. Year 0 is the year of IP implementation. The percentage changes are estimated following
Equation 1 with 100× (exp(βh)− 1). Standard errors are clustered at the country level and the regression includes
as well as sector-year, country-year and country-sector fixed effects. The dashed lines represent 90 percent confidence
intervals. The histograms in the bottom parts of the Figure represent the first stage KP rank Wald F-statistic for
the considered horizon. The black line delineates F=10.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
of global value chains and the size of their markets.
Liberalizing industrial policies promote cross-border patenting, particularly from AEs to other
countries. In EMDEs, liberalizing policies are not significantly associated with increased patenting
activity (Figure in the Appendix). This may be partly related to the low variation in liberalizing
IPs in EMDEs, as most of them have been implemented in AEs around 2015, as shown in Figure 1A.
In AEs, the effect on received foreign patent applications follows a similar trajectory to the aggregate
results in Figure 3B, though large standard errors—due to smaller sample size—limit statistical
significance. However, a clearer pattern emerges when examining patent applications submitted
abroad: liberalizing IPs significantly increase outward patenting by inventors based in AEs (Figure
6B). Between 1990 and 2021, AEs accounted for 96% percent of patent filled abroad. These findings
suggest that firms in AEs respond to trade liberalization by increasing cross-border patenting, likely
to leverage new market opportunities and adapt to heightened competitive pressures.
Targeting
Although on average, industrial policies (IPs) do not increase patenting by domestic inventors
within the targeted sector over the considered horizon, appropriately targeted policies may yield
different outcomes. A relevant policy question is, therefore: which sectors should be prioritized?
The literature provides guidance on targeting strategies, emphasizing the need to address market
failures and distortions, enhance competition, engage the private sector in their design, and avoid
INTERNATIONAL MONETARY FUND 21
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Figure 6: Industrial policies and patent applications, by country income group.
A. Protectionist IPs and patent applications re-
ceived from abroad in EMDEs.
−10
−5
0
5
10
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50
K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
B. Liberalizing IPs and patent applications sub-
mitted abroad in AEs.
−20
0
20
40
60
80
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
0
10
20
30
40
50
K
P
r
k
W
a
ld
F
−
s
ta
t
−3 −2 −1 0 1 2 3 4
Notes: The upper part of the Panels presents the main results. In Panel A, the y-axis represents the percentage
change in patent applications received from abroad in percent. Sample is limited to EMDEs. In Panel B, the y-axis
represents the percentage change in patent applications from domestic inventors submitted abroad in percent. Sample
is restricted to AEs. In both panels, the x-axis represents the considered horizon in years. Year 0 is the year of IP
implementation. The percentage changes are estimated following Equation 1 with 100×(exp(βh)−1). Standard errors
are clustered at the country level and the regression includes as well as sector-year, country-year and country-sector
fixed effects. The dashed lines represent 90 percent confidence intervals. The histograms in the bottom parts of the
Figure represent the first stage KP rank Wald F-statistic for the considered horizon. The black line delineates F=10.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
permanent support (Harrison, 2024). While this section only focus on maximizing innovation and
does not provide normative recommendations, it presents empirical evidence that well-targeted IPs
can support innovation. Informed by the literature and recent policy efforts, we focus on three
specific sectors: infant industries, low-carbon technologies, and innovation-central industries.
Infant industries: Industrial policies (IPs) targeting infant industries appear more effective at
fostering domestic innovation in the medium run than the average IP. As shown in Figure 7A,
these policies are followed by a statistically significant increase in patenting by domestic inventors
after two years, reaching about a one percent increase after two years. IPs applied to non-infant
industries are less strongly associated with domestic innovation, and IV estimates in Figure 3A
shows that the small and temporary increase in the OLS specification stems from endogeneity
and the targeting of sectors with upward trending innovation8. We further discuss the difference
between OLS and IV results for the average IP in Section .
These results are consistent with evidence showing that targeted support for young, financially
constrained, and high potential for learning-by-doing can generate positive innovation responses
(Aghion et al., 2015; Mazzucato, 2011). Different mechanisms may be at play. For instance, Akcigit
and Goldschlag (2023) find that inventors in large incumbent firms receive higher wages but tend
8As detailed in Section , OLS estimates are used to study infant-industry IPs because these policies are strategic
rather than retaliatory, limiting the strength of our instrument. The absence of upward pre-trends for infant industries
in Figure7 supports the plausibility of a causal interpretation or of an lower bound estimation.
INTERNATIONAL MONETARY FUND 22
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
to be less productive, suggesting limited innovation gains if dominant firms capture IPs’ gains.
Moreover, Machado-Parente et al. (2025) and Baquié et al. (2025) show that IPs have stronger
effects on younger and smaller firms, likely by alleviating financial frictions (Brandão-Marques and
Toprak, 2024).
Finally, IPs targeting infant-industries do not increase foreign patent applications, as presented
in Figure 7B. Since these industries are nascent, there are likely fewer immediate technological op-
portunities and foreign inventors likely have fewer pre-existing innovations in the pipeline. There-
fore, foreign inventors may not be able to secure access to the IP-impacted markets following policy
implementation, limiting their short-term patenting response compared to the average IP.
Figure 7: Protectionist industrial policies targeting infant (purple) or non-infant industries (grey)
and patent applications submitted by domestic or foreign inventors using OLS.
A. Domestic inventors.
−.5
0
.5
1
2
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3
Horizon (years)
B. Foreign inventors.
−5
0
5
10
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3
Horizon (years)
Notes: The y-axis represents the percentage change in patent applications submitted by domestic inventors (left) or
foreign inventors (right) in percent. The x-axis represents the considered horizon in years. The purple line shows the
effect of IPs targeting infant industries and the grey one the effect of IPs targeting non-infant industries. Definitions
are explained in Sections and . Year 0 is the year of IP implementation. The percentage changes are estimated
following Equation 1 with 100 × (exp(βh) − 1). The regression controls for non-protectionist IPs and non-IP trade
policies as well as sector-year, country-year and country-sector fixed effects. Standard errors are clustered at the
country level. The dashed lines represent 90 percent confidence intervals.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
Low-carbon technologies: Climate-related IPs are found to be more effective in fostering do-
mestic innovation than the average IP. As shown in Figure 8A, protectionist IPs with climate
objectives are associated with a gradual and statistically significant increase in patenting by do-
mestic inventors, reaching over percent three years after implementation. Figure in the
Appendix indicates that the coefficients on the extensive margin9 are significant, suggesting that
climate-related IPs foster the development of new innovation ecosystems in sectors previously un-
patented. By contrast, the average IP has a muted effect on domestic innovation as shown by
Figure 3A’s IV estimates that tackle the selection bias of non-climate-related IPs targeting sectors
9The extensive margin effects are estimated by using a dependent variable equal to 1 if the sector-year-country
has more than one patent and zero otherwise. As such, the corresponding regression estimates the effect of IPs on
developing unpatented sectors.
INTERNATIONAL MONETARY FUND 23
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
already experiencing innovation gains 10.
These findings align with sector-specific analyses. For instance, Barwick et al. (2024) show
that a 10 percent increase in financial incentives to electric vehicle (EV) producers results in a
4 percent rise in sector-specific patent applications. Likewise, Hasna et al. (2023) document a
positive association between environmental subsidies and green patenting activity. Sector-specific
characteristics of LCT industries–such as high learning-by-doing potential, increasing returns to
scale, and a high share of new entrants–may explain the larger effect of climate-related IPs compared
to the others (Bartelme et al., 2019; Garcia-Macia and Sollaci, 2024).
Our results suggest that the difference between climate-related and non-climate-related IPs is
insignificant when it comes to foreign patenting. Both policies facilitate foreign patenting to the
same extent. This is shown by the overlap of the two curves in Figure 8B and Figure of the
Appendix.
Figure 8: Protectionist climate-related industrial policies and patent applications submitted by
domestic or foreign inventor using OLS.
A. Domestic inventors.
−.5
0
.5
1
2
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4 5
Horizon (years)
Climate−related
Non−climate−related
B. Foreign inventors.
−1
−.5
0
.5
1
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4 5
Horizon (years)
Climate−related
Non−climate−related
Notes: The y-axis represents the percentage change in patent applications submitted by domestic inventors (left)
or foreign inventors (right) in percent. The x-axis represents the considered horizon in years. The green line shows
the effect of climate-related IPs and the brown one the effect of non-climate related IPs. Definitions are explained
in Section . Year 0 is the year of IP implementation. The percentage changes are estimated following Equation
1 with 100 × (exp(βh) − 1). The regression controls for non-protectionist IPs and non-IP trade policies as well as
sector-year, country-year and country-sector fixed effects. Standard errors are clustered at the country level. The
dashed lines represent 90 percent confidence intervals.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
Innovation-central industries: Our empirical results align with Liu and Ma and Garcia-Macia and
Sollaci’s theoretical models, as they confirm that the positive and temporary effects of industrial
policies on foreign patenting are larger in innovation-central industries. Figure 9 shows that one
10As explained in Section , our analysis of LCT-targeting IPs relies on OLS estimation of local projections
because our proposed instrument is weak due to climate-related IPs not being often implemented as a retaliatory
response. While this limits causal inference due to potential endogeneity—particularly the concern that LCTs may
be targeted precisely because they show innovation promise—the absence of pre-trends in Figure 8 supports the
plausibility of a causal interpretation.
INTERNATIONAL MONETARY FUND 24
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
additional protectionist IP targeting an innovation-central industry is associated with a percent
increase in received foreign patent applications, nearly double the average effect. Similarly, one ad-
ditional liberalizing industrial policy applied to an innovation-central industry is associated with a
percent significant increase in received foreign patent applications after 4 years. These patterns
could potentially stem from foreign firms strategically adjusting their patenting behavior to secure
access to frontier industries that receive government support and are expected to generate a broad
range of future innovations. In contrast, IPs applied to non-central industries show no statistically
significant impact. If anything, liberalizing IPs temporarily decrease received patent applications,
potentially as non-central firms in the innovation network adjust to a more competitive environ-
ment. As such, the average results presented in Figures 3B and 5B mask meaningful heterogeneity:
large and significant effects in innovation-central industries and insignificant effects in non-central
ones. This finding implies that, for industrial policies to be conducive to cross-border patenting,
policymakers may need to prioritize industries that are central in the country’s innovation network.
Figure 9: Protectionist (left) and liberalizing (right) industrial policies and received patent appli-
cations from foreign inventors, by centrality of the targeted sector (color)
A. Protectionist IPs and patent applications re-
ceived from foreign inventors, by innovation cen-
trality of the targeted industry.
−20
0
20
40
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
Targets central sector
Targets non−central sector
B. Liberalizing IPs and patent applications re-
ceived from foreign inventors, by innovation cen-
trality of the targeted industry.
−40
−20
0
20
P
e
rc
e
n
ta
g
e
c
h
a
n
g
e
(
%
)
−3 −2 −1 0 1 2 3 4
Horizon (years)
Targets central sector
Targets non−central sector
Notes: The y-axis represents the percentage change in patent applications received from abroad in percent. The x-axis
represents the considered horizon in years. The purple and light blue curves show the effect of IPs targeting innovation-
central sectors and gray and dark blue curves the effects of policies targeting non-central industries. Industries are
innovation central if their eigenvalue centrality in their country’s innovation network is above the 80th percentile of
the country’s innovation centrality distribution, as explained in Section . Year 0 is the year of IP implementation.
The percentage changes are estimated following Equation 1 with 100 × (exp(βh) − 1). The regression controls for
IPs implemented in upstream and downstream sectors as well as sector-year, country-year and country-sector fixed
effects. The instrumented variable is the one of interest. Standard errors are clustered at the country level. The
dashed lines represent 90 percent confidence intervals.
Sources: LaBelle et al. (2024), GTA (2022), Juhász et al. (2025), and authors’ calculations.
Robustness
We conduct a series of robustness checks to validate the above results. These tests address potential
concerns related to instrument validity, sample composition, inference methods, policy heterogene-
INTERNATIONAL MONETARY FUND 25
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
ity, network definitions, and estimation strategies.
Hansen J-test: Since liberalizing IPs are instrumented with two instrumental variables, we test for
over-identifying restrictions with the Hansen J-test. As shown in Figures and , we cannot
reject the null hypothesis across all horizons that the over-identifying restrictions are valid11.
Excluding major trading/patenting economies: The absence of pre-trends when using the IV strat-
egy supports the idea that IPs implemented at the same time as IPs in other sectors and politically
distant countries are not targeted based on promising innovation trends. However, the instrument
could capture reverse causality if the considered country is a major trading economy implementing
an IP, to which other countries react with IPs in other sectors. This would weaken the exogene-
ity argument of the instrument. To address this concern, we exclude the three largest trading
economies—China, the United States, and Germany—from the estimation. Results remain robust
(Figure ), mitigating concerns about the validity of the IV strategy. Since these major trading
ecomies are also the main patenting ones, these results also tackle potential concern related to pos-
sible measurement biases related to the underreporting of sub-national IPs in the GTA database
and the over-reporting of patents in countries with aggressive patenting incentives (., China).
Clustering standard errors at the country-sector level: The baseline specification clusters standard
errors at the country level to account for potential cross-sectoral spillovers. We test robustness to
a less conservative (more granular) clustering at the country-sector level. As shown in Figure ,
results remain consistent; the first-stage F-statistics improve and overall IV results are unchanged.
All trade policies: Broadening the set of trade policies with which countries can respond to IPs
in other countries and sectors does not change the results. Figure shows that results are
stable, with the magnitude and shape of the effects of protectionist and liberalizing trade policies
on received patent applications largely unchanged. If anything, estimates are smaller—possibly
reflecting the fact that retaliatory IPs drive the main results.
Policy instrument: Figure shows that results for subsidies and export incentives are un-
changed when clustering the standard errors at the country-sector level. However, the first stage
F-statistics are above 10 under this less conservative assumption. In addition, Figure presents
alternative results when using the same policy tool in other sectors and countries as an instrumental
variable while controlling for other types of trade policy tools. The shape of the results is similar,
but the magnitude is slightly smaller and insignificant for export incentives, likely due to the lower
variation in retaliatory export incentives compared to retaliatory IPs.
11Although the weights and underlying variables differ across our instruments, both are constructed from foreign
policy movements, whose correlation could limit the power of the Hansen J-test.
INTERNATIONAL MONETARY FUND 26
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Centrality: Figure 9’s results on the targeting of innovation-central sectors rely on the eigenvector
centrality definition described in Section . Although this is the measure that Liu and Ma find
to be relevant for the socially optimal allocation of innovation resources in their theoretical model,
other measures could better capture innovation spillovers in the network. We test the sensitivity of
results on innovation-central sectors to alternative definitions of network centrality. First, we test
the robustness of the results to varying the threshold. Changing the cutoff from the 80th percentile
to the median (Figure ) yields similar results, although estimates for non-central sectors are
less precise. Second, we replace the eigenvector centrality measure with the PageRank centrality
one (Figure ), accounting for the extent of patent citations including indirect influence through
citation chains12. Results are also unchanged. Together, these findings suggest the main results
are not sensitive to the centrality metric or the threshold used.
OLS: Figure presents the OLS estimation of Equation 1, providing a benchmark for inter-
preting the IV results shown in Figure 3. Indeed, as explained in Section , the OLS results shed
light on the average association between innovation and IPs, no matter targeting, even if estimates
cannot be causally interpreted. The comparison between OLS and IV results supports the above
point on the importance of targeting to unleash domestic innovation. An additional protectionist
IP correlates with a percentage point increase in domestic patent applications after one year.
However, the presence of an upward pre-trend suggests selection bias with industrial policies dis-
proportionately targeting sectors with upward-trending innovation. Therefore, the OLS estimation
suggests that targeting “promising” sectors is important for IPs to be associated with a boost in
domestic innovation, as shown in Section . For foreign patents, OLS and IV estimates align in
magnitude and timing, but the pre-trends are also pronounced in the OLS results, especially for
protectionist IPs–again indicating possible selection effects–. A decomposition into extensive and
intensive margin shows that the pre-trend results from an increasing pre-trend on the extensive
margin (., entry of new patenting sectors) and a decreasing pre-trend on the intensive margin.
These findings reinforce the rationale for the IV method. While OLS captures the average associ-
ation, IV better identifies causal effects by addressing endogenous targeting, as supported by the
absence of pre-trends in the IV specification.
5 Conclusion
Overall, our results suggest that industrial policies can foster patent applications when well de-
signed. Protectionist IPs temporarily increase received foreign patents, with export-oriented poli-
cies outperforming subsidies. This temporary effect may stem from foreign inventors protecting
their invention or securing access to the markets targeted by the industrial policy by patenting
innovations they already have in the pipeline. In contrast, liberalizing IPs, such as lifting trade
barriers, produce more persistent and broader cross-border patenting. However, both protectionist
12This measure was initially used by web searching engines to measure the relative importance of webpages de-
pending on links referring to them.
INTERNATIONAL MONETARY FUND 27
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
and liberalizing IPs have no significant effect on domestic patenting, suggesting that, on average,
new innovation likely takes more than four years to materialize.
Targeting IPs to promising sectors is essential to unleash innovation. For instance, targeting
infant industries, such as LCTs, is effective to boost domestic patent applications in the medium
term. Indeed, targeted support can relax financing constraints and accelerate innovation in early-
stage, high-potential sectors (Aghion et al., 2015; Criscuolo et al., 2019; Mazzucato, 2011). In
addition, IPs focused on innovation-central exhibit larger gains in terms of received foreign patent
applications. This is consistent with broader evidence on other outcomes showing that industrial
policies are more effective when targeting sector with important market failures and distortions
(Baquié et al., 2025; Machado-Parente et al., 2025; IMF, 2025).
The effects of IPs also vary by country income group. Protectionist IPs disproportionately
facilitate the receipt of patent applications in EMDEs, as foreign inventors may try to protect
their invention or secure access to markets impacted by these policies. Hence, further research is
necessary to determine whether patent applications in EMDEs actually translate into increased
productivity. Machado-Parente et al. (2025) suggest that this may not be the case in the short
run, as firms benefit more from IPs in countries with better fundamentals, such as governance or
financial market development. At the same time, liberalizing IPs encourage cross-country patenting
from AEs, potentially as firms access lower-cost or higher-quality imported inputs that facilitate
innovation or as heightened product market competition incentivizes them to engage more actively
in patenting (Goldberg et al., 2009; Amiti and Konings, 2007; Bloom et al., 2011).
Thus, while IPs have the potential to boost innovation, their effectiveness in shaping innova-
tion depends on their targeting and implementation. Supporting infant industries, or prioritizing
measures that facilitate cross-border patenting along the innovation network could yield higher
benefits. Moreover, IPs tend to increase technology diffusion to EMDEs, although more evidence is
needed to assess whether this translates into technological development and productivity increases.
In addition, broader outcomes should be carefully evaluated prior to implementation, including
expected benefits, spillovers, and policy alternatives. Indeed, while our findings provide a positive
medium-term assessment of IPs’ effects on patenting, they do not capture general equilibrium dy-
namics—such as trade adjustments or cross-country spillovers (Hodge et al., 2024; Huang et al.,
2025; Brandão-Marques and Toprak, 2024).
INTERNATIONAL MONETARY FUND 28
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
References
Aghion, Philippe, Antoine Dechezleprêtre, David Hemous, Ralf Martin, and John Van Reenen,
“Carbon Taxes, Path Dependency, and Directed Technical Change: Evidence from the Auto
Industry,” Journal of Political Economy, 2016, 124 (1), 1–51. doi: Publisher:
University of Chicago Press.
, Jing Cai, Mathias Dewatripont, Luosha Du, Ann Harrison, and Patrick Legros, “Industrial
Policy and Competition,” American Economic Journal: Macroeconomics, October 2015, 7 (4),
1–32.
, Lint Barrage, Eric Donald, David Hémous, and Ernest Liu, “Transition to Green Technology
along the Supply Chain,” Working Paper 33934, National Bureau of Economic Research June
2025.
Akcigit, Ufuk and Nathan Goldschlag, “Where Have All the “Creative Talents”Gone? Employment
Dynamics of US Inventors,”Working Paper 31085, National Bureau of Economic Research March
2023.
Alston, Julian M., Philip G. Pardey, Devin Serfas, and Shanchao Wang, “Slow Magic: Agricultural
Versus Industrial R&D Lag Models,” Annual Review of Resource Economics, October 2023, 15
(Volume 15, 2023), 471–493. Publisher: Annual Reviews.
Amiti, Mary and Jozef Konings, “Trade liberalization, intermediate inputs, and productivity: Evi-
dence from indonesia,”American Economic Review, December 2007, 97 (5), 1611–1638.
Bailey, Michael A., Anton Strezhnev, and Erik Voeten, “Estimating dynamic state preferences from
united nations voting data,”The Journal of Conflict Resolution, 2017, 61 (2), 430–456. Publisher:
Sage Publications, Inc.
Baquié, Sandra, Yueling Huang, Florence Jaumotte, Jaden Kim, Rafael Machado Parente, and
Samuel Pienknagura, “Industrial Policies: Handle with care,” IMF Staff Discussion Notes, March
2025, 002, 1. Publisher: International Monetary Fund (IMF).
Bartelme, Dominick G., Arnaud Costinot, Dave Donaldson, and Andrés Rodŕıguez-Clare, “The
Textbook Case for Industrial Policy: Theory Meets Data,” Working Paper 26193, National Bu-
reau of Economic Research August 2019.
Barwick, Panle Jia, Hyuk soo Kwon, Shanjun Li, and Nahim Zahur, “Drive Down the Cost: Learn-
ing by Doing and Government Policies in the Global EV Battery Industry *,” Working Paper
33378, National Bureau of Economic Research March 2025.
, Hyuk-Soo Kwon, Shanjun Li, Yucheng Wang, and Nahim B. Zahur, “Industrial Policies and
Innovation: Evidence from the Global Automobile Industry,” Working Paper 33138, National
Bureau of Economic Research November 2024.
Blind, Knut, Jakob Edler, Rainer Frietsch, and Ulrich Schmoch, “Motives to patent: Empirical
evidence from Germany,”Research Policy, June 2006, 35 (5), 655–672.
Bloom, Nicholas, Mirko Draca, and John Van Reenen, “Trade Induced Technical Change? The Im-
pact of Chinese Imports on Innovation, IT and Productivity,”Technical Report w16717, National
Bureau of Economic Research January 2011.
INTERNATIONAL MONETARY FUND 29
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Brandão-Marques, Luis and Hasan H Toprak, “A Bitter Aftertaste: How State Aid Affects Recipient
Firms and Their Competitors in Europe,” IMF Working Papers, 2024, (250).
Branstetter, Lee, Raymond Fisman, and C. Fritz Foley, “Do Stronger Intellectual Property Rights
Increase International Technology Transfer? Empirical Evidence from . Firm-Level Data,”
August 2005.
Chang, Ha-Joon, “Industrial policy in East Asia: Lessons for Europe,” EIB Papers, 2006, 11 (2),
106–132. Publisher: Luxembourg: European Investment Bank (EIB).
Cherif, Reda and Fuad Hassanov, “The Return of the Policy that Shall Not Be Named: Principles
of Industrial Policy,” IMF Working Papers, 2019, (074).
Choi, Jaedo and Andrei A. Levchenko, “The long-term effects of industrial policy,” Journal of
Monetary Economics, June 2025, 152, 103779.
Criscuolo, Chiara, Ralf Martin, Henry Overman, and John van Reenen, “Some causal effects of an
industrial policy,”The American Economic Review, 2019, 109 (1), 48–85.
Evenett, Simon, Adam Jakubik, Fernando Mart́ın, and Michele Ruta, “The return of industrial
policy in data,” IMF Working Papers, 2024, (001).
Evenett, Simon J. and Johannes Fritz, “The GTA Handbook: Data and methodology used by the
Global Trade Alert initiative,” Technical Report, Global Trade Alert 2020.
Garcia-Macia, Daniel and Alexandre Sollaci, “Industrial Policies for Innovation: A Cost-Benefit
Framework,”Working Papers, IMF August 2024.
Goldberg, Pinelopi, Amit Khandelwal, Nina Pavcnik, and Petia Topalova, “Trade Liberalization
and New Imported Inputs,”American Economic Review, May 2009, 99 (2), 494–500.
Goldberg, Pinelopi K., Réka Juhász, Nathan J. Lane, Giulia Lo Forte, and Jeff Thurk, “Industrial
Policy in the Global Semiconductor Sector,”Working Paper 32651, National Bureau of Economic
Research July 2024.
Goldschlag, Nathan, Travis J. Lybbert, and Nikolas J. Zolas, “Tracking the technological com-
position of industries with algorithmic patent concordance,” Economics of Innovation and New
Technology, 2020, 29 (6), 582–602.
Gopinath, Gita, Şebnem Kalemli-Özcan, Loukas Karabarbounis, and Carolina Villegas-Sanchez,
“Capital allocation and productivity in south europe*,” The Quarterly Journal of Economics,
June 2017, 132 (4), 1915–1967.
GTA, “Global Trade Alert Database,” 2022.
Hall, Bronwyn H., Adam B. Jaffe, and Manuel Trajtenberg, “The NBER Patent Citation Data File:
Lessons, Insights and Methodological Tools,”Working Paper 8498, National Bureau of Economic
Research October 2001.
Harrison, Ann, “What Makes Industrial Policy Work?,”CEPR Discussion Paper, November 2024,
19693.
INTERNATIONAL MONETARY FUND 30
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Hasna, Zeina, Florence Jaumotte, Jaden Kim, Samuel Pienknagura, and Gregor Schwerhoff, “Green
Innovation and Diffusion: Policies to Accelerate Them and Expected Impact on Macroeconomic
and Firm-Level Performance,” IMF Staff Discussion Notes, November 2023. Publisher: Interna-
tional Monetary Fund (IMF).
Hodge, Andrew, Roberto Piazza, Fuad Hasanov, Xun Li, Maryam Vaziri, Atticus Weller, and
Yu Ching Wong, “Industrial policy in europe: a single market perspective,” 2024.
Hsieh, Chang-Tai and Peter J. Klenow, “Misallocation and manufacturing TFP in china and india*,”
The Quarterly Journal of Economics, November 2009, 124 (4), 1403–1448.
Hu, Albert Guangzhou and Gary H. Jefferson, “A great wall of patents: What is behind China’s
recent patent explosion?,” Journal of Development Economics, September 2009, 90 (1), 57–68.
Huang, Yueling, Sandra Baquié, Florence Jaumotte, Jaden Kim, Yucheng Lu, Rafael Machado
Parente, and Samuel Pienknagura, “Do Industrial Policies Increase Trade Competitiveness?,”
IMF Working Papers, 2025, (098).
IMF, “Industrial policy: Managing trade-offs to promote growth and resilience?,”World Economic
Outlook: Chapter 3, October 2025, Washington, DC.
Juhász, Réka and Nathan Lane, “The Political Economy of Industrial Policy,” Journal of Economic
Perspectives, November 2024, 38 (4), 27–54.
and , “A Short Guide to Thinking About Industrial Policy: Takeaways from the New Eco-
nomics of Industrial Policy,” SocArXiv, August 2024. Number: 4sra7 v1 Publisher: Center for
Open Science.
, , and Dani Rodrik, “The New Economics of Industrial Policy,”Annual Review of Economics,
2024, 16, 213–242.
, , Emily Oehlsen, and Verónica C. Pérez, “Measuring Industrial Policy: A Text-Based Ap-
proach,”Working Paper, National Bureau of Economic Research June 2025.
Kowalski, Przemyslaw and Clarisse Legendre, “Raw materials critical for the green transition:
Production, international trade and export restriction,” Technical Report 269, OECD, Paris
2023.
Krueger, Anne O., “Government Failures in Development,” Journal of Economic Perspectives,
September 1990, 4 (3), 9–23.
, “Trade Policy and Economic Development: How We Learn,”American Economic Review, 1997,
87 (1), 1–22. Publisher: American Economic Association.
LaBelle, Jesse, Immaculada Martinez-Zarzoso, Ana Maria Santacreu, and Yoto Yotov, “Cross-
border Patenting, Globalization, and Development,” Working Papers 202507, Center for Global
Policy Analysis, LeBow College of Business, Drexel University. June 2024.
Lane, Nathan, “Manufacturing Revolutions: Industrial Policy and Industrialization in South Ko-
rea,”Working Papers 2021-10, Monash University, SoDa Laboratories.
Liu, Ernest and Song Ma, “Innovation Networks and R&D Allocation,” Working Paper 29607,
National Bureau of Economic Research December 2021.
INTERNATIONAL MONETARY FUND 31
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Machado-Parente, Rafael, Samuel Pienknagura, Sandra Baquié, Yueling Huang, Florence Jau-
motte, and Jaden Kim, “Industrial Policies and Firm Performance: A Nuanced Relationship,”
Working Papers, IMF 2025.
Mazzucato, Mariana, The Entrepreneurial State 2011.
Mealy, Penny and Alexander Teytelboym, “Economic complexity and the green economy,”Research
Policy, October 2022, 51 (8), 103948.
Melitz, Marc J., “When and how should infant industries be protected?,” Journal of International
Economics, May 2005, 66 (1), 177–196.
OECD, “An institutional framework for industrial policies,” Policy Papers 180, OECD, Paris June
2025. Edition: 180 Series: OECD Science, Technology and Industry Policy Papers.
OECD/Eurostat, “The Environmental Goods and Services Industry: Manual for Data Collection
and Analysis,” Technical Report, OECD, Paris 1999.
Pack, Howard and Kamal Saggi, “The Case for Industrial Policy: A Critical Survey,” Working
Papers 3839, The World Bank February 2006.
Pigato, Miria A., Simon J. Black, Damien Dussaux, Zhimin Mao, Miles McKenna, Ryan Rafaty, and
Simon Touboul, “Technology Transfer and Innovation for Low-Carbon Development.,” Technical
Report, The World Bank, Washington . 2020.
Reed, Tristan, “Export-Led Industrial Policy for Developing Countries: Is There a Way to Pick
Winners?,” Journal of Economic Perspectives, November 2024, 38 (4), 3–26.
Rodrigues, Mauro, “Import substitution and economic growth,” Journal of Monetary Economics,
March 2010, 57 (2), 175–188.
Rodrik, Dani, “Industrial Policy: Don’t Ask Why, Ask How,” Middle East Development Journal,
January 2009, 1 (1), 1–29.
Rosenow, Samuel Kaspar and Penelope Ann Mealy, “Turning risks into rewards: Diversifying the
global value chains of decarbonization technologies,”Policy Papers 10696, The World Bank 2024.
Wang, Shanchao, Alston Julian M., , and Philip G. Pardey, “R&D lags in economic models,”
Economics of Innovation and New Technology, 2024, pp. 1–21. Publisher: Routledge eprint:
Weiss, John, “Export growth and industrial policy: Lessons from the East Asian miracle experi-
ence,”ADBI Discussion Paper, 2005, 26.
WITS, “Product Concordances,” 2025.
World Bank, ed., The East Asian miracle: economic growth and public policy A World Bank policy
research report, New York, : Oxford University Press, 1993.
Òscar Jordà, “Estimation and inference of impulse responses by local projections,”American Eco-
nomic Review, March 2005, 95 (1), 161–182.
and Alan M. Taylor, “The time for austerity: Estimating the average treatment effect of fiscal
policy,”The Economic Journal, February 2016, 126 (590), 219–255.
INTERNATIONAL MONETARY FUND 32
IMF WORKING PAPERS Shaping Innovation: Can Industrial Policies Boost Patent Applications?
Supporting material
A Additional Data Description
Figure : Cross-sectoral distribution of the average number of patent applications received or
submitted
0
500
1000
1500
2000
A
v
e
ra
g
e
n
u
m
b
e
r
o
f
p
a
te
n
t
a
p
p
l.
(
2
0
0
9
−
1
9
)
C
o
ll
e
c
ti
o
n
,
p
u
ri
fi
c
a
ti
o
n
a
n
d
d
is
tr
ib
u
ti
o
n
o
f
w
a
te
r
C
o
n
s
tr
u
c
ti
o
n
E
le
c
tr
ic
it
y
,
g
a
s
,
s
te
a
m
a
n
d
h
o
t
w
a
te
r
s
u
p
p
ly
E
x
tr
a
c
ti
o
n
o
f
c
ru
d
e
p
e
tr
o
le
u
m
a
n
d
n
a
tu
ra
l
g
a
s
[
..
.]
M
a
n
u
fa
c
tu
re
o
f
b
a
s
ic
m
e
ta
ls
M
a
n
u
fa
c
tu
re
o
f
c
h
e
m
ic
a
ls
a
n
d
c
h
e
m
ic
a
l
p
ro
d
u
c
ts
M
a
n
u
fa
c
tu
re
o
f
c
o
k
e
,
re
fi
n
e
d
p
e
tr
o
le
u
m
p
ro
d
u
c
ts
a
n
d
n
u
c
le
a
r
fu
e
l
M
a
n
u
fa
c
tu
re
o
f
e
le
c
tr
ic
a
l
m
a
c
h
in
e
ry
a
n
d
a
p
p
a
ra
tu
s
n
.e
.c
.
M
a
n
u
fa
c
tu
re
o
f
fa
b
ri
c
a
te
d
m
e
ta
l
p
ro
d
u
c
ts
,
e
x
c
e
p
t
m
a
c
h
in
e
ry
a
n
d
e
q
u
ip
m
e
n
t
M
a
n
u
fa
c
tu
re
o
f
fo
o
d
p
ro
d
u
c
ts
a
n
d
b
e
v
e
ra
g
e
s
M
a
n
u
fa
c
tu
re
o
f
fu
rn
it
u
re
;
m
a
n
u
fa
c
tu
ri
n
g
n
.e
.c
.
M
a
n
u
fa
c
tu
re
o
f
m
a
c
h
in
e
ry
a
n
d
e
q
u
ip
m
e
n
t
n
.e
.c
.
M
a
n
u
fa
c
tu
re
o
f
m
e
d
ic
a
l,
p
re
c
is
io
n
a
n
d
o
p
ti
c
a
l
in
s
tr
u
m
e
n
ts
,
w
a
tc
h
e
s
a
n
d
c
lo
c
k
s
M
a
n
u
fa
c
tu
re
o
f
m
o
to
r
v
e
h
ic
le
s
,
tr
a
il
e
rs
a
n
d
s
e
m
i−
tr
a
il
e
rs
M
a
n
u
fa
c
tu
re
o
f
o
ff
ic
e
,
a
c
c
o
u
n
ti
n
g
a
n
d
c
o
m
p
u
ti
n
g
m
a
c
h
in
e
ry
M
a
n
u
fa
c
tu
re
o
f
o
th
e
r
n
o
n
−
m
e
ta
ll
ic
m
in
e
ra
l
p
ro
d
u
c
ts
M
a
n
u
fa
c
tu
re
o
f
o
th
e
r
tr
a
n
s
p
o
rt
e
q
u
ip
m
e
n
t
M
a
n
u
fa
c
tu
re
o
f
p
a
p
e
r
a
n
d
p
a
p
e
r
p
ro
d
u
c
ts
M
a
n
u
fa
c
tu
re
o
f
ra
d
io
,
te
le
v
is
io
n
a
n
d
c
o
m
m
u
n
ic
a
ti
o
n
e
q
u
ip
m
e
n
t
a
n
d
a
p
p
a
ra
tu
s
M
a
n
u
fa
c
tu
re
o
f
ru
b
b
e
r
a
n
d
p
la
s
ti
c
s
p
ro
d
u
c
ts
M
a
n
u
fa
c
tu
re
o
f
te
x
ti
le
s
M
a
n
u
fa
c
tu
re
o
f
to
b
a
c
c
o
p
ro
d
u
c
ts
M
a
n
u
fa
c
tu
re
o
f
w
e
a
ri
n
g
a
p
p
a
re
l;
d
re
s
s
in
g
a
n
d
d
y
e
in
g
o
f
fu
r
M
a
n
u
fa
c
tu
re
o
f
w
o
o
d
[
..
.]
M
in
in
g
o
f
c
o
a
l
a
n
d
l
ig
n
it
e
;
e
x
tr
a
c
ti
o
n
o
f
p
e
a
t
M
in
in
g
o
f
m
e
ta
l
o
re
s
M
in
in
g
o
f
u
ra
n
iu
m
a
n
d
t
h
o
ri
u
m
o
re
s
O
th
e
r
m
in
in
