Innovate to Resist: Are Innovators Shielded from External Shocks?

Economic shocks are often difficult to predict, as they are independent of the will of the economic agents affected and, therefore, exogenous to their choices. However, the effects of such shocks when they occur impact on national economies, business performance, and employment of individuals, thus shaking countries’ economies, and often with long-lasting effects. Although it is difficult to predict their occurrence, it is possible to withstand the consequences of exogenous shocks and be more resilient should they occur. Investing in order to be at the frontier of innovation is one of the tools that economies can adopt to protect their agents, thus defending them from the downturns that can result from unexpected economic events. In this paper, we study whether being more innovative does protect Italian provinces from the negative effects of exogenous shocks, with a focus on employment levels. The object of the analysis is, more specifically, the role of the innovation stock in protecting provinces’ employment levels from trade and economic shocks. The analysis, conducted in the period 2000–2018, examines the effect of trade shock on Italian provinces’ employment levels, and the role of the innovation stock in preserving those levels. Our results confirm that innovative provinces are more protected from external trade and economic shocks, with this effect being evident in both Southern and Northern regions, and despite the level of internationalization of provinces.


Introduction
The last two decades have witnessed many structural changes in the globally increasing interest in the economic literature, which has extensively investigated the effects of exogenous shocks, finding the effects to be substantially different across industries, markets, and firms.One the most relevant shocks in the last decades can be traced back to China entering the WTO in 2001, and to the subsequent increase in international competition (e.g. increase in imports from China).
Following the seminal work of Autor et al. (2013), a vast body of literature does indeed analyze the effects of the increase in imports from China (the so-called "China Shock") on employment, wages, and innovation outcomes in the United States (US), almost exclusively focusing on industry variation (Autor et al. 2016(Autor et al. , 2021;;Acemoglu et al. 2016;Hombert and Matray 2018).A large body of literature is also covering EU countries: Donoso et al. (2015) for Spain, Balsvik et al. (2015) for Norway, Malgouyres (2017) for France, Citino and Linarello (2022) for Italy, Cabral et al. (2018) and Branstetter et al. (2019) for Portugal, and Dauth et al. (2014) for Germany.Meanwhile, on the effects of the Great recession, several studies have exploited how the sector and firm heterogeneity induces different responses to external demand shocks in the US (Charles et al. 2016;Yagan 2019) and in other countries such as Italy (Brancati et al. 2022), France (Domini and Moschella 2018), Portugal (Carreira and Teixera 2016), and a set of OECD countries (Criscuolo et al. 2014).In this paper, we add to this literature by adopting a different point of view, asking whether innovation could be an effective shield against external shocks, protecting more innovative economic agents from the potential negative downsides that they could experience from shocks that are exogenous and out of their control (Hombert and Matray 2018).
Focusing on the trade shock, we link this problem to the recent rise of China, which is facilitated by its transition to a market-oriented economy and its ever-increasing integration into global trade.This rise has been identified in the literature as one of the main causes of destabilisation of the economies of high-income countries.The increasing integration of China into the global economy has triggered a lively debate on what the effects of trade with low-wage countries have on firms and workers in high-income countries, and what conditions are in place for firms and workers to withstand and absorb such shocks.
In this changing context, innovation is often seen as a useful tool for firms to differentiate their products from imports from low-wage countries by serving as an effective shield against foreign competition.Competition based solely on mere cost, in fact, given such profound wage differences between countries, is doomed to failure, and only investment in innovation, and subsequent product improvement, enables economic agents-particularly firms-to compete successfully against low-cost imports.Moreover, being innovative and standing on the frontier of innovation, could also help to prevent being overwhelmed by the effects of economic crises, as firms are more able to differentiate their products.
Traditionally, the innovation literature has analysed related but fundamentally different questions about how economic agents endogenously adjust their R&D investments after such shocks have occurred (Autor et al. 2016;Bloom et al. 2016).In the present paper, instead, we estimate how specific labour outcomes are affected by (exogenous) shocks, conditional on their innovation stock before the shocks (Hombert and Matray 2018).We test our hypotheses on the context of Italian provinces (NUTS 3) on the period from 2000 to 2018, investigating whether the provinces with a greater stock of innovation when the shock occurs are more able to resist against the downturns of the economic cycle.In particular, we explore with province-level data on employment for Italy if local units are less adversely affected by an increase in import competition when ex-ante they invested more in successful innovation (i.e. higher stock of innovation).Then, we take a direct approach to estimating the effect of innovation on the ability of Italian provinces to resist competition arising from imports from low-income countries.In fact, we test whether provinces' performance, measured by changes in employment, is less affected by increased import competition when the province has invested more in innovation, as measured by patent output, prior to increased import penetration, in order to measure the role of innovation as an ex-ante moderator of the effects of imports from China.
In this context, studying the effect of unexpected shocks on Italy and the role of innovation are of particular interest; in that, the Italian context is traditionally characterised by disparities between regions, provinces, and industrial district economies, that during periods of crisis can show different dynamics, and also in terms of industrial demography.In the aftermath of another unexpected crisis, i.e. the Great Recession of the last decade, some regions faced prolonged decline characterised by widening productivity gaps, escalating firm closures, mounting unemployment, and dwindling competitiveness on the international stage (Di Giacinto et al. 2012).These trends exerted a pronounced impact on industrial districts (IDs), leading to a partial deindustrialisation, limited tertiary sector growth, sluggish productivity, and a growing divergence between Northern and Southern regions.
Recent empirical studies have elucidated that the once-prevailing competitive advantages rooted in informal socio-economic relations within IDs, and the benefits of geographic proximity, have waned over the past decade, due to a confluence of transformative changes.The Italian economy has undergone a progressive shift towards the tertiary sector, driven by structural changes.Additionally, the specialisation of most Italian industrial districts predominantly pertains to low, medium-low, and medium-high tech sectors.Italy's underwhelming specialisation in high-tech sectors, given the country's proximity to the technological frontier, has significantly contributed to its eroding competitiveness, especially in light of technological catch-up by emerging economies that share similar specialisations (notably China, Eastern European countries, and more recently, competitive Asian counterparts like Thailand and Vietnam).
Furthermore, the evolving international competitive landscape, coupled with the surge in delocalisation and the protracted economic crisis, has eroded the resilience of IDs and local economies.Indeed, district firms are exhibiting diminished resilience compared to their historical performance, further exacerbating the lag in Southern regions and widening the disparity vis-à-vis Northern Italy and more advanced regions within the European Union (e.g.Iuzzolino and Micucci 2011;Alampi et al. 2012).Therefore, the primary contribution of this paper to the existing literature lies in two main aspects.Firstly, it offers empirical evidence concerning the impact of trade competition, as well as the significant role played by innovation in safeguarding employment at the province level.Secondly, it contributes to the body of knowledge addressing geographical disparities within Italy, shedding light on this pertinent issue.
The paper is organised as follows.Section 2 reviews the relevant literature.Section 3 presents a set of stylised facts.Then, Sect. 4 describes our empirical approach and the data used.Section 5 explains the methodology used to estimate the impact of the shocks, and the results are presented in Sect.7 which provides a discussion of these results, while Sect.8 concludes the paper.An Appendix complements the main body of the paper.

Related Literature
There is evidence of an unemployment increase/wage decrease trend in more advanced economies starting in the 1980s, while trade liberalisation triggered and China was starting its fast transition to a market-oriented economy, which gave rise to a longstanding debate about the impact of trade with low-wage countries on the labour market in the US and Europe.
This paper draws upon this literature [surveyed by Bernard et al. (2012); and by Autor et al. (2016)] on the impact of import competition on high-income economies.Several seminal papers analysed this issue at firm level for the US, demonstrating the unconditional effect of trade shocks on various dimensions of firm performance such as output and survival (Bernard et al. 2006), cost of debt and leverage (Valta 2012), capital expenditure (Frèsard and Valta 2016), and employment and outsourcing (Pierce and Schott 2016).As far as studies for employment effects in various countries are concerned, all existing studies document negative employment effects.
However, some important differences emerge at the level of individual countries compared to those found in the US by Autor et al. (2013) with employment effects greater in Spain (Donoso et al. 2015), or lower in countries such as Norway (Balsvik et al. 2015), France (Malgouyres 2017), Portugal (Cabral et al. 2018and Branstetter et al. 2019), or Germany.In the latter case, Dauth et al. (2014) find that while areas specialised in import-competing industries lost employment, this was more than compensated by gains in areas specialised in export-oriented industries.
Regarding Italy, the extent and significance of the impact of Chinese import competition on employment in the manufacturing sector has been studied in previous research.Most of these studies underline a significant impact compared to other countries, as well as compared to overall employment trends.In the study by Citino and Linarello (2022), a back-of-the-envelope calculation shows that the "China shock" led to the displacement of about 24,000 jobs in the period 1991-2001, and a more significant displacement of 119,000 jobs in the period 2001-2007.This decline, which represents around 12% of total manufacturing employment, significantly exceeds the job losses in other European countries such as France, Germany, Portugal, and Norway.In these countries, the number of manufacturing jobs lost to Chinese import competition ranged from 1 to 4% of manufacturing employment in 1995.
Moreover, these figures also indicate that Italy has experienced an asymmetric shock: the impact has been concentrated in certain sectors and has shown a pronounced asymmetry in its effects on firms, sectors, and geographical regions.
In this context, innovation is often viewed as an effective shield against low-cost foreign competition.Only firms that have invested in R&D, upgraded product quality, and are likely to protect their innovation with patents, are able to compete successfully against low-cost imports by climbing the quality ladder and differentiating their products from the exports of low-wage countries (Leamer 2007).Bernard et al. (2006) showed that capital-intensive plants are more likely to survive and grow in the wake of import competition.However, little empirical evidence exists on whether firms that have invested in R&D are shielded from trade shocks.An exception in point is the study of Hombert and Matray (2018), which questions whether R&D-intensive firms are more resilient to trade shocks.Those authors find that rising imports from China lead to slower sales growth and lower profitability, but these effects are significantly smaller for firms with a larger stock of R&D (about half when moving from the bottom quartile to the top quartile of R&D).Hence, firms in import-competing industries could tend to cut expenditures and employment, while R&D-intensive firms do this considerably less.
There is a growing body of literature on the employment effects of innovation activities, which can shed light on the mechanisms which may protect from trade shock, considering how different types of innovation can have opposite effects on employment.Indeed, product innovations, (i.e. the introduction of new goods or services) can lead to the emergence of new markets and, thus, can induce positive employment effects, while process innovations (i.e. the implementation of a new and significantly improved production method) can lead to technological unemployment due to increased labour productivity (Harrison et al. 2014;Piva and Vivarelli 2018).
Furthermore, innovation also produces a skill-biased technological change effect on employment (Berman et al. 1994;Amoroso and Castello 2018;Autor et al. 2003), reducing the need for low-skilled employees while increasing the demand for highskilled employees.At the firm level, R&D raises the demand for employment in occupations requiring abstract skills (e.g.engineers).At the same time, it may lead to improved productivity and new products, which may result in greater demand, and employment, in all occupations (Moretti 2010;Piva and Vivarelli 2018).An important issue to consider is that if R&D growth occurs in areas with highly routinised labour, and this may have a negative effect on the demand for local employment, which does not have the skills to work in the new jobs that are required by the R&D investments.Hence, Bogliacino et al. (2012) and Barbieri et al. (2019) have found a positive employment effect of R&D in high-tech industries, while reporting non-significant findings for low-tech sectors.
Despite the abundant literature on the topics described above, there remains a gap of knowledge about whether firms that have invested in R&D are shielded from trade shocks, and on how the firms' innovation activities (measured by R&D) undertaken before a trade shock do impact on the labour market.The empirical literature on these two topics is mainly based on microeconomic evidence that uses firm-and sector-level data.The previous literature on the Italian case has pointed out that industries which were hit by import competition from low-wage countries lost employment, compared to other manufacturing industries, and that this is especially true in low-skill and labour-intensive industries (Federico 2014).At the regional level, the links between employment growth, import competition, and innovative capacity have largely been ignored.Thus, very little is known about how the effects observed at the firm or sector level relate to the regional level.For Italy, the recent study by Citino and Linarello (2022) describes the effects of increased import competition from China on the Italian local labour markets (LLM).The authors' idea was to check both the effects on the LLM and on manufacturing workers.The study describes how the Italian local labour markets, which were more exposed to Chinese trade by means of their industry composition, ended up suffering larger manufacturing and overall employment losses.However, workers initially employed in more exposed manufacturing industries did not suffer long-term losses.While they were indeed less likely than other similar workers to continue working in manufacturing, they were also able to carry out successful transitions toward the non-tradable sector, in other areas with better job opportunities.
As for the evidence on the unemployment increase/wage decrease trend during another unexpected shock, such as the global financial crisis (GFC), the literature is mainly focused on analyses at the firm or industry level.Firm heterogeneity induces different responses to external demand shocks in the US (Charles et al. 2016;Yagan 2019) and in other countries (Brancati et al. 2022;Carreira and Teixera 2016;Criscuolo et al. 2014;Domini and Moschella 2018).Criscuolo et al. (2014) focus on young firms which are founded by net job creators throughout the business cycle, even during the financial crisis.Domini and Moschella (2018), for French manufacturing firms, show that during the GFC more productive firms decreased their advantage with respect to less productive firms, in terms of both employment growth and probability to survive, in disagreement with the "cleansing hypothesis" (i.e. the relocation of resources due to a recession).Carreira and Teixera (2016) also show that in the extreme scenario of deep recession, efficiency in resource reallocation can be reduced.The explanation is that credit market stringency in conjunction with an unfavourable economic cycle is likely to generate a long-lasting destructive process.However, the firm-level analysis of Italy underlines how critical aspects such as technological capabilities, innovativeness, and R&D investment strategies triggered superior performances during the global recession (Brancati et al. 2022).Except for the paper by Yagan (2019) on US local areas, the links between employment growth, global shock, innovative capacity, and the long-term impacts of the Great Recession have been largely ignored at the regional level.

Stylised Facts
As illustrated above, the objective of this paper is to consider how the innovative performance of Italian provinces can protect them or help them react to exogenous shocks by protecting their employment levels.The Italian case is of particular interest due to the traditional disparities that mark the country being characterized by huge differences in the quality of institutions at the subnational level, particularly among regions of Southern Italy, commonly referred to as the Mezzogiorno (comprising Abruzzo, Puglia, Basilicata, Campania, Calabria, Molise, Sicily and Sardinia), and Northern Italy.As highlighted by previous studies, these features can be explained by historical factors, such as the consolidation of clientelistic networks or social trust and economic inequalities, making the differences between the Italian regions of Bolzano and Calabria larger than the differences in the national averages between Sweden and Spain, or even between Germany and Slovakia (Charron and Lapuente 2013).
It is noteworthy that, traditionally, studies exploring the effects of exogenous shocks on labour market outcomes in Italy have commonly utilized local labour systems, also known as "Sistemi Locali del Lavoro", as the primary unit of analysis (e.g.Citino and Linarello 2022).These non-administrative units provide a lens through which to examine the intricacies of labour market dynamics.In our study, however, we shift our focus to the Italian provinces as our analytical framework.This shift brings to light the role of innovation in coping with the impact of shocks, potentially yielding a variety of valuable insights.First and foremost, the traditional Italian economic landscape is characterised by structural territorial disparities in terms of employment, innovation investment, and firm performance.Thus, the initial rationale for focusing on the provinces is rooted in the role of innovation as a key driver of economic growth.However, among the main countries of the European Union, Italy has the most pronounced geographical economic disparities.Consequently, it is essential to examine whether innovation matters in the context of trade shocks, in order to understand regional disparities and their potential policy implications.Therefore, using Italian provinces as the unit of analysis, and considering both observable and unobservable variables, allows for a more nuanced analysis of how exogenous shocks affect employment levels.
A second reason for adopting this territorial perspective is the use of a narrowly defined geographical unit, such as provinces, which corresponds to the more detailed Eurostat classification (NUTS-3).This approach allows to also consider the key role of industrial agglomeration in the Italian local economy-the concentration of specific industries in certain provinces.Previous studies have examined the impact of agglomeration economies in Italy (Iuzzolino and Micucci 2011;Bugamelli et al. 2012;Alampi et al. 2012;Di Giacinto et al. 2012) as an important determinant of firm growth, a conduit for technology acquisition, and a productivity enhancer.There is a remarkable degree of spatial concentration of industrial activity in Italy, which hosts 101 of the 141 specialised clusters identified in Germany, France and Italy by Alampi et al. (2012).
In addition, a third and equally important motivation for an analysis at the provincial level stems from the fact that this level of governance allows for a more comprehensive understanding of the central role of provinces after decentralisation in innovation, resource allocation, and employment policies.Following the constitutional reform of 2001, Italy operates within a quasi-federal system in which competences, including innovation, employment, and enterprise policies, are shared between the regions and the state on the basis of the principle of vertical subsidiarity.As a result, regional initiatives coexist with national programmes managed by the Italian government.The regions play an important role because their electoral cycles are, on average, longer than those of the national government.As a result, regional governments have the temporal capacity to implement specific strategies (Caloffi and Bellandi 2017).In some cases, due to limited financial autonomy and decreasing state transfers, their policies have been significantly driven and supported by EU structural funds, which have promoted advances in innovation and investments in cutting-edge high-tech sectors or Key Enabling Technologies (KETs), as part of Industry 4.0 strategies.However, historical 123 and institutional differences still persist and they have, over the years, also involved aspects concerning economic policies and, consequently, innovation policies, whose competence, following the constitutional reform of 2001, has become increasingly decentralised in favour of the regions in matters of enterprise and innovation policy, based on the idea that peripheral governments are better able to respond to local needs than central governments.However, innovation policies have not always helped to achieve positive effects on the innovation performance of regions (and provinces) that are lagging in this regard (Peiró-Palomino and Perugini 2022).
In fact, regions in Southern Italy, despite having a greater need to spend on innovation, were less able to absorb and spend public funds allocated to support innovation than the more innovative regions in the Centre-North.Furthermore, the ability to allocate resources, even within the same region, was sub-optimal for the most innovative regions.It has also been shown that the amount of EU structural funds allocated to R&D is geographically concentrated within the same region (D'Adda et al. 2022), and that in some regions previous innovation policy experiences or the quality of accumulated know-how in administrative structures influenced the way funds were allocated (Bellini et al. 2021), thus also leading to a high intra-regional variation in terms of innovation performance.
Our data (Fig. 1) suggest a similar pattern to the one we outlined so far, with the provinces located in Northern Italy being those with the best innovation performance Source: Authors' own elaboration using OECD data before China entered the WTO, measured by the average number of patents in the period from 1993 (the first year in which we observe patent data for Italian provinces) to 2015.In particular, Fig. 1 shows a concentration of innovation within bigger cities-a trend aligned with the established literature (Paunov et al. 2019).Northern provinces are more innovative, showing the larger number of patents on average in the period considered (Milano 557.6,Torino 233.3,Bologna 209.3,Varese 110.5,Bergamo 92.9,Vicenza 90.6,Treviso 87.6), with Roma (142.5)being the only city that is not from the North among the top inventors.On the contrary, provinces that show less than one patent per year are mostly located in the Centre or in the South of Italy as in period 2000-2015 (e.g.Caltanissetta 0.9, Isernia 0.9, Reggio Calabria 0.8, Campobasso 0.8, Nuoro 0.7).
If we look, moreover, at the change in employment in the manufacturing sector in the Italian provinces between 2000 and 2018, Fig. 2 seems to confirm the existence of a trend in which the southern and central regions are the regions in which the employment in the manufacturing sector has declined more.While, overall, the employment in manufacturing has declined on average by 16.20% in Italy, there are some regions (negatively) driving the results, such as Basilicata experiencing a decrease in employment in manufacturing of more than 38% in 2018 compared to 2000, followed by Lazio (− 26.52%), Molise (− 26.49%), Campania (− 26.28%), and Sardegna (− 26%).Northern regions, on the contrary, apart from Valle D'Aosta (− 27.27%), although registering negative variations, experienced a smaller decrease, as in the case of Veneto Source: Authors' own elaboration using Eurostat data 123 (− 5.35%) and Emilia-Romagna (− 0.83%) or a positive variation, as in Trentino-Alto Adige (9.49%).Of course, there is also heterogeneity across provinces but, overall, the most affected are in central and southern Italy (Matera.− 46.46%; Caltanissetta − 45.88%, Rieti − 37.09%; Cosenza, − 39.22%), with the northern provinces, on the other hand-and in particular in the North-East of the country-preserving the employment levels of the beginning of the century (like Mantova and Piacenza) or increasing their employment rates in manufacturing (Parma + 14.47%; Belluno + 14.22%, Bolzano + 13.62%, Trieste + 6.06%).Clearly, it must be said that the reasons for this loss of importance in the manufacturing sector can be linked to many factors that go beyond critical events, however important and with structural effects on the Italian economic system, such as the movement of workers from the south to the north of the country, or the progressive change in the national production structure, with an increasing importance of the services sectors at the expense of the manufacturing sector.But there is evidence suggesting that the impact of exogenous shocks on processes, that are already underway, plays an important role.
One of these can be traced back to the entering of China in WTO in 2001, which has exposed countries, and the workers and firms, to external competition with a subsequent loss, in terms of jobs, in those sectors whose outputs compete with imports from China (Autor et al. 2013(Autor et al. , 2021;;Acemoglu et al. 2016;Hombert and Matray 2018;Citino and Linarello 2022).Between 1990 and 2011, the share of world manufacturing exports originating in China increased from 2 to 16 percent and, as shown in Fig. 3, also Italy, like most other countries, has seen its imports from China grow significantly.Since 2001, imports in manufacturing sectors increased from 5.67 mil USD to 35.45 mil USD, with exports in the same sectors that did not equally benefit from this integration process going from 3.17 mil USD to just 14.54 mil USD.Equally, the If we consider both the import penetration from China and the average level of employment in manufacturing in Italian provinces, as shown in Fig. 4, after China entering WTO, there is an increase in import penetration from China at the beginning of the 2000s and an almost contemporaneous decline in the employment in manufacturing in Italian provinces.It is noteworthy that a growing body of literature is exploring the pro-competitive consequences of international trade (Arkolakis et al. 2019;Edmond et al. 2015).Based on the cited literature, we expect that firms in more direct competition with Chinese imports may experience a decline in their mark-ups, potentially leading to the displacement of certain local industries.Several works have developed models that incorporate endogenously variable markups and firm heterogeneity and have predicted lower markups due to increased competition within product markets (Atkeson and Burstein 2008;Mayer et al. 2014;Melitz and Ottaviano 2008).
To explain the theoretical mechanism linking import penetration and employment, price competition could affect local industries by reducing markups, thereby potentially displacing certain local industries.At the same time, it is imperative to incorporate the notion that these displaced industries could import low-cost intermediate inputs, thereby increasing their competitiveness.The result is a dichotomous outcome in which some firms and workers lose out while others gain.In particular, the benefits of trade may manifest themselves in the form of new export prospects, the availability of more affordable and diverse imports, and the potential increase in innovation and quality standards within firms. 1uite understandably, even when considering the extent to which Italian provinces have been exposed to imports from China, there is some heterogeneity at the territorial level if we consider, in fact, the exposure of Italian provinces to Chinese imports.Since Autor et al. (2021) consider how the impact of exposure to the "China Shock" is mainly to be found in the years from 2001 to 2012, given that, after this period, the effect Source: Authors' own elaboration using COEWEB data of this trade shock seems to fade, in Fig. 5 we consider how the level of individual provinces' imports from China has changed in the period that is one year prior to China's WTO entry and ending in 2018.
In this case, the emerging pattern is more heterogeneous.There are provinces in which exposure to imports from China grew very significantly over the 2000-2018 period.In particular, the province of Lodi (Lombardia) saw Chinese imports grow 295 per cent over the years considered to account for 40.4% of Lodi's imports in 2022 (ESTER 2022), as did Oristano (Sardegna), which recorded + 185%.Other than these particularly exposed provinces, the other provinces most exposed to this type of import and, therefore, competition, are Matera + 85.64%, and Campobasso + 61.44%.

Estimation Strategy
In this overall picture, therefore, it is relevant to understand what tools policy makers can use to ensure that, in the face of exogenous and unexpected shocks, they can preserve the economic system they administer and, consequently, the employment levels of the citizens who live in it.The first aim of this work is, as noted, to look at if (and how) showing better performances in innovation could help Italian provinces to better resist against negative external shocks.The idea, indeed, is that being more (less) innovative could protect more (less) against shocks that could undermine the productivity of firms and thus making provinces more (less) resilient against external shocks.
Our estimation model is the following: (1) where p and t are indexes for provinces and time, respectively.In our baseline model, we consider as E M P pτ the change in employment in manufacturing over the total working age population (15-64), since the start of the shock and considering 2000 as the base year.Then, as controls X p,T we include GDP pT , that is the base year (T 2000) GDP per capita and endowment of human capital of province p at the beginning of the period (T 2000) proxied by the enrolment in secondary school (HC pT ).All our regressions are weighted for the working age population (15-64).Finally, η r are the regions' FE to capture any potential unobserved heterogeneity at regional level, and θ t are the decadal dummies to control for decadal trends.2In Eq. ( 1), our variable of main interest is the interaction term between ( S p,t × INN p,t ) where S p,t captures the difference in the share of total imports in manufacturing from China (Borusyak et al. 2022), and INN p,t is the "stock of innovation" (proxied by the output measure of patents).The variable measured in stocks allows to take into account not only the net amount of patents generated, but also the cumulated knowledge generated by past patenting activity.In this paper, we measure the stock of patents of province p, following the perpetual inventory method (Peri 2005; Quatraro and Scandura 2019).We initiate the stock in year 1993, which is the first year when we observe Italian patents in the OECD database, and use the recursive formula INN p,t (1 − δ)INN p,t−1 + nINN p,t where nINN p,t is the net flow of province level patent applications (between t and t − 1) and δ is the obsolescence rate applied to depreciate the stock of past patent applications INN p,t−1 .The value chosen for δ is 15%, which is commonly used in the literature (Keller 2002;Quatraro and Scandura 2019).As mentioned, there are multiple reasons for using stock variables in our case.Firstly, because patent applications fluctuate widely over time, it is possible to get a more complete picture by considering the amount of cumulative knowledge, taking into account both the stock component and the net amount of patents generated in each year.
In addition, this approach also allows us to consider the time between patent applications and patent grants (on average 18-24 months for EPO patents), ensuring that the annual values do not misleadingly consider the stock of knowledge generated through patents at any given time (Quatraro and Scandura 2019).Instead, the stock variable capitalises past and current generated knowledge, thus providing an unbiased measure of the amount of technical knowledge.Moreover, compared to patent flows, patent stocks allow for the fact that patent benefits are likely to persist in future years (Quatraro and Scandura 2019;Bloom and Van Reenen 2002).Therefore, it is advisable to include past patent applications, appropriately discounted, to account for their potential future value.

Data
To estimate the effect of shocks on provinces' performance, we rely on different data sources.Firstly, for bilateral trade, we rely on data from the BACI Database managed by the Centre for Prospective Studies and International Information (CEPII).The data are classified at 6 digits according to the Harmonised System (HS) code 92.Then, to estimate the effect on employment and industries, we use data from the Italian National Institute of Statistics (Istat), which provides historical data on employment at the sectoral level.The data provide information on 14 sectors according to the Nace rev.2 classification.To make trade data comparable with employment data, we use correspondence tables, first from HS92 to SITC Revision 3; then we match the SITC Revision 3 with Nace rev.1.1 data, for which correspondence tables exist.Finally, we match Nace rev.1.1 classification with Nace rev. 2.Moreover, to proxy for the innovation (INN p,t0 in our Eq.1), we use data on patents from OECD at NUTS3 level.Furthermore, for data on controls at province level in Eq. ( 1), we rely on the data made available by the Territorial Statistical Atlas of Infrastructures (ASTI) from Istat.Finally, data on internal output are derived from the INDSTAT 4 2022, ISIC Revision 3 database from the United Nations Industrial Development Organization (UNIDO).
In our work, as already mentioned, we concentrate on the provincial level of analysis.As explained in the introduction, this choice has both advantages and disadvantages and is not without its limitations, especially in terms of data availability.While it offers a different perspective on employment dynamics, we have encountered certain limitations due to data availability.For example, in terms of employment data, we are limited in our ability to delve deeper into sectoral heterogeneity beyond the sections outlined in the Ateco 2007 classification.Moreover, as far as innovation is concerned, we have chosen to focus on patents as a measure of innovation.While this choice is also dictated by the lack of province-wide data on innovation input variables (such as R&D expenditure), we recognise that using patents as a measure of innovation has both advantages and disadvantages, although we are aware of some limitations associated with this choice.It is important to recognise that patents do not necessarily translate into innovations that can be commercialised.Moreover, differences at the sectoral level are particularly significant, especially in the context of the Italian economy, which is largely made up of small and micro-sized enterprises (SMEs), which are particularly prevalent in the southern regions.For such firms, patents may not be a viable option.It is noteworthy that not all innovations are patented, and that the propensity to patent may vary across sectors and over time.Innovation may come from external suppliers, informal sources, and process improvements that are aimed at reducing costs rather than introducing new products.
Nevertheless, we recognise the inherent value of patents as a robust measure of innovation.Compared to proxies such as the number of new products and processes introduced by firms, patent filings suggest a higher level of innovation.This is because firms are aware that obtaining a patent requires a thorough evaluation of the invention by experts who assess its novelty.In addition, patents provide firms with the means to protect their innovative efforts, thereby enhancing their ability to compete effectively against low-cost imports from low-wage countries (Leamer 2007).In conclusion, using patents as a measure of innovation fits well with the scope of our analysis of import competition for several reasons that have been outlined above.However, it is important to acknowledge that with access to firm-level data, further research could deepen our understanding of the factors and mechanisms underlying resilience to shocks.Such research could lead to the identification of policies that can strengthen the ability of firms to anticipate, manage, and respond to these challenges.3

Identifying Labour Market Adjustment to Trade Shocks
This section presents our measure of the China trade shock and main empirical specification, which builds on a body of seminal literature (Autor et al. 2013(Autor et al. , 2021;;Acemoglu et al. 2016).We aim to identify the causal effect of import competition from China on labour market outcomes for Italian provinces and evaluate impacts as the shock intensifies in the early 2000s, and reaches peak intensity around 2010, stabilising thereafter (Autor et al. 2021).
For this reason, we examine exposure to import competition from China for Italian provinces.Following the extensive literature on the impact of increasing import competition on labour market, our measure of trade exposure is the average change in Chinese import penetration across industries, weighted by industry shares in initial province employment.Following an extensive body of literature on China shock, in this section we compute the S pt in Eq. (1) as the share of total imports in manufacturing sector from China (with s pt j s pjt , being the sum of all j sectors in section C;"amufacturing" of Ateco 2007 classification) (Autor et al. 2013(Autor et al. , 2021;;Acemoglu et al. 2016;Bloom et al. 2016;Hombert and Matray 2018;Borusyak et al. 2022).
where I P t M t /(Y 1995 + M 1995 − X 1995 ) is the growth of Chinese import penetration for Italian manufacturing industry in the time interval (2000 to 2018) over the pre-sample level of consumption, t is the base year (2000, immediately before China's entry), and s pt L pt /L T ot pt is the share of manufacturing industry in province p's total employment in the base year. 4We compute import penetration as the growth in Italian industry imports from China in period t and t − 1, M t , divided by pre-sample industry domestic absorption (Italy industry shipments plus net imports minus internal demand (Y 1995 + M 1995 − X 1995 ).Geographical differences in I P t across provinces stem from variation in local industry employment in the base year, which arises from differential specialisation in manufacturing and in import-intensive industries within manufacturing.
One of the concerns about the measure in Eq. ( 1) could be the correlation with unobserved industry shocks in Italy, which also explain employment dynamics.In order to obviate this issue, we follow an extensive literature employing an instrumental variable strategy to isolate changes in Chinese trade that are due to productivity improvements in China, by replacing changes in exports to Italy with changes in Chinese exports to other developed countries (DC) (e.g.Acemoglu et al. 2016;Acemoglu and Restrepo 2020;Autor et al. 2013Autor et al. , 2016Autor et al. , 2021;;Citino and Linarello 2022).
The reasoning behind the widespread use of this tool in the literature is based on the premise that China's transition to a market-driven economy and its rapid integration into global trade, achieved through structural reforms, enhanced its productive capacities in selected sectors in which it held a competitive advantage by allowing it to increase the export in the destinations it served.
The common approach in the literature (to ensure mitigating the risk that Chinese trade patterns, which are common to developed economies, and do not reflect correlated demand or technology shocks among high-income countries) is to select a set of developed countries to identify foreign supply-driven components of Italy imports from China.The countries we use in our setting are the same as those used by Autor et al. (2013), but we include the United States and exclude European countries, which are the main destinations of Italian exports and trade flows.Our set of countries includes the US, Australia, Canada, Japan, and New Zealand.Shared import flows between Italy and this group of countries are more likely to capture the common supply element in Chinese trade, rather than a related demand component (Citino and Linarello 2022).

Results
In our analysis of the relationship between trade shock, innovation, and employment, we focus on the period from 2000 to 2018.In order to measure, on average, the actual contribution of innovation in protecting employment levels in the Italian provinces, we have to mitigate concerns about any unobserved shocks that could explain employment dynamics.As noted in Sect.5, we follow the extensive literature stemming from the work of Autor et al. (2013), by replacing changes in exports to Italy with changes in Chinese exports to other developed countries (DCs).
In addition, the innovative activity of firms, their patents, and thus the accumulation of innovation stock may in turn be endogenous to the productivity of firms located in each province, which, therefore, may be reflected in higher levels of employment.For this reason, to partially mitigate this concern, we interact the trade shock with the stock of innovation of provinces at time T 2000, predicted as per Eq. ( 4), absorbing with fixed effects all time invariant characteristics of provinces (η p ) and annual time trends that might influence their innovative performance.

Stock_I nn pt
Results in Table 1 of the first stage from our preferred estimation of Eq. ( 1)-with the interaction between region and decade fixed effects-show that the predicted stock of innovation act as a good predictor of the actual stocks, and the increase of import competition in other developed countries has a positive correlation with the increase in imports from China for Italy.
In Table 2, we present the correlation among variables (column (1)) from OLS estimation, and the results of 2SLS estimation in columns from (2) to (4).The correlation of trade shock on the variation of employment in manufacturing over the working age population is negative and significant as expected, since workers located in a province can be more exposed to external competition coming from the increase in imports from China.
From OLS estimation results, the direction of the correlation of innovation with the variation of employment, albeit not significant, has the expected sign and, looking at the interaction term, more innovative provinces seem to be able to diminish the impact of the shock coming from increased imports in manufacturing sectors.
The impact of the trade shock has a negative effect on employment, confirming that when economic systems are exposed to exogenous shocks, these can trigger competitive dynamics that can have negative repercussions on employment levels.However, when interacted with the stock of innovation, the effect of such shocks is mitigated by the fact that economic agents can react more effectively to events beyond their control by differentiating their output due to the latter's higher level of innovativeness.

Discussion
Italy's distinct historical economic characteristics make it an intriguing subject for investigating the aftermath of exogenous shocks, providing valuable insights that can guide the formulation of policy recommendations aimed at fostering investments in innovation.
Firstly, this intrigue is heightened by the substantial geographical disparities between Southern Italy (the Mezzogiorno, as aforementioned).These disparities, which trace their roots back to the formative stages of internal economic integration, serve as a rich vein of heterogeneity among the provinces (Rungi and Biancalani 2019;Basile and Ciccarelli 2018).In our context, we tested this heterogeneity in Appendix A.1 by splitting the sample into provinces situated in Southern Italy and the Islands, Central Italy, and Northern Italy.From the insights presented in Sect.2, as delineated by the stylised facts, a concentration of innovation within prominent urban cities-a trend aligned with the established literature (Paunov et al. 2019)-is evident, while the surge in Chinese imports after the WTO accession has exhibited no discernible provincial pattern, as evidenced in Fig. 4. The results presented in Table 3 in the Appendix echo our idea of our argument that the impact of heightened imports from China has manifested as a uniform adverse influence spanning provinces.Particularly noteworthy is the observation that when looking at the interaction term, it becomes evident that elevated levels of innovation, in both Southern and Northern regions, play a role in ameliorating the repercussions of augmented competition stemming from the upsurge in imports.
Secondly, another critical facet to consider when interpreting our results is to propose a plausible mechanism that could underlie both the effects of increased import competition and the role of innovation in protecting provinces from the effects of exogenous shocks.To this end, in Appendix A.2 we extend our investigation by using the variation in the number of registered firms in province p as the dependent variable.This serves as a proxy for measuring the impact on the industrial structure of a given province, and the idea is that the more a province is exposed to a strong increase in imports, the more firms face higher international competition, and the more harsh is the effect on local employment.As shown in Table 4, we use data from Movimprese, accessed via Unioncamere, to examine the potential negative impact of increased competition on the number of active firms, and whether the existing reservoir of innovation within a province serves to mitigate this effect.Furthermore, our analysis extends to the impact on the total number of enterprises (Total) and the impact on enterprises that opt for partnership structures ('società di persone') and corporations ('società di capitali').The results presented in Table 4-taking into consideration all the precautions outlined in Appendix A.2-suggest that the brunt of increased competition due to increased imports falls mainly on partnerships.These typically include smaller enterprises, which seem to benefit more, especially if they are supported by a richer knowledge base in times of crisis.Conversely, increased innovation seems to be more beneficial for corporations.These firms, due to their larger and more structured nature, seem to be less vulnerable to the surge in imports.However, a more detailed analysis at the firm level would be essential to gain a fuller understanding of the precise types of firms that are most affected by increased competition.
Furthermore, the aforementioned heterogeneity in the characteristics of different provinces could also influence the impact of a trade shock on manufacturing employment levels.To further explore the heterogeneity stemming from pre-shock characteristics, we examine the industrial composition of the provinces in Appendix A.3.By controlling for the share of active firms in manufacturing and segmenting the samples into terciles, we find that, despite some instability, the role of innovation continues to exert its protective influence.Furthermore, a higher share of manufacturing firms-indicating a more robust manufacturing landscape-acts as a shield against employment decline compared to provinces with a less robust industrial base in manufacturing (Table 5).
Finally, another potential concern that may arise from our findings pertains to whether the increase in imports from China uniformly affects provinces with existing internationalisation or trade connections with China.Despite the unprecedented nature of the post-WTO accession increase, examining whether the impact on employment levels varies, based on existing internationalisation levels, could yield insights.In Appendix A.4, we address this concern by incorporating controls for provinces' trade relationships.Initially, we introduce the trade openness of province p at the inception of our analysis period (T 2000) into our baseline regression (Eq. 1) to capture the significance of international trade with all partners.The results (Table 6) demonstrate that the effects of the variables of interest remain consistent, with the negative impact on manufacturing employment variation persisting, and innovation still serving as a partial buffer against it.Next, we consider the level of trade that provinces had with China itself at the beginning of the period.This is achieved by integrating the normalised trade balance (N T B pT ), which measures province p's exposure to trade with China at T 2000.Once again, the results (Table 7) validate our findings, as N T B pT is not found to be statistically significant.This suggests that the initial level of exposure to trade with China does not have a substantial impact on the results.Consequently, it appears that the influx of imports from China has had a noteworthy effect even on provinces with pre-existing robust trade relationships, in terms of volume.
Overall, it is noteworthy that our work faces some limitations, both in the analysis of the innovation output measured by patents-for reasons of data constraints on R&D which is not available at province level-and on the employment side since provincelevel data on employment are not publicly available with a disaggregation finer than sections in Ateco 2007.More research is needed to develop a deeper understanding of factors and mechanisms underlying resilience to shocks, and of policies to strengthen the capabilities of firms to anticipate, tackle, and respond to them.Suggested policies should favour a broad range of training and investment activities, leading to a generalised reduction of barriers to technical and product innovation, and also, and most significantly, to more effective organisational structures and capabilities.In this respect, our work has clear limitations that need to be overcome in future research.The characteristics of successful innovation, as well as structural and strategic strengths and weaknesses of provinces cannot be captured in detail at a more granular level by means of the available data.As an example, our analysis is unable to differentiate between smaller-scale (larger-scale) product innovation endeavours, which might lead to relatively slower (faster) demand growth and varying degrees of job losses.Similarly, we are unable to discern whether the dominance of process innovations is driving strategies that are aimed at maintaining price competitiveness in response to heightened international competition.This could result in a reduction of the domestic industrial foundation without facilitating expansion into new, rapidly growing, and product-oriented sectors (Antonucci and Pianta 2002;Pianta et al. 2003).However, given the focus of our study on province level determinants, future research would definitely benefit from firm-level data examining how characteristics affect, and interact with, firm-level factors and strategies.
Our analysis being carried out at province level did not allow us to catch the role of spillovers.Such evidence can only be found via the analysis of the spatial concentration of firms' innovation in cluster or IDs, and searching for evidence on the impact due to intra-sectoral innovation at firm and local level.Hence, to catch the spillover effects we should have a sample made up of firms operating in districts and clusters.In such a production environment, the circulation of knowledge occurs through (formal or informal) networks of division of labour, through the exchange and contagion of ideas and of innovation embedded in the local district atmosphere (Accetturo et al. 2013;Ferragina and Nunziante 2018).This will be an important further direction of research to be pursued.

Conclusions
This paper contributes to the debate on employment changes in turbulent times, emphasising the lasting impacts of temporary shocks on output and employment.Our analysis of Italian provinces confirms the negative effect of trade shocks on employment, highlighting the competitive dynamics that lead to adverse repercussions.Notably, we find that innovation can positively influence employment change, even in industries that are most affected by exogenous events.This mitigating role becomes prominent during deep economic crises.
Our emphasis on the stock of innovation enables us to capture the effect of structural innovation, laying the groundwork for policy implications that are aimed at long-term growth strategies.Our results underscore that innovation acts as a tool for protection and resilience, even in unprecedented circumstances.This insight can guide policy makers in creating an environment that is conducive to business innovation, yielding benefits even in times of global upheaval.
In our analysis, we account for regional heterogeneity and time trends through various fixed effects and by splitting the sample.Despite controlling for all observable and unobservable variables, we consistently observe the strategic role of innovation in shaping province-level employment performance, providing agents with the capacity to react more effectively to events.
While the positive effect of innovative activities is not new, our exploration unveils territorial heterogeneity and emphasises the impact of the stock of innovation over temporary changes in patents during deep economic crises.This focus offers a robust foundation for policy implications in the context of resurging industrial policies and the persistent issue of territorial divides.
As Italy grapples with post-COVID-19 economic recovery, the effectiveness of policies targeting private firm revival is contingent on allocation criteria.Policies directed towards high-return firms have the potential to expedite economic recovery and mitigate regional disparities.Conversely, ill-allocated policies risk deadweight losses and resource misallocation (Cingano et al. 2022).
Over the next six years, Italy will in fact allocate some 104 billion euros to the development of its southern regions in particular.This amount represents more than 25% of the Mezzogiorno's GDP in 2019-a significant share with promising potential.However, these encouraging prospects are tempered by historical precedents.In particular, since 1950 the government has consistently invested around 3% of the Mezzogiorno's GDP each year in regional development.Unfortunately, an examination of the data shows that the southern regions continue to lag behind the northern regions today, just as they did seven decades ago.
Under these circumstances, it is crucial to direct public policies towards rewarding productive activities.As our results suggest, investment in, and incentives for, innovation can serve as a valuable tool with which to protect employment from unexpected shocks.
Further extensions of our analysis could delve into whether the most vulnerable provinces, characterised by low productivity and innovation, stand to gain the most from innovation strategies, in terms of employment growth.In times of turbulence, public policies would likely be more effective if they prioritize less advantaged territories, which are poised to reap substantial benefits from adopting innovative approaches during deep economic downturns.Conversely, economically challenged provinces, especially those with low performance, face significant barriers to engaging in innovation efforts, despite the potential gains.Urgent attention is warranted, in order to implement tailored innovation policies that are aimed at bolstering the weakest and most peripheral provinces.
Funding Open access funding provided by Università degli Studi di Salerno within the CRUI-CARE Agreement.
of Chinese imports on provinces with existing trade connections.The results affirm the consistent impact of the variables of interest, with innovation acting as a partial buffer.Notably, the initial level of trade with China does not substantially alter the results, underscoring the significant impact of Chinese imports even on provinces with established trade relationships.
Testing the Results Among Different Areas: North-South Divide Italy, with its distinct North-South divide, offers a compelling case study for examining the effects of exogenous shocks.This divide, historically rooted in the early stages of economic integration, has been extensively debated (Rungi and Biancalani 2019;Basile and Ciccarelli 2018).
As outlined in Sect.2, innovation tends to cluster in major urban centres, aligning with the established literature (Paunov et al. 2019) Notably, the surge of Chinese imports post-WTO accession exhibits significant heterogeneity, without a discernible size-related provincial pattern (see Fig. 4).
Given these considerations, it is pertinent to explore whether our results manifest distinct patterns when analysing the dataset divided into the Mezzogiorno and the rest of Italy.In Table 3, we present Eq. ( 1) estimations, by splitting the sample between provinces located in Southern Italy and the Islands, Central Italy, and Northern Italy.Broadly, as detailed in the main text, the impact of increased Chinese imports uniformly affects provinces.Notably, in both southern and northern regions, a higher level of innovation serves to mitigate the effects of heightened competition stemming from increased imports.

The Impact on Active Firms
In this section, we examine a possible effect of increased competition due to the increase in imports from China.Specifically, we modify our Eq.( 1) by substituting the dependent variable and considering the change in the number of active firms in province p between the base year (2000) and year t (see Eq. 5). (5) Of course, there are limitations in this approach since the number in registered firms could change because of firms merging, or changing their organisational structure, but by using data from Movimprese, obtained through Unioncamere, and evaluating the change in the number of registered firms, we aim to provide a very suggestive evidence as to whether increased competition has a negative impact on the number of active firms, and whether the existing stock of innovation within a given province partially mitigates this impact.
We also examine the impact on the total number of enterprises (Total), on enterprises opting for partnership structures ('società di persone') and on corporations ('società di capitali') The results in Table 4 show that the impact of increased competition due to increased imports is mainly felt by partnerships.These are typically smaller enterprises that seem to benefit more, especially if they are accompanied by a higher reservoir of knowledge in times of crisis.Conversely, a larger stock of innovation seems to be more beneficial for corporations.Due to their larger and more structured nature, corporations seem to be less affected by the upturn in imports.Consequently, when analysing the variation in the number of registered businesses within the province, categorised as partnerships and corporations, it becomes evident that partnerships exhibit a larger negative variation in the number of registered firms associated with increased imports due to the higher risk to which they are exposed.Their structure (e.g.unlimited partner liability, no asset segregation), indeed, inherently makes them more susceptible to equity, market, and business risk, increasing their risk of failure and consequently, making them less resilient.
However, to gain a fuller understanding of which types of firms have been most affected by increased competition, more detailed analyses at the firm level would be essential.

Industry Structure
In this subsection, we assess how different industrial structures in Italian provinces may influence employment responses.Using data from the Italian Union of Chambers of Commerce (UnionCamere), we calculate the share of active firms in the manufacturing sector relative to the total number of active firms in a given province in 2000.
We categorise provinces into terciles based on the share of manufacturing enterprises.The first tercile (T1) ranges from 6.75% to 10.66%, the second tercile (T2) spans up to 13.90%, and the last tercile includes provinces with up to 34.50%.
The results in Table 5 indicate that considering this aspect of a province's industrial structure does not alter the overall trends.This affirms the negative impact of import competition and the positive influence of innovation, albeit with slightly reduced significance (as observed in column (3)).

Trade Relationships
A potential concern with our estimation is whether provinces that already have robust international trade links might benefit from the surge in imports following major events, such as China's accession to the WTO.To address this concern, this subsection introduces controls for the level of trade links that provinces had at the start of our analysis period.
Firstly, we include the trade openness of province p at time T 2000 in our baseline regression (Eq.1), in order to account for the importance of international trade with all partners.The results (Table 6) show that the effects of the variables of interest remain consistent, with the negative impact on the variation of manufacturing employment persisting and innovation acting as a partial buffer against it.We then take into account the level of trade that the provinces had with China at the beginning of the period.This  is done by integrating the normalised trade balance (N T B pT ) to measure the exposure of province p to trade with China at T 2000.Again, the results (Table 7) confirm our findings, as N T B pT is found to be not statistically significant.This suggests that the initial level of exposure to trade with China does not significantly affect the results.This also suggests that the influx of imports from China has had a significant impact, even on provinces with pre-existing robust trade relations in terms of volume.

Fig. 1
Fig. 1 Innovative performance of Italian provinces (1993-2000).Notes: The figure shows the innovative performance of Italian provinces in the period 1993-2000 measured by the average number of patents.Source: Authors' own elaboration using OECD data

Fig. 2
Fig. 2 Variation of employment in manufacturing in Italian provinces (2000-2018).Notes: The figure shows the change in employment in manufacturing sectors for the Italian provinces between 2000 and 2018.Source: Authors' own elaboration using Eurostat data

Fig. 3 Fig. 4
Fig. 3 Evolution of trade betweenItaly and China (1996-2020).Notes: The figure shows the evolution of imports (solid) and exports (dashed) between Italy and China, and the Import penetration ratio of China in manufacturing (dashed and dotted).The values are in millions of USD for exports and imports (left scale), and as a share of GDP for the penetration ratio (right scale).The vertical line identifies China entering WTO in 2001.Source: Authors' own elaboration using CEPII gravity database

Fig. 5
Fig. 5 Chinese Import Penetration (2000-2018).Note: This figure shows the change in import penetration from China over the period 2000-2018.Source: Authors' own elaboration using COEWEB data

Table 1
First stage results Results from the first stage of 2SLS estimation, where the dependent variable is the change in employment with respect to the base year 2000.All the specifications include region fixed effects and decade fixed effects.Standard errors are clustered at the province level (N 103) and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively

Table 2
The effect of trade shock and innovation on the employment in manufacturing Notes: Results from OLS estimation in column (1) and from 2SLS estimation in columns (2) to (4) where the dependent variable is the change in employment with respect to the base year 2000.Specifications in columns (1) to (3) include region fixed effects and decade fixed effects, in column (4) the interaction.The regressions are weighted using the working age population at the initial year.Columns (1), (3) and (4) include the controls (G D P pT , HC pT ).Standard errors are clustered at the province level (N 97) and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively

Table 3
Heterogeneity among different areas: the North-South divide the dependent variable is the change in employment with respect to the base year 2000.Column (1) shows the results for the provinces located in Southern Italy and Islands; Column (2) for those in Central Italy; Column (3) for provinces in Northern Italy.All regressions include controls.Standard errors are clustered at the province level and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively

Table 4
Heterogeneity among different organisational structures: the effect of trade shocks and innovation on the variation in the number of active companies Results from OLS estimation in column (1) and from 2SLS estimation in columns (2) to (4) where the dependent variable is the change in employment with respect to the base year 2000.All the specifications in columns (1) and (2) include the interaction between region fixed effects and decade fixed effects.Column (1) shows the results for the total number of active firms; Column (2) for firms that choose to organize as partnerships ('società di persone'); Column (3) for corporations ('società di capitali') Columns (2) and (3) include controls.Standard errors are clustered at the province level (N 97) and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively

Table 5
Robustness check on provinces' characteristics: industry structure Results from OLS estimation in column (1) and from 2SLS estimation in columns (2) to (4) where the dependent variable is the change in employment with respect to the base year 2000.Specifications in columns (1) and (2) include region fixed effects and decade fixed effects, in column (3) the interaction.We include the second and third terciles in the distribution of the share of manufacturing firms over the total active firms in province p.The first terciles is the reference category.Columns (2) and (3) include controls.Standard errors are clustered at the province level (N 97) and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively Results from OLS estimation in column (1) and from 2SLS estimation in columns (2) to (4) where the dependent variable is the change in employment with respect to the base year 2000.Specifications in columns (1) and (2) include region fixed effects and decade fixed effects, in column (3) the interaction.We include the trade openness of province p at T 2000.Columns (2) and (3) include controls.Standard errors are clustered at the province level (N 97) and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively

Table 7
Robustness check on provinces' characteristics: trade relationships with China Results from OLS estimation in column (1) and from 2SLS estimation in columns (2) to (4) where the dependent variable is the change in employment with respect to the base year 2000.Specifications in columns (1) and (2) include region fixed effects and decade fixed effects, in column (3) the interaction.We include the Normalised Trade Balance with China of province p at T 2000.Standard errors are clustered at the province level (N 97) and reported in parentheses.*, **, and *** indicate statistically different from zero at the 10%, 5%, and 1% level of significance, respectively