• No results found

University of Groningen Faculty of Economics and Business Master Thesis – International Economics and Business

N/A
N/A
Protected

Academic year: 2021

Share "University of Groningen Faculty of Economics and Business Master Thesis – International Economics and Business"

Copied!
38
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

University of Groningen

Faculty of Economics and Business

Master Thesis – International Economics and Business

Student: Gianmaria Genetlici

Student ID: S3160432

Student email: g.genetlici@student.rug.nl

Date: June 18th, 2019

Supervisor: prof. dr. R.C. (Robert) Inklaar

(2)

The Electoral Consequences of Import Competition

Evidences From Italian Constituents

___________________________________________________________________________

Abstract

Is the rising import competition from China positively affecting the ideological support of constituents toward populist parties? Do these voters belong to specific socio-economic groups of our societies? The present thesis examines the impact of Chinese import competition on electoral outcome for three different levels of election (regional, national and European), over 21 Italian regions and across the period 1994–2013. Results indicate the presence of a considerable ideological realignment toward populist stances, which took place in those regions that suffered the most from increased Chinese import pressure over the years. However, despite previous studies on the matter showed significant concurrent presence of regional class-voting, no strong evidence has been found for the Italian case so far – maybe suggesting the underlying presence of other factors that could not be included in our model.

Keywords: economic voting, import competition, populism

(3)

1. Introduction

It is hard to deny that we are living in an age of political turmoil. On the one hand, we see how the widely accepted benefits deriving from (or attributed to) trade openness and economic interconnectedness have reached a law-like status within the scientific literature: countless publications are providing compelling evidences of the positive effects, in terms of productivity, competitiveness and income growth, for nations removing barriers to free trade. On the other hand, the very foundations of these ideas have recently been opposed by an increasing share of economic actors, calling for protectionist measures often in contrast with these principles.

The political instances of these actors, whose socio-economic whereabouts have been profiled with the low-income, low-educated shares of population, are usually advocated by populist, protectionist, and alt-right movements (Piketty, 2018). It is worth mentioning that neither globalization, nor its correlation with the rise of populist movements, are brand-new phenomena in the economic landscape: their roots and dynamics have been sufficiently analyzed by a consistent body of literature, although the precise connections between economic shocks (whether internal or external) and increased partisanship is yet to be fully understood.

Just as mutual trade benefits, the concept of economic voting reached, in the last decade, a very broad consensus in the academic literature: the idea that voters reward the incumbents for positive economic conditions, while punishing them for negative ones, is widely accepted (Lewis-Beck & Stegmaier, 2000). Furthermore, we know that economic interconnectedness can generally exacerbate these circumstances and, in the absence of systemic redistributive measures, the so-called

losers of globalization might be more harmed than benefitted from trade openness (Colantone &

Stanig, 2018). The resulting wide discontent is then proven to be manifested through an increased support toward parties or movements that support populist, nationalist and protectionist stances (Autor, et al., 2016).

The present research will build up from this point onwards. By adopting a combined approach based on those used by Autor et al. (2016) and Colantone & Stanig (2018) in their respective studies, I will try to expand their findings starting from the shortcomings that these publications present. First, the research by Autor’s et al. (2016) investigates the general right-alignment in U.S. elections due to trade exposure. Two limitations already emerge: the article concerns ideological shifts toward right-alignment only, and the electoral system considered is majoritarian. The present study explores instead the rise of populism, within a proportional electoral system.

(4)

a single country has been picked for this research: Italy, whose electoral system has always been purely proportional, without any considerable deviations from this pattern. While for Western European countries populism can be a rather new phenomenon (and generally populist never really governed), Italy has a well-established history of populist governments all across its history (Verbeek & Zaslove, 2016). It can be argued that Colantone & Stanig (2018) already encompassed Italy in their sample, having focused their research on the effect of import competition over the increased support for nationalist and protectionist parties in the whole Western Europe. As a matter of facts, an aggregated focus could instead be missing valuable details, specially concerning Italian peculiarities. Moreover, their data takes into account only national elections – which could represent a strong shortcoming, considering that different levels of elections could entail different levels of political expression. For this reason, three different types of election have been used in the present research: regional, national and European.

More specifically, Italy presents interesting macroeconomic features that pair well with the purpose of the present research: a high unemployment rate, along with a rather skewed wealth and sectoral distribution. I attempted to correlate import competition as a proxy of these adverse conditions against the surge of populist parties and movements within the Italian constituents. In order to pursue this objective, two specific indices have been computed: the degree of import shock suffered from China, normalized by each region’s employment levels and structure, and the ideological center of gravity, recorded for every district in every election, which provides a quantitative measure of how populist the district’s vote has been. This relation between trade competition, employment and economic voting is rather accepted in the literature (e.g. in Hellwig (2001), that investigates how trade openness dampens the strength of popular support toward traditional parties in certain occupational groups), thus significant results were expected in the Italian case as well.

The relationship between import competition and employment proved to be significant for regions more exposed to Chinese competition. This is completely in line with all the other studies on this subject and confirms the presence of a strong regional pattern among the Italian territory – with the northern regions being far more affected than southern ones, given their sectoral specialization. Unfortunately, no evidence of this pattern has been found across voting behavior. Nevertheless, the relationship between import competition and increased support for populist parties and movements proved to be strongly significant in general and retained a strong explanatory power across all year and regions taken into account. This is once more in line with the main findings of Colantone & Stanig (2018) for Western Europe, even though the magnitude of these effects appeared to be far stronger for the Italian case vis-à-vis the aggregate European sample.

(5)

2. Globalization and voting behavior 2.1. Foundations of economic voting

As Lewis-Beck and Nadeau (2011) report, the very first authors that researched a possible correlation between economic performance and electoral outcome are Campbell et al. (1960), in their book The American Voter. They ask a simple, yet ground-breaking question: “is a person’s economic outlook associated with his partisan choice between presidential candidates?” (Campbell, et al., 1960, p. 397). This question sparked a large subsequent literature. It is helpful here to distinguish, first, the literature that focuses on explaining the relation between economic conditions and electoral participation.1 A second branch of literature explores the effect of economic performances on voters’ behavior, once they reach the ballot-box2. The latter will be the main source of interest of this research.

Despite the many extensions of the work of Campbell et al., their postulate of economic voting remains relevant after almost eighty years: voters will reward the incumbent for good economic times, while punishing them for bad ones. The underlying assumption behind this line of reasoning is that the electorate will share a common (positive) preference toward the broad concept referred to as “economy”. More precisely, it is unquestionably hard to find someone who wishes the economic conditions to be adverse, or distressing: everyone wants to live in a place that is blessed by stable economic growth and general prosperity. This line of reasoning recognizes the economy (and the perception thereof) as a valence issue, a term used to identify an “issue that is uniformly liked or disliked among the electorate, as opposed to a position issue on which opinion is divided” (Brown, et al., 2018, p. 532), according to the formal definition provided by the Concise Oxford Dictionary of

Politics and International Relations.

It is worth mentioning, for the sake of completeness, that an increasing number of studies is recently emphasizing a different point of view, meant to tackle the approach to economic voting from a completely different angle. This more recent body of literature considers the economy to be instead a position issue, hence reducing its overall concurrence to the partisanship-shaping dynamics. The line of reasoning behind this assumption is simple: although everyone could theoretically agree that the economic conditions should ultimately ensure citizens’ well-being, there is much less concrete consensus around the policies that should be pursued in order to reach such a purpose – or, in other words, “the fact that voters all seek economic prosperity has blinded analyst to the notion that they do not all seek the same economic policies” (Lewis-Beck & Nadeau, 2011, p. 293). The bottom line of this approach is that the economy is a divisive (rather than inclusive) subject, that its effects are neither dominant nor negligible, and hence economic conditions should merely be considered as one of many determinants of voters’ attitude and behavior. For example, a study on Canadian elections demonstrates how, from 1945 to 1972, adverse economic conditions benefited the incumbent party, in contrast with the standard result on economic voting (Carmichael, 1990). Moreover, constituents

1 See for example, Rosenstone that, in a very early study on this matter, highlights how “economic problems both increase the opportunity cost of political participation and reduce a person’s capacity to attend to politics” (Rosenstone, 1982, p. 26). More recent findings are illustrated by Carrerasa & Castañeda-Angarita (2019), that analyze the heterogenous impact that macroeconomic conditions have on voters’ turnout, mostly based both on individual socio-economic status and general country’s integration within global economy.

(6)

might, for example, keep supporting the incumbent even in bad economic times, in the presence of strong ideological partisanship (Kayser, 2007). Notwithstanding the relevance of this debate, there is also ample evidence that, while voters are certainly not solely concerned by economic issues, they generally hold them in much more consideration against any other issue taken into account. In their broad review of the economic determinants of electoral outcomes, Lewis-Beck and Stegmaier (2000) find that the economic indicators hold significantly at different levels of multivariate analysis, even when controlling for ideological or historical background. This seems to suggest that, on average, the economy can still be accepted as valence issue, and this is the direction that the present research will follow; further studies on its position property will be left to the specialized literature on the matter. This is also the main reason for which the economy has been used as a proxy of globalization within this research. Globalization falls indubitably within the position issues, with different socio-economic groups having completely different opinions on the matter – ranging from unconditional support to complete aversion (Bhagwati, 2004). The economic well-being, on the other hand, is free from these divisive perceptions, hence providing a somehow neutral benchmark.

Extensive analyses of contemporary expressions of this widely accepted phenomenon can be found rather easily: for example, using individual-level data, Nezi (2012) compares the electoral performance of the New Democracy in Greece, between the elections held in 2004 and those held in 2009. The final results show how, when the economic indicators are unfavorable, the incumbent’s likelihood of victory is close to zero. Being the ruling party in bad economic times, The New Democracy lost 2009’s elections against the Panhellenic Socialist Movement. Similarly, Scotto et al. (2010) demonstrate how the increasing concern related to the macroeconomic conditions played a key role in Barack Obama's victory – with more than 40% of the growth rate of its support attributable to the economic issue. Such a consideration is indeed alarming and, before proceeding any further, it might be worth considering the nature of economic perceptions in mass public. To put it simple, to what extent voters are capable of understanding the economy? Is their perception adherent to the reality, or can it be manipulated by political propaganda?

The very foundations of economic voting are grounded in the assumption that the majority of voters, although likely uncapable of fathoming academic-level issues, still retains an accurate perception of the current state of the economy. A study published by Holbrook and Garand (1996) uses individual-level data to analyze the accuracy of economic perception as a function of personal characteristics, general perceptions of economic threat, interest in economic and politics in general, and exposure to media sources. Their fairly skewed findings highlight how the accuracy of voters’ perception is strongly influenced by “personal characteristics, such as socioeconomic status, gender, race, and age, as well as retrospective personal evaluations, political interest, and media exposure” (Holbrook & Garand, 1996, p. 363). As a matter of facts, such an effect has lately been proved to be dampened by a detailed research published by Deaton (2012), which correlates daily data of self-reported well-being within a pooled sample of respondents interviewed by the American National Election Studies between 1968 and 2008 against the economic conditions; the results prove that voters “perform well in that [since] they respond to income, unemployment, and the stock market in the directions that we would expect” (Deaton, 2012, p. 23), although with few minor caveats1. The same

(7)

conclusion is reached by Lewis-Beck et al. (2013), who investigate the perception of American citizens on their national economic trend, finding compelling evidence concerning their accuracy; these findings are consistent not only on individual base, but they also acquire robustness when aggregated, with awareness of timely economic trends “clearly and accurately” tracking changes both in GDP and in unemployment.

Having therefore ruled out alternative paths and dead ends concerning the (vast) literature on economic voting that is applicable to the present research topic, the review will thereafter be mainly focused on the impact that economic globalization has on the electoral outcome.

2.2. The rise of populism

A conceptual starting point for the present research has been provided by Bhagwati (2004) and Kayser (2007). The first review takes into account the different dynamics that contribute to fuel the general skepticism toward globalization in the recent decades, and how these dynamics are increasingly being represented (and polarized, according to some researchers1) by populist and protectionist parties. The second study investigate instead the degree of influence that globalization can exercise on domestic (electoral) politics, finding considerable proves of increase interconnectedness. The following concept can sum up the bottom line of both reviews: the optimism of the academic world toward the gains associated with an increasing degree of economic integration is recently being replaced by a rising amount of prudence and thoughtfulness toward the issues raised by those that are commonly defined as the “losers of globalization” (Williamson, 2002). To be more specific, these individuals usually belong to the segments of population that suffer most of the adjustment costs, largely in terms of wage and employment share, associated with the shifts in economic production and the slicing of global value chains (Grossman & Rossi-Hansberg, 2006).

With regards to their socio-economic whereabouts, many publications found how they can be clustered in the low-income, low-educated share of the population, usually employed in low-skilled activities whose share of value-added on the global value chains is constantly being worn away by economic interconnectedness. More specifically, since advanced economies are specializing in highly technological, high-skilled activities carried out by high- (and sometimes medium-) skilled workers, only the last, capital-intensive stages of production are retained onshore, while the remaining activities are generally being offshored where unskilled labor is cheaper (Timmer, et al., 2014)2. Thus, since unskilled laborers perceived their intertemporal well-being to be dramatically decreasing due to factors that are completely outside their scope or control, they have been rather fast to identify globalization as the root of all their troubles: as Rodrik (2018) points out, it is definitely easier for the

how, at least three times, self-reported well-being peaked and plummeted, in a seemingly unjustified way (i.e. nothing really important happened, when the anomalous data has been collected).

1 It is worth mentioning on the matter the findings of Rooduijn et al. (2016), which came to the conclusion that a causal effect exists even in the opposite direction, despite what has been considered so far: “populist parties fuel political discontent by exposing their supporters to a populist message in which they criticize the elite” (Rooduijn, et al., 2016, p. 32), usually represented by the other parties within the system. Yet, although their implications are surely valuable food for thoughts, they are far beyond the scope of the present research.

(8)

general public to perceive as “unfair” the foreign competition coming from a far-away, nameless worker that “steals our jobs” by violating both competition laws and social understandings, than actually acknowledging domestic inefficiencies in production or outdated business practices. To sum this up, “what arouses popular opposition is not inequality per se, but perceived unfairness” (Rodrik, 2018, p. 7). But what happens once this anger reaches the ballot-box?

Unsurprisingly, it favors protectionist stances usually advocated by populist (and protectionist) parties, according to the vast academic literature that covered this research topic so far. Reporting only the most recent production on the matter, Alabrese et al. (2018) studied the Brexit vote by combining individual- and regional-level data, in an attempt to sketch a socio-economic profile of the average Leave voter; using individual-level data on thousands of respondents, they find that “voting Leave is associated with older age, white ethnicity, low educational attainment, infrequent use of smartphones and the internet, receiving benefits, adverse health and low life satisfaction” (Alabrese, et al., 2018, p. 132). These findings are not only extremely significant but do present a considerable degree of robustness when the regional-level aggregates are used, when other election years are taken into account and when different measures of globalization are considered. Interestingly enough, this seemingly common geographical connection between impoverished regions and populist vote has drawn a big share of attention in the recent academic production. Two studies can be reported: the first one is (being) published by Piketty (2018), and find evidence of this connection by correlating individual-level electoral data with post-electoral surveys of France, Britain and U.S. elections between 1950 and 2015. The second one has been produced by Rodríguez-Pose (2017), and examines more in depth the common patterns behind the geographical heterogeneity of social discontent: using economic theory derived from the New Economic Geography, he finds that geographical areas of Europe that are lagging behind in terms of development prospects are more likely to channel their discontent through the ballot-box. The foundations of the political populism that they express seems be far more territorial than social. Needless to say, these areas are characterized by a majority of unskilled laborers, are often dependent by welfare redistribution, and present in turn lower-than-average labor productivity, wages, and social development indicators.

(9)

spillovers associated with agglomeration economies, usually represented by big cities (Rodríguez-Pose, 2017)1. Nevertheless, this rise in inequality is, in turn, strongly intertwined with voters’ socio-economic status: high-income highly-educated people usually living in big cities seem to perceive inequality differently, probably due their average better positioning within the job market (Joon Han, 2016).

2.3. The trade origins of populism

So far, we have seen how the unevenly distributed benefits of globalization pushed voters belonging to the low-educated, low-income parts of rich countries’ population to support populist and protectionist coalitions, turning their back to traditional parties. As Rodrik (2018) stated, such phenomenon is largely due to the model of globalization that the Western world pursued in the last decades, “built on a fundamental and corrosive asymmetry […]: make investors happy and the benefits will eventually flow down to the rest of society” (Rodrik, 2018, p. 16). Nevertheless, although all the evidences gathered up to this point indicate globalization as a potential issue, the very concept of “globalization” is still rather vague and all-encompassing. To overcome this inconvenience, the academic literature already produced countless measures and proxies, in the attempt to narrow down, quantify and analyze such a wide phenomenon. The measure that this study uses is the import competition coming from the Chinese market; this choice seemed to be the most fitting, mostly due to the strongly intertwined nature of globalization, that makes endogeneity a serious concern. On the contrary, Chinese import shock can represent a valuable exogenous event for a control and treatment type of study: before its access to the World Trade Organization in 2001, Chinese share of world exports has not been noteworthy; afterwards, it skyrocketed from 4.8% in 2000, to 15.1% in 2010 (Autor, et al., 2016), bearing with it many unforeseen consequences for both the United States and the European Union. One of the most observable result is the dislocation of labor-intensive manufacturing activities, due to the competitive pressure coming from the cheap labor price of Chinese workers. Exposure from Chinese competition seems to have a role in the declining of U.S. manufacturing employment (Autor, et al., 2013), and appears to be responsible for many job lost in the adjustment process: for example, a research by Acemoglu, Dorn, Hanson & Price (2016) shows how the amount of job losses in the U.S. due to import competition from China fluctuates between 2 and 2.4 million.

A word of caution is needed before proceeding any further: indeed, one may object that technological change, primarily represented by workplace automation, affected job displacement consistently. The replacement of low-skilled, repetitive tasks by machines and computers is an ongoing process that caught the attention of many scholars, as well as sparking a good amount of public debate around the world; under this interpretation, all these “lost jobs” could be due to creative

destruction. References to this issue can easily be found within all those narratives that, in most recent

years, have inflamed public debate around topics like “workplace automation” or “job substitution”. Not surprising, populist parties are often rather close to these protectionist (and sometimes luddite) stances. Truth be told, these concerns find a shaky ground within the academic literature: already Autor (2015) drew a thick conceptual line between “technological complementation” and

(10)

“technological substitution”, demonstrating how all tasks complemented (rather than substituted) by technology will not only be safe, but will even possibly increase as the level of technological integration with our daily routine tasks increases at the present level. Moreover, despite Acemoglu and Restrepo (2017) offer a rather drastic picture, where an additional robot per thousand workers is capable of reducing the employment to population ratio of 0.34 percentage points in the worst case scenario, Graetz and Michaels (2018) do not seem to generally share such a negative view on the rate of employment in general. Their results show indeed a rather predictable reduction in the employment share, but mostly at the expenses of low-skilled workers – being medium- and high-skilled complemented by technological integration. For these reasons, the present study will not take technological change into account. Although it would be interesting to check whether the increased workplace automation has a significant effect on job precariousness, hence resulting in increased support for populist and protectionist parties, such an analysis is far beyond the scope of the present research.

Following the dynamics that have been described up to this point, we expect this striking decrease in employment to have a noticeable consequence in voters’ attitude toward the ruling parties and elites. This expectation is entirely met and extensively confirmed by the literature on the matter: for example, Jensen et al. (2016) used national- and county-level data to show how the U.S. trade balance strongly affected voters’ behavior: incumbents in swing states that present both a high share of low-skilled workers and experienced a considerable amount of trade exposure proved to be particularly vulnerable to losing votes1. A connection between the more specific Chinese import competition and electoral outcome has instead being published by Autor, et al. (2016), who analyze the outcomes from the 2002 and 2010 congressional elections and the 2000, 2008, and 2016 presidential elections. Over these years, a significant ideological realignment has been detected, mostly centered in trade-exposed local labor markets, that started well before the divisive 2016 U.S. presidential election. Throughout the period taken into account, U.S. districts more exposed to Chinese competition have been found more likely to polarize their voting attitude against the candidate’s ruling party, and more in favor of populist and protectionist stances, generally attributable to runners of the Republican Party. Proceeding from these findings, Colantone and Stanig (2018) apply the same methodology on 15 Western European countries over the 1988–2007 timespan; their results are generally in line with the ones obtained by the previous authors, confirming the existence of a connection between the level of Chinese import competition and the increased support for populists and protectionist parties.

As mentioned before, in order to properly investigate this hypothesis, as well as its social, economic and geographical implication, a single country has been chosen – namely Italy – instead of a group of countries or a region. The next chapter will extensively provide reasons for such a decision.

2.4. Our study case: Italy

Narrowing down the scope of a whole research to a single country might, at a first glance, sound like a rather limiting decision; in fact, it is unexpectedly appropriate, when it comes to the

(11)

Italian case: the peculiar geo-economic situation of this chosen country bears with it many unexpected advantages capable of enriching, instead of restricting, the discussion on this field. Henceforth, I will claim that three main arguments can be presented in favor of this decision, and these are: first, the considerable economic relevance of Italy within the European Union, paired with its disappointing macroeconomic conditions since the end of the global financial crisis of 2009; second, the extremely skewed per capita income, inequality distribution and polarized production structure, along with its sharp territorial differences in terms of efficiency, labor, skill allocation and unemployment that, altogether, can produce completely different results on a regional basis; third, the complete absence of literature concerning the trade origins of Italian populism, despite the constant historical presence of this phenomenon among Italian politics.

The first point concerns the general Italian economic relevance, which is still far from being neglectable: according to The World Bank (2019), Italy is the ninth biggest economy in the world in terms of GDP; it ranks third in the Eurozone (after France and Germany), and fourth for total volume of export (after Germany, the Netherlands and France). The decomposition of Italian GDP is largely in line with the one of a high-income nation, where services account for almost three quarters of total gross domestic product, and industry for approximately one quarter; Italy’s main share of export is currently concentrated in machinery and vehicles, which account for more than 32% of the total exports, while travel, business and transport services account for 75% of its exported services (International Trade Centre, 2019). Yet, despite this seemingly positive economic outlook, Italy suffers from decades-long political instability, chronical economic stagnation and lack of structural reforms; on top of that, the global financial crisis of 2009 worsened the already compromised situation, causing a 5.5% contraction in GDP levels only during the first year. Since then, Italy has considerably underperformed and shown no clear trend of recovery; nowadays, the GDP growth rate is still close to zero (around 0.9% per year, according to the OECD), and the government debt-to-GDP ratio is stable around 130% in the last years. In view of the fact that more than 62% of this public debt is being held by countries within the Euro area (della Corte & Federico, 2016), any fluctuations in Italian economic conditions are being regarded as extremely concerning for the general macroeconomic stability of the European Union as a whole. Bearing in mind all the strong connections between economic performances and political outcomes that have been highlighted so far, I consider the Italian case worth of receiving a good deal of attention, given all the possible future implications in terms of political and macroeconomic stability, both at the national and at the supranational level.

Secondly, Italy shows a very peculiar (yet spatially homogenous) economic outlook: as image 1 shows in terms of GDP per capita distribution and unemployment rate, a clear division line can be drawn between the highly-industrialized northern part of the country where more than 75% of the GDP is produced, and the agricultural-dependent southern part, accountable of the rest of gross domestic product. Not surprisingly, unemployment is definitely higher in the latter. According to the BES 2018 report, issued by the Istituto Nazionale di Statistica (National Institute of Statistics, or Istat in short), these differences do not show any converging trend: during the past decades the south of Italy kept lagging behind in nearly any aggregated score taken into account within this document1

(12)

(Istat, 2018), leaving little confidence in any deviation from such a trend in the near future. As it can be easily imagined, these regions and their cities fall perfectly within Rodríguez-Pose’s description of “places that don’t matter”: they are poorer, strongly dependent from welfare assistance and centralized income redistribution, mostly specialized in the production of labor-intensive goods (e.g. agriculture, manufacture, fishing); since the past chapters highlighted a recurring trend concerning the territorial nature of populism and its inherent concentration among the low-educated, low-income part of the population, we would expect the comparison between northern and southern regions of Italy to be more than fruitful under an academic point of view: we are expecting the (poorer) south of Italy to be strongly polarized toward the stances of populist and protectionist parties, while a bigger support for traditional parties should be found among northern (richer) regions. The findings will highlight how this is not necessarily the case, since populist propaganda seemed to work with a concerning efficiency pretty much everywhere across the Italian borders.

Third, it is worth stressing out that there is, at the present moment, no literature concerning the impact of trade competition over the Italian voting polarization toward populist and protectionist parties. Indeed, different scholars have studied Italian populism under various points of view: for example, Barone et al. (2016) investigate the effect that immigration has on populism over this region, while Albertazzi and McDonnell (2008) focus more on the social and cultural cleavages (e.g. the strong differences between national belonging, the strong regional belonging, the historical rhetoric, etc). Other studies, such as Colantone & Stanig (2018), take Italy into account only as a part of a more general (usually aggregated) sample, hence overlooking all the peculiarities that have been stressed out so far concerning the Italian economic and social conditions. Nevertheless, a more careful analysis of the Italian case could reveal historical insights that have been overlooked so far: as mentioned before, the Italian politics is often regarded as being dominated by populism, ever since the end of the past century. During the period between 1994 and 2011, Italy has been ruled by four coalition governments, largely dominated by Forza Italia1, Silvio Berlusconi’s party. Afterwards, in 2013 parliamentary elections, one fourth of the Italian electorate voted for a Beppe Grillo’s

Movimento Cinque Stelle2, a party that has been openly labelled as populist and protectionist, supporting strong anti-establishment stances. Has populism always been fueled by trade? Or is this connection relatively new and strongly dependent from the latest wave of globalization?

Having therefore ruled out all possible alternatives and provided the reader with a consistent review of the state of the literature concerning the object of this study, along with the motivations that led a single country to be choose, we can proceed to formalize our hypothesis for the research question, namely: “Chinese import competition increases voters’ partisanship toward populist and

protectionist parties”. The following chapter will highlight the econometric approach that has been

adopted, in order to provide all these questions with a proper answer.

safety (murders), safety (other crimes), life satisfaction, landscape and cultural heritage, environment, research and innovation.

1 Commonly referred to as Go Italy on international publications.

(13)

Image 1

Comparison between average regional per-capita GDP (left) and average regional unemployment rate (right) during the period from 1994 to 2013

Source: author's calculation based on data from the Istat. The map shows the average values for the period taken into account within the whole research (1994-2013), and it is meant to highlight the high degree of geographical heterogeneity among the Italian peninsula. The left-hand map shows the GDP per capita distribution, while the right-hand map shows the unemployment distribution. The color scale is meant to follow an increasing logic: the darker the color, the higher the measure. The per-capita GDP value ranges from a minimum of € 14.000/year (Calabria) to a maximum of € 30.496/year (Valle d’Aosta). The unemployment rate ranges from a minimum of 0,0497 (Trentino Alto Adige) to a maximum of 0,2372 (Sicilia).

3. Models and estimators

The model that has been used to formalize our hypothesis is deeply grounded into the existing literature on the matter (e.g. Autor, et al. (2016) and Colantone & Stanig (2018) for similar formulations). Nevertheless, it has been tailored to our study case with few important modifications, that will be highlighted momentarily. First of all, electoral result will measure the increased partisanship toward populist and protectionist parties and will be considered as a function of import competition from China. Afterwards, control variables and fixed effects for both years and regions have been introduced, in order to capture any variation in outcome that it is relative to these two dimensions and not attributable to any other explanatory variable. The main regression line will then look as follows:

𝐺"# = 𝑎"+ 𝛼#+ 𝛽)𝐼𝑆"#+ 𝒛

-"# + 𝜀"#. (1)

(14)

𝐺"#is meant to be a function of the alpha intercept terms αr, which captures the region-specific effect, and αt that is, instead, supposed to collect the year-specific effect. Here we can highlight a substantial difference with the publication from Colantone & Stanig (2018), where the intercept term is aggregated in a general region-year effect αrt. This decision sparks from the idea of properly investigating geographical and yearly pattern in the evolution of Italian populism. More formally, we expect region specific effect to be positive and highly significant in all region that, throughout the years, have been exposed to a greater deal of Chinese competition. Moreover, given the strong historical presence of populism in Italian politics, we expect year-fixed effect to be significative for the whole period taken into account – with its magnitudes varying accordingly to yearly macroeconomic factors.

The beta estimator collects instead the effect of the independent variable: Chinese import shock IS, calculated on the yearly Italian import flows from China, measured in thousand Euros and normalized over the employment structure of each region r, for each year t. The variable has not been lagged, in accordance with the rest of the studies that have been addressed so far – a more extensive discussion on this matter will follow in chapter 4.2. Furthermore, according to the type of relation that has been drafted so far, import shock’s coefficient is expected to present a positive value: the bigger the regional exposure within a specific year to Chinese import competition, the stronger the populist turnout. No specific expectations upon its magnitude are present, although we could imagine it to be somehow smaller than the one observed by Colantone & Stanig (2018) for the whole Western Europe, since Chinese import competition proved, for the Italian labor-intensive, low-technology production, to be below the European average (Bugamelli, et al., 2010).

Finally, control variables are encompassed into a general vector z’, that include regional employment composition, changes in regional wealth, inequality index and changes in regional immigration from non-European countries. The use of these controls constitutes another deviation from the usual econometric approach: to my knowledge, no study has so far included controlled, within this specific relation, for anything more than employment effect. Naturally, as I will argue in the next chapter, the presence of these controls is backed up by the academic literature on trade competition and populism.

Moreover, since our interest is to investigate whether the local macroeconomic conditions can result in systematic regional differences in voting behavior among Italian constituents, a new regression line will be estimated. This equation, although being in line with the literature that we discussed so far, adds a factor that has not being considered before. It will then look as follows:

𝐺"# = 𝛼#+ 𝛽)𝐼𝑆"# + 𝛽/𝛿1+ 𝒛

-"# + 𝜀"#, (2)

(15)

manufacturing sector that could be more exposed to Chinese import competition – thus increasing their support toward populist parties.

3.1. Econometric approach and endogeneity

Concerning the econometric approach, the type of data allows both ordinary least squares (OLS) and instrument-variable regression (IVR) through a two-stages least squares (2SLS) method. Both methods will then be used and compared. By using the first method I argue that the electoral outcome is a linear function of our given parameter, that clustered standard error will be used in order to tackle any issue related to heteroskedasticity, and that the residuals are normally distributed and serially uncorrelated. Moreover, in the attempt of minimizing any form of collinearity, correlated variables will be regressed separately and, additionally, the variance inflation factor (VIF) will be computed for each estimate as an additional test.

The second method uses instrumental-variable estimation to address any possible concern of endogeneity on the import shock with respect to gravity score. This decision is motivated by earlier literature (e.g. Autor et al. (2013), Autor et al. (2016), Colantone & Stanig (2018)); the logic behind it is that the observed differences in import volume are caused both by exogenous factors (such as changes in the Chinese supply), and endogenous ones. Like any other demand for a specific good, economic theory predicts many possible reasons for its outward or inward shifts, ranging from simple changes in real income to more complex changes in intertemporal consumption. The government could influence these changes as well: as Colantone & Stanig (2018) suggest, policy makers could strategically protect specific regions, for example those where manufacturing industry is concentrated, from exogenous import shocks, in the attempt to favor local production (hence securing additional votes). This concern is far from being remote: although Italy belongs to the EU, which has exclusive competence on trade policy, domestic politicians are still capable of influencing domestic demand through a system of positive incentives and local market regulations. Instrumenting trade shock addresses such an issue and ensures that our regression exclusively captures any variation that is due to exogenous changes in Chinese supply, rather than any other region-specific domestic factor that might be affecting electoral results.

The instrument will then be computed as follows: 𝐼𝑛𝑠𝑡𝑟𝑢𝑚𝑒𝑛𝑡"# = ∑ :;<(>?)

:;(>?)

∆CDEFGHIJKL<(>)

:<(>?)

M . (3)

This equation for the instrumental variable is virtually identical to (6), the equation used to calculate the import shock. It correlates sectoral employment with import competition from China and will thus be treated extensively within the section reserved for such a purpose (section 4.2). The only difference, which is the core of the instrument, is the presence of ∆𝐼𝑀𝐶ℎ𝑖𝑛𝑎𝐴𝑈𝑆M(#) instead of

∆𝐼𝑀𝐶ℎ𝑖𝑛𝑎M(#): in other words, we will be using the Australian imports from China during the same time-period; data concerning this trade flow will be sourced from the UN Comtrade Database.

(16)

Goldsmith-Pinkham, et al (2019) for a more extensive discussion concerning instrumental variable estimation).

Applied to our specific case, one could worry about (correlated) trade shocks across countries that simultaneously affect imports from China, both for Italy and for Australia. Notably, after formal testing, this seems not to be the case; however, the testing statistics on the instrument have been included under each regression table.

4. Data and measures

As mentioned before, building up a consistent dataset that pairs trade statistics with electoral data is far from being an easy task, when it comes to our chosen country. Concerning its political system, Italy has been capable of producing a massive amount of parties since its first democratic elections in 1948: a simple computation on the electoral archives can highlight the presence, across 70 years of history, of more than 380 different parties – namely, an average of 32 parties available for each given election. Concerning instead its trade statistics and sectoral employment, there is no consistent data across any given time period: the ATECO1 classification used by the Istat has been adopted in 1991, updated in 2002 with completely different categories, and then entirely replaced in 2008. Nevertheless, the following chapter will deal more extensively with these issues, giving a detailed description of all the setbacks that have been encountered, paired with the strategy that has been used to overcome them.

Under a geographic point of view, Italy is a relatively large peninsula with a population of 60.4 million, spread quite homogenously along the whole territory of approximately 300,000 km2. In order to properly investigate the consistent geographical differences that we expect to find between the northern and the southern part of the country in a standardized and codified manner, the official Nomenclature of Territorial Units for Statistics (NUTS) partitions, developed and regulated by the European Union, will be used to refer to Italian administrative subdivisions. The level NUTS-0 represent the nation-state in its entirety, with its territory and borders as recognized by the international community. The level NUTS-1 divides the national territory in 5 geographical macro-regions, that do not have any legal or political representation. This classification is typically used for aggregated statistical purposes by the Istat and it will be later used by the present research to investigate the presence of any structural aggregated differences between macro-regions. The level NUTS-2 represent our main geographic scope: it divides the state in 21 official regions (instead of the 20 that are described by the Italian Constitution)2 with legal and political representation; these are the first-level administrative divisions of Italy and will be the main focus of our study. This is the most common level of territorial aggregation used in Italy for statistical purposes: regions are recognized by the Italian Constitution Law as “autonomous entities with their own statutes, powers and functions according to the principles arranged in the Constitution”3. Last, the level NUTS-3

1 ATECO, which stands for ATtività ECOnomiche (i.e. economic activities in Italian), is an alphanumeric classification system adopted by Italy and based on European NACE.

2The “additional” region is represented by the split of “Trentino-Alto Adige/Südtirol” within its two territorial components “Trentino” and “Alto Adige/Südtirol”, mostly due to the autonomy that this territory has been granted with by the Italian Constitutions; this decision has its roots in the strong, inter-regional cultural differences that characterize this region. Since this peculiarity has relatively neglectable economic influence, any data concerning the two sub-regions will simply be aggregated into the main one.

(17)

further divides the 21 regions in 103 provinces, or institutional bodies of second level with legal and political representation, that include 80 active provinces, 2 autonomous provinces, 6 free municipal consortia, 14 metropolitan cities, and Aosta Valley region.

Being our purpose to assess the effect of Chinese trade competition on the Italian electoral outcome, the following data has been acquired: data on electoral outcome (i.e. which party is running for which elections, the voting shares for each party in each election, the ideological alignment of each party in each election), data on industrial structure of the regions and on their economic outcomes (e.g. sectoral employment, unemployment rate, production structure, per-capita income, etc.). All figures are both at a national (NUTS-0) and a regional (NUTS-2) level and cover a timespan of approximately 20 years (i.e. from 1994 to 2013). This specific timespan has been picked for two specific reason: on the one hand, to maximize cross-sources data availability; on the other, we expect fruitful comparison between data prior and following both the Chinese access to the World Trade Organization (2001) and prior and following the global financial crisis (2009).

Regrettably, since the institutional territory and borders of provinces (NUTS-3) have sometimes changed along the years (mostly to accommodate either economic or demographic fluctuations), all data that refers to this NUTS level is rather shattered, difficult to extract, manipulate and attribute, being its observations referred to territorial units that are not consistent throughout the years. For this reason, this research will use regional data (NUTS-2), with a short macro-regional comparison (NUTS-1) when it comes to estimating both (1) and (2).

The final data sample covers 28 regional, national and European elections between 1994 and 2013, for a total of 240 observations. Detailed descriptions of each variable taken into account, its measurement, its source and its method of collection will follow in the next sub-chapters: the first one will concern the electoral outcomes and party classification, the second one will deal with economic statistics and import competition, while the third one will analyze all control variables that have been selected to fit into regression equations (1) and (2).

4.1. Electoral outcomes and party classification

(18)

the number of observations per year is far smaller. Our dataset covers regional elections for the years 1995, 2000, 2002, 2003, 2004, 2005, 2008, 2009, 2010 and 2013. No substantial differences in political alignment between elections is expected, except those that are already explained by our model. This expectation will be formally tested later on, in chapter 5.

We can then define our variable pslt as the votes share for party s, at the regional level r, in election year t. Moreover, since our research focuses on a very specific party type (i.e. populist parties), a scientific method capable of classifying their stances and assigning them a specific populist score has been adopted: first, we make use of the Manifesto Corpus (Volkens, et al., 2018), a project that provides updated coding of each party’s political programme alongside different political dimensions; then, making use of the same methodology described by Lowe, et al. (2011) and lately applied by Colantone and Stanig (2018), we calculate a specific “ideology score” that is party-election specific (i.e. accounts for potential party realignment, over the ideology spectrum, between every elections) but consistent across all geographical levels taken into account. This allows us to define the score Populismsrt as the populist score for party s, at the NUTS level r, in election t, calculated as follows:

𝑃𝑜𝑝𝑢𝑙𝑖𝑠𝑚X"# = log(. 5 + 𝑧X"#_ ) − log(. 5 + 𝑧

X"#a ). (4)

Following what has been said so far concerning populist stances and positions, the indicators that have been used to classify a party’s populist score are essentially four: the expression of negative feelings or messages toward the European Union (e.g. exiters or eurosceptics), a political propaganda generally stressing the importance of national values and traditional way of living, any form of skepticism shown toward multiculturalism, multicultural societies, or integration of different cultures, and a favorable view toward protectionist policies and market protection1. By construction, the

Populismsrt index can indeed range between positive and negative figures; it presents a minimum value of -0.7522 assigned to Italia dei Valori in year 2008, that is pretty in line with this party’s policy: being a left-oriented party with consistently progressive ideologies, it represents stances completely opposite to those considered to be populist (e.g. multiculturalism, economic and social integration, etc.). Comparatively, on the same year, the Lega Nord2 shows a score of 1.2471. The maximum value is instead of 1.7113 and belongs to the Partito Popolare Italiano in year 1994; strange it might sound, this party supported, along with Christians, all stances that we would now observe as compatible to the populist agenda. The average populist score for the entire dataset is 0.701, thus highlighting an unsurprising tendency, for Italian political dialogue, to express a certain degree of support toward populist stances in general (e.g. check for example Verbeek & Zaslove (2016) for an in-depth historical analysis of Italian populism).

Afterwards, using Colantone and Stanig’s (2018) gravity equation, we can then compute the ideological center of gravity of each region, that will be nothing else but the average populist position scores of each party involved in the round of election, weighted by their vote shares within each region. This measure is indeed sensitive to the distribution of policy positions (i.e. the quantity of stances toward populism) and vote shares (i.e. increased or reduced support for populist parties

1Namely, using the Manifesto classification, the values that fall within z+ are 108, 601, 608 and 406. Conversely, the values that fall within z- are 110, 602, 607, 407.

(19)

à-vis for traditional parties); for example, it might increase if, ceteris paribus, a populist party increases its populist stances. The center of gravity G is then calculated as follows:

𝐺"#=∑ bc;>∗debfgGX1c> h

cij

∑hcijbc;> , (5)

where once more s indicates the party, at the NUTS-2 level l, in election t, while the populism score is calculated using the formula described in (4); by using instead equation (5) we can extract an efficient map of how populist each region is, within each time and election taken into account. Image 2 applies the method that have been described so far: the regional scores for the first and last three years of our sample have been aggregated, in order to include all levels of election within our analysis. The result is, with the notable exception of few regions, a general surge in populism across all Italian peninsula, paired with a relatively visible polarization of Italian constituents: as it can be observed on the legends, the minimum and maximum gravity scores for the second sample cover a much wider interval.

According to our map, northern regions seems to be more affected. This is generally in line with the first expectation formulated in the previous chapter: the northern part of Italy presents a considerably higher degree of industrialization, compared to both center and southern part. Since the surge in Chinese import mainly affected the manufacturing sector – rather than agricultural or services one – it is not surprising to observe a surge of populism in areas whose aggregate employment is intensive in labor-intensive manufacturing activities.

Image 2

Regional-level comparison between initial- and final-period elections in terms of average gravity score

(20)

higher the gravity score – hence stronger the populism. The minimum and maximum values for each year-range are indicated in the legends.

In order to thoroughly investigate the evolution of populism across our sample, an in-depth yearly analysis of our gravity score has been undertaken, with the resulting figures being plotted on Graph 1. Concerning the latter, a brief historical overview of Italian voting attitudes can provide additional context, highlighting the high degree of connection between electoral history and computed data. First of all, it is worth noting that the minimum average score is generally smaller than 1 and constantly above 0, across the whole period taken into account; this is rather in line with Italian voting trend, characterized by a rather moderate median voter (e.g. Milanovic (2000) for a deeper analysis on the median voter characteristic across 24 countries – among which Italy), whose favor has always been casted toward political coalitions formed by centrist parties: elections between 1943 and 1992 have been dominated by the Democrazia Cristiana1 (DC), a moderate Christian party supporting temperate interventionism and conservative social stances on national policy, Atlanticism and international integration on foreign policy; regarded as the archetype of the traditional party system, DC’s average gravity score is 0.259, with a standard deviation of 0.156.

Graph 2

Regional gravity scores and national trend by year (1994-2013)

Source: author's calculation. The electoral data has been gathered from the election’s database of the Ministry of Internal Affairs, while the populist score for each region, in each period, is based on the classification provided by the Comparative Manifesto Project. The resulting gravity score has been calculated using (5). The graph shows the trend (red line) in the unweighted mean score (red diamonds) throughout the years 1994-2013, and the regional outcomes of each election considered for that specific year (blue circle).

(21)

Afterwards, the involvement of DC’s leadership into the Mani Pulite1 scandal in 1992 marked the end of Italian’s traditional party system: elections between 1994 and 2008 have been characterized by the stable presence of Forza Italia (FI), guided by Silvio Berlusconi, as the leading party within the majority of government’s coalitions – with few unsuccessful left-centrist governments in between. While being recognized as a centrist party as well, FI supported policies that can be regarded as far more populists compared to its Christian counterpart. Such a choice is indeed reflected by its mean gravity score of 0.845, with a standard deviation of 0.011 – 0.586 points higher compared to the DC. Afterwards, both Berlusconi involvement in countless personal scandals and the effects of the Great Recession called for a general contraction in trade volume, resulting in the so-called austerity policies put into place by Mario Monti’s government (from 2011 to 2013).

On a final note, few caveats must be addressed before moving on to import competition. More specifically, and as mentioned in the previous chapter, Italian history is characterized by a disproportionate amount of parties that, with only few notable exceptions, do not last long on the political landscape. As a consequence, any party-based cross-electoral comparisons is made a rather fruitless task (i.e. we cannot observe directly the rise of populism as the decline of support toward traditional parties after each round of election): any given party might be running for few electoral rounds, and then suddenly break up due to either changes in internal alliances or rebranding efforts. For this reason, only parties that have scored a regional voting share of 5% or more have been considered for the purpose of this research. Such a choice is not random: they approximately correspond to those whose manifesto has been analyzed and coded by the Comparative Manifesto Project, and it is the legal threshold that parties have to achieve, in order to gain at least on seat in the Parliament. Parties below this share do not gain any seat, thus do not exert any form of control on the official political agenda.

Nevertheless, a similar line of reasoning applies to party coalitions: it is common practice, for parties that are unsure about their electoral result in any given region, to gather up in bigger coalitions in order to attract a more consistent share of votes; the populist score of the list is then approximated to the score of the biggest party present in the list (i.e. the party that, at the national level and for the year taken into account, receives the bigger share of votes).

4.2. Economic statistics and import competition

The purpose of this research is to regress electoral outcomes at the regional (NUTS-2) level against Chinese trade shock. For this reason, an indicator of regional exposure to Chinese competition is needed; employing once more the same methodology already consolidated by the academic literature on the subject (e.g. Autor, et al. (2013) and Colantone & Stanig (2018)), the import shock

IS will be composed as follows:

𝐼𝑆"# = ∑ :;<(>?) :;(>?) ∗ ∆CDEFGHI<(>) :<(>?) M . (6)

Here, two main components concur in shaping the import shock index: sectoral employment and Chinese import shock. The first term represents the sectoral employment component, which is simply calculated as the ratio between employment L in industry j for region r (i.e. the number of

(22)

workers employed for specific industry in a specific region), divided by the total number of workers

L employed in the same region (i.e. the total number of workers for that region) – both taken at the

beginning of sampling period t0 (i.e. that would be January of each specific year). The right-hand side represents instead the Chinese shock: it enters the equation as ∆IMCN, namely the yearly change in national imports from China, for industry j in year t; this is normalized by the total employment L of industry j, at the beginning of sampling period t0. The idea behind this approach, initially developed by Autor, et al. (2013), is that every region will be more or less exposed to changes in Chinese competition, depending on its industrial specialization pattern existing before the impact of the shock (hence the t0 sampling period).

This means in turn that every change in country-wide imports from China will affect every region differently, depending on the share of workers that have been initially employed in that specific sector: at any given point, any given change in import at the country-industry level will have a stronger effect on regions whose labour force is more intensively employed in that specific sector. Our expectation is that import competition will be stronger in those regions that are characterized by a greater share of workers employed in low-skilled labor-intensive activities and industries, such as manufacturing of textiles or electrical products. As previously mentioned, these regions roughly correspond to the northern part of Italy.

The data needed to compose such an index has been extracted from three specific sources. First, employment data has been sourced both from Istat’s data bank (Istituto Nazionale di Statistica, s.d.), and from the OECD (2019) archive whenever not available through the first source. Second, trade data has been retrieved from Comext, Eurostat’s database about trade and goods, in the form of import trade flow at the product level. Afterwards, due to the presence of consistent discrepancies in product codification between European NACE (used to codify trade data) and Italian ATECO (used to codify sectoral employment), all data have been painstakingly aggregated and recoded into three main categories (no comparison would have been possible otherwise): agriculture, hunting, fishing and primary sector (a), manufacture and industrial sector (b), tertiary and services sector (c).

Table 1

Comparison between initial- and final-period’s statistics for import competition index

Source: author’s calculation. The trade data has been extracted from Eurostat Comext trade database, while the sectoral employment data has been extracted from Istat’s data bank. The import shock for each region has been calculated using (3) and has then been aggregated for each year. Statistics shown are mean, standard deviation, 10th percentile and 90th percentile. The first four and last four

years of our 1994-2013 sample have been picked, for comparison purposes.

(23)

competition on these two sub-sectors can vary considerably. Unfortunately, data on sectoral employment has not been gathered consistently by the Istat for the whole period taken into account by this research. Time-series estimation could eventually overcome this complication, but such a reconstruction is far beyond the scope of the present research and can only be included within the possible extensions of our topic. Nevertheless, the data displays a behavior that is rather in line with both our expectations and official trade statistics, despite the absence of such a specification.

Table 1 shows descriptive statistics for the explanatory variable across the first and last four years of our sample. A positive number indicates a growth in terms of Chinese import competition in the region, compared to the previous year; similarly, a negative number suggest a decrease of the same measure, hence a contraction in trade inflow coming from China. Two observation can be made: the first one is that the import shock always presents a bigger effect within the secondary sector, hence manufacturing and industry, ever since the first year taken into account. Secondly, there is a striking difference between the magnitude of shock indices (hence the volumes of trade involved) during the first and the last years of our sample, across all sectors.

Image 3

Regional-level comparison between initial- and final-period import shock indices

Source: author’s calculation. The map shows the unweighted average values for the first and the last four years of our sample, which ranges from 1994-2013. The trade data has been extracted from Eurostat Comext trade database, while the sectoral employment data has been extracted from Istat’s data bank. The import shock for each region, in each year, has been calculated using (3). The color scale is meant to follow an increasing logic: the darker the color, the higher the import shock. The minimum and maximum values for each year-range are indicated in the legends.

(24)

virtually identical, when Image 3 and Image 2 are compared. The only exception seems to be represented by Trentino Alto-Adige (the northernmost region on the map) that, despite experiencing a negative trade shock for the last four-years period, appears to show an increased deal of support toward populist parties. This could be related to strong presence of the Südtiroler Volkspartei, a regional party supporting strong secessionist stances that, since the 2009 economic crisis, gained a considerable deal of popularity.

Graph 2

Import trend by year and sector (1994-2013)

Source: author’s calculation. The trade data has been extracted from Eurostat Comext trade database, while the sectoral employment data has been extracted from Istat’s data bank. The import shock is shown for the three aggregated sectors taken into account: agriculture (blue), industry (orange), services (red).

(25)

of employment and general expenditures and has been followed by a strict period of economic

austerity between 2011 and 2013.

Further descriptive statistics, as well as graphs for the most interesting indicators, have been placed into the appendix in order to be easily accessed by the reader.

4.3. Control variables

In order to properly isolate the effect of Chinese trade shock on electoral outcome, few control variables are needed, since consistent strands of literature proved them to somehow affect the relation we want to investigate. All figures have been collected at a regional level and refer to i) changes of immigration, ii) changes in inequality, iii) changes in the gross domestic product. All these variables enter the regression line as vector 𝒛

-"#, and their estimates are expected to bear different signs and

magnitudes; an extensive analysis of each estimator will follow in the next paragraphs. Moreover, both changes in the unemployment rate and subsequent changes in the employment structure will be investigated, in order to properly assess the magnitude of the impact of Chinese shock on the work market.

The first control variable will concern immigration levels. We can see how the specific issue of immigration is gaining a considerable deal of attention on the mass public, particularly on places that are more exposed to it. A study by Otto and Steinhardt (2014) on 103 districts of Hamburg finds, for example, a significant impact of immigration on the political success of extreme right-wing parties, paired with a decreased electoral support toward the political forces promoting liberal immigration and asylum policies; unfortunately, the unavailability of individual-level data prevented a real analysis on the impact of these factors. More details on the casual link between immigration and right positioning at municipality-level can be found in Barone, et al. (2016), whose findings highlight how Italian municipalities that experienced larger incoming migration flows have been more willing to vote, on national elections, for the political coalition supporting stances less favorable to the immigrants; yet, interestingly enough, immigration seems to have no effect on electoral outcome when the size of the city is big, suggesting once more a conspicuous deal of geographic heterogeneity within political polarization. A last piece of supporting evidence can be found on a study by Arnorsson and Zoega (2018), regarding the Brexit referendum; their results show how, on a regional level, districts with low per capita GDP, higher proportion of low-income, low-educated inhabitants and that have been affected more strongly by immigration are more likely to be against the European Union – and thus to cast their vote on the Leave option. Concerning our data, information concerning immigration will be extracted from the Istat database and will be measured in terms of the yearly change in the number of non-EU people that a registered within a specific region.

The second and third control variables concern instead the general economic conditions. We have seen in the past chapters how lower income levels and (perceived) inequality might result,

ceteris paribus, in increased support toward populist parties. For this reason, both the Gini

Referenties

GERELATEERDE DOCUMENTEN

In another article (Ito &amp; Lee, 2005b) the authors compared the impact of 9/11 and its after-effects on airline demand in Australia, Canada, Europe, Japan and the U.S. Using the

When a set of control variables are added (2), the significance for middle- income share becomes stronger (0.1%) and when control variables are added for industrial jobs (4),

The papers that went closer to the cited goal are series of studies conducted by Marques et al. In these studies, the authors tried to analyze the motivations driving energy

The link between regional competitiveness and the development of human capital is primarily a result of resources gained because of the region’s competitive position vis-à-vis

These plots have been constructed for the traditional multifactor productivity growth measure, the elasticity of scale, the elasticity of cost with respect to capital and the

The average deposit rate of the other banks operating in each location is calculated which then is used to create an instrument which exogenously influences the deposit rate for

Results in table 6 (model 5) imply that there is no difference between state owned firms and private owned firms on basis of export propensity when there are control variables

This significant government balance interaction variable shows that for the CEE10 a higher government balance does lead towards a higher economic growth rate, whereas the effect