The Increasing Appeal of Far-Right Parties

Plant disease epidemics are best understood as collective traumatic events that affect the economic sustenance and community life of places associated with the vulnerable crop (Erikson Reference Erikson1978; Pañeda-Fernández Reference Pañeda-Fernández2021). While outbreaks may generate less immediate physical devastation than other natural disasters, we argue that they create long-lasting grievances due to the combination of economic hardship and loss of community life (Bolet Reference Bolet2021; Cramer Reference Cramer2016; Hochschild Reference Hochschild2018). Beyond serving as an important means of subsistence, the typical crops grown in a geographic area play a pivotal role in structuring community life and giving residents a deep-rooted “sense of place” (Agnew Reference Agnew1987; Rowling Reference Rowling2022). The economic and sociocultural hardship associated with plant disease epidemics shapes how individuals relate to their surroundings and with each other (Bell Reference Bell1992; Cohen Reference Cohen1985; Macgregor Reference Macgregor2013; Wuthnow Reference Wuthnow2015).

Building on prior cultural sociology research (DiMaggio Reference DiMaggio1997; Small, Harding, and Lamont Reference Small, Harding and Lamont2010; Swidler Reference Swidler1986), we argue that plant disease epidemics have the potential to bolster far-right support in affected communities due to the socioeconomic and sociocultural grievances they generate (Ballard-Rosa et al. Reference Ballard-Rosa, Malik, Rickard and Scheve2021; Bolet Reference Bolet2021; Cramer Reference Cramer2016), but preexisting conditions determine the extent to which they do. Historically rooted community narratives—which we define as the stories people tell each other about the place they live in—give residents of affected communities a lens through which to interpret the outbreak and attach political meaning to their experience. These narratives serve as guiding frameworks that influence interpretations of new experiences and shape strategic responses to them (Harding Reference Harding2007; Rosen Reference Rosen2017; Small Reference Small2002). Building on this notion, we argue that support for far-right parties after a plant disease epidemic is higher in communities that already espoused a narrative of being left behind and abandoned by the state.

This narrative reflects a sense of feeling peripheral to an area, and such feelings of being left behind have been shown to matter for rural residents (Cramer Reference Cramer2016; Cramer Walsh Reference Cramer Walsh2012). We link the community narrative of being left behind to places that resent the state for historically poor public service provision (Cremaschi et al. Reference Cremaschi, Rettl, Cappelluti and De Vries2024; Nyholt Reference Nyholt2024; Rickardsson Reference Rickardsson2021; Stroppe Reference Stroppe2023). Access to services such as hospitals, schools, and transport links is one of the most direct ways in which the state and political elites affect citizens’ lives. It provides a tangible basis for evaluating how taxes are being spent, and how willing the state is to invest in certain communities (Dowding and John Reference Dowding and John2012; Hern Reference Hern2019). In public-service-deprived areas, residents share a community narrative of repeated exposure to “distributive injustices” by those in power (Ziblatt, Hilbig, and Bischof Reference Ziblatt, Hilbig and Bischof2024, 1483). Communities that are less connected to national decision-making centers begin to feel they have been left behind by the state and ruling political elites, and that they have not received their fair share of public spending and services (Cramer Reference Cramer2016; Hochschild Reference Hochschild2018; Patana Reference Patana2020; Ziblatt, Hilbig, and Bischof Reference Ziblatt, Hilbig and Bischof2024).

The community narrative of being left behind shapes residents’ perceptions of themselves as well as others, including politicians. It thus provides a powerful lens through which to interpret the outbreak and the associated hardship as further evidence that the state and ruling elites do not care about their communities. Building on prior work which has established that having limited access to public services erodes trust in state institutions and bolsters support for far-right parties (Cremaschi et al. Reference Cremaschi, Rettl, Cappelluti and De Vries2024; Nyholt Reference Nyholt2024; Rickardsson Reference Rickardsson2021; Stroppe Reference Stroppe2023), we argue that exposure to plant disease epidemics fuels support for the far right in areas with poor access to public services that embrace a narrative of being left behind by the state. In these areas, left-wing messaging about state redistribution is unlikely to gain support due to assumptions that universalist welfare spending would not benefit the local community (Cramer Reference Cramer2016; Cremaschi et al. Reference Cremaschi, Rettl, Cappelluti and De Vries2024).

Our argument stems from a combination of deductive expectations and inductive theory building. The literature on economic shocks and place-based resentment leads us to develop concrete expectations about how the economic and sociocultural hardship caused by the outbreak shapes its political effects, and how public service deprivation moderates it. Our qualitative fieldwork deepens our mechanistic understanding of the outbreak’s impact on the local economy and community life, and helps us identify residents’ motives, perceptions, and meanings that shape the electoral outcomes we observe in affected communities. The fieldwork reveals how a community narrative of being left behind by the state determines the extent to which exposure to the Xylella outbreak fuels far-right support. Our evidence comes from a high-income country in which the state traditionally delivers key services such as health care or public transport, and a multi-party political system in which populism is an established political force. While Italian politics has in many ways been at the forefront of the wave of far-right electoral success in the advanced industrial world, and is therefore an important case to study, future research should empirically test our argument’s scope conditions about the political consequences of plant disease epidemics.

The Xylella Outbreak

Puglia has a strong primary sector economy; olive tree cultivation and olive oil production constitute an essential component. In the marketing year 2013–14, Italy was the second-largest producer of olive oil in the EU, with 18.6% of the total EU production. Puglia was the country’s largest olive-producing region that year, accounting for 40.1% of the national production, followed by Calabria (20.2%), Sicilia (11.3%), and Campania (8.1%). As Figure 1b indicates, olive tree plantations are evenly spread across the region with a higher concentration in four of its six provinces—the northern Barletta, Andria, Trani, and Bari and the southern provinces of Brindisi and Lecce. Olive orchards constitute a prominent feature of Puglia’s landscape and cultural heritage, encompassing an estimated 60 million olive trees, including 5 million over 100 years old and 300,000 that are at least 1,000 years old (Conti Reference Conti2019). A 1950 land reform and subsequent division of inherited land into small parcels broadly distributed ownership of the orchards across the population. In the areas most affected by the Xylella outbreak, about 80% of the owners have a small plot of 50–70 trees, which are mainly used for family consumption.Footnote ²

Figure 1. The Xylella Outbreak

Note: Panel a shows 2022 provincial and municipal borders. Panels b, c, and d show 2022 municipal borders.

Xylella causes olive quick decline syndrome, which restricts the flow of sap within the tree (Schneider et al. Reference Schneider, van der Werf, Cendoya, Mourits, Navas-Cortés, Vicent and Lansink2020). Affected trees lose their leaves and stop producing olives; they usually die within 2–5 years (Saponari et al. Reference Saponari, Giampetruzzi, Loconsole, Boscia and Saldarelli2018). It came to Italy via a commercial plant shipment from Costa Rica to the port of Gallipoli and was spread by the spittlebug Philaenus spumarius (Bajocco, Raparelli, and Bregaglio Reference Bajocco, Raparelli and Bregaglio2023; Kottelenberg et al. Reference Kottelenberg, Hemerik, Saponari and van der Werf2021). Scientists have established a positive correlation between the prevalence of insects carrying the Xylella bacterium and increasing temperatures in the Mediterranean region (Farigoule et al. Reference Farigoule, Chartois, Mesmin, Lambert, Rossi, Rasplus and Cruaud2022). The projected milder winters in the region are expected to increase the risk of further Xylella outbreaks.

By the time the Xylella outbreak was discovered in Puglia in October 2013, 30–40 square miles of olive trees were already infected (Scortichini et al. Reference Scortichini, Loreti, Pucci, Scala, Tatulli, Verweire and Oehl2021). The known infected area almost tripled during the following year, extending to the entire province of Lecce (Kottelenberg et al. Reference Kottelenberg, Hemerik, Saponari and van der Werf2021). By 2018, the successful implementation of containment measures considerably slowed the spread of the bacterium and protected the region’s northern areas from infection. By that time, over half (about 6.5 million of 11 million) of the trees in the area had been infected (Bajocco, Raparelli, and Bregaglio Reference Bajocco, Raparelli and Bregaglio2023; Conti Reference Conti2019), resulting in yield losses equivalent to 10% of the national production (Kottelenberg et al. Reference Kottelenberg, Hemerik, Saponari and van der Werf2021). Figures 1c and 1d visualize the spread of the disease according to official data released from April 2014.

Ongoing containment measures include the eradication and destruction of suspected infected plants, constant monitoring in surrounding areas, a ban on transporting infected organic material, and a mandate for landowners to keep land clean (e.g., cutting weeds) in areas considered at risk. Since 2015, the national government has also provided economic support to affected farmers and encouraged them to replace infected trees with species resistant to the bacterium. As of 2022, EUR 243 million had been allocated to manage the emergency (Camera dei Deputati 2022). However, slow and intricate political decision-making and bureaucratic implementation processes have delayed the adoption of containment and replanting measures.

The Political Environment

During the Xylella outbreak, the Italian political landscape featured a diverse array of parties from the entire political spectrum. This increases our confidence that enhanced far-right support does not simply reflect the absence of far-left supply. Parties on the far left include those in the tradition of the Italian Communist Party, such as Left Ecology Freedom (Sinistra Ecologia e Libertà) and, since 2013, the Five Star Movement (Movimento 5 Stelle).Footnote ³ Parties on the far right include the originally regionalist League (Lega) party, which had become a “national” far-right party by 2013, and the nationalist Brothers of Italy (Fratelli d’Italia), established in 2012 following in the footsteps of the Fascist National Party’s republican heir, the Italian Social Movement (Movimento Sociale Italiano).

While right-leaning parties have traditionally enjoyed significant support in Puglia, the left has made inroads in the region in recent years. In 2010, a broad center-left coalition led by far-left leader Nichi Vendola won regional elections. Since then, Puglia has been governed by a left-wing coalition led by Democratic Party (Partito Democratico) leader Michele Emiliano after 2015. The Five Star Movement received the highest vote share in the 2018 (44.94%) and 2022 (28.28%) national elections, followed closely in the latter by the far-right Brothers of Italy (23.5%).

As the leading force in the regional government throughout the outbreak, the Democratic Party assumed political responsibility for managing the disaster and distributing relief.Footnote ⁴ Prior work on incumbent approval in the aftermath of natural disasters (Healy and Malhotra Reference Healy and Malhotra2013) would lead us to expect the public authorities’ initial inability to contain the spread of the bacterium and the sluggish implementation of the replanting and compensation programs to generate an electoral penalty for the Democratic Party. During the outbreak, all major parties shifted positions (ranging from advocating to opposing eradication) or supported contrasting positions at the regional versus national levels (see Section A.2 of the Supplementary Material). These shifting stances partially reflect difficulties in understanding the spread of the outbreak and contributed to delays in containment—which could have led to voter disengagement and reduced turnout in affected areas. We address these additional expectations in our analysis.

Quantitative Analysis

Our identification strategy exploits the geographic variation and timing of the Xylella outbreak. We rely on official data provided by the regional government’s Geographic Information Service (SIT Puglia 2022), which tracked the time of infection across municipalities starting 6 months after the bacterium was first isolated.Footnote ⁵ As Figures 1c and 1d illustrate, while some (139 out of 257) municipalities in Puglia were declared infected between April 2014 and May 2019, others (118) were not touched by the bacterium by the end of 2022. This setting allows us to employ a DID strategy to compare the pre- and post-infection change in the outcomes between infected and uninfected municipalities, and thus estimate the average treatment effect on treated (ATT) units.

Our main model examines how the outbreak affected the share of valid votes for different political blocs and turnout in national elections for the lower chamber (Camera dei Deputati) based on official electoral data (Ministero dell’Interno Reference Ministero2024). We exploit two aspects of the timing of the outbreak and elections to estimate our two-way fixed-effect (TWFE) model: (1) Xylella was first detected in October 2013, almost a year after the February 2013 parliamentary election and (2) the bacterium had reached all but six (133) of the eventually infected municipalities by April 2016, almost 2 years before the next national election was held in March 2018. We use the following equation:Footnote ⁶

⁽¹⁾ $$ \begin{array}{rl}{Y}_{i,t}=\beta {X}_{i,t}+{\mu}_i+{\lambda}_t+{L}_{p(i)}(t)+{\varepsilon}_{i,t},& \end{array} $$

where

• • $ {Y}_{i,t} $ measures the electoral outcome for municipality $ i\in \mathcal{I} $ in election year $ t\in \mathcal{T}:=\{2001,2006,2008,2013,2018,2022\} $ ;

• • $ {X}_{i,t}:={\unicode{x1D7D9}}_{{\mathcal{I}}_{inf}\times {\mathcal{T}}_{post}}(i,t) $ ; (i) $ {\mathcal{I}}_{inf}\subset \mathcal{I} $ denotes the subset of infected municipalities and (ii) $ {\mathcal{T}}_{post}:=\{2018,2022\} $ , that is, $ {\mathcal{T}}_{post} $ the subset of post-infection election years;

• • $ {\mu}_i $ and $ {\lambda}_t $ are municipality- and year-specific intercepts;

• • $ (p,t)\mapsto {L}_p(t) $ is a province-level linear time trend,Footnote ⁷ while $ i\mapsto p(i) $ maps municipality i onto the group of municipalities defining the province to which i belongs; and

• • $ {\varepsilon}_{i,t} $ is an error term, which we allow to be correlated within individual municipalities across years.

To avoid a staggered DID setting and the well-documented theoretical difficulties this would pose for TWFE estimation (see, e.g., De Chaisemartin and d’Haultfoeuille Reference De Chaisemartin and d’Haultfoeuille2020), we exclude the six municipalities infected after 2018 from the sample used to estimate Equation 1. This allows us to interpret the coefficient $ \beta $ as an estimate of the ATT while retaining the flexibility of the TWFE specification, which lends itself to natural extensions for HTE analysis and further robustness checks.

We estimate four additional models that we describe in detail in Section B.1 of the Supplementary Material. First, we estimate TWFE event-study equations (Equation B.1.1 in the Supplementary Material) to test the pre-treatment parallel-trends assumption underlying our DID design. Second, we estimate staggered DID equations, following the methodology proposed by Callaway and Sant’Anna (Reference Callaway and Sant’Anna2021),Footnote ⁸ to examine the outbreak’s economic and sociocultural consequences using yearly measurements of municipal-level indicators. Third, we test our hypothesis about the moderating effect of public services and validate our treatment measure by estimating interaction specifications (Equation B.1.2 in the Supplementary Material) that generalize the binning estimator suggested by Hainmueller, Mummolo, and Xu (Reference Hainmueller, Mummolo and Xu2019). Finally, we run a set of additional empirical tests that we describe in the “Robustness Checks” section.

Qualitative Analysis

We conduct an in-depth qualitative study of a “positive case” (a municipality where the DID analysis reveals that the Xylella outbreak increased far-right support) and a “negative case” (a comparable municipality where we detected no electoral consequences). Our approach differs from common mixed-methods strategies in two respects. First, it aims to identify the most suitable cases for ethnographic fieldwork to reveal the mechanisms underlying the causal effect identified in the quantitative analysis. This objective differs from enhancing model specification (Lieberman Reference Lieberman2005) or identifying intermediate variables and refining estimates of causal effects (Humphreys and Jacobs Reference Humphreys and Jacobs2023; Weller and Barnes Reference Weller and Barnes2014). It allows us to be sensitive to mechanistic understandings of causality by uncovering the motives, perceptions, and meanings behind the political choices of those affected by the outbreak. Second, our case selection procedure incorporates the key features of the DID’s dynamic potential outcome causal framework, unlike static frameworks based on linear regression or matching (Galvin and Seawright Reference Galvin and Seawright2021; Seawright Reference Seawright2016b).

To select our cases, we estimate the TWFE event-study model (Equation B.1.1 in the Supplementary Material) 133 times, each time including one treated municipality and all control municipalities. This estimation helps us identify affected municipalities that exhibit similar pre-treatment trends in far-right vote share to those in the control group (i.e., those with pre-treatment event-study coefficients not significantly different from zero at a 0.05 p-value threshold). Since their behavior closely resembles the natural baseline represented by the control group, we can assume that municipalities in this set would have continued behaving similarly in the absence of Xylella. From this set, we select a municipality that had a high estimated treatment effect in the individual model as the positive case, and one that had a zero estimated treatment effect as the negative case. We ensure that these cases are comparable across multiple baseline dimensions, including population, timing of infection, olive tree prevalence, and geographic location, all measured before the Xylella outbreak.Footnote ⁹ We deliberately remain blind to any measure of public service provision to further probe the relevance of this moderator.Footnote ¹⁰

Assessing if (and how) public service deprivation shapes the electoral consequences of the outbreak is a key objective of the design. However, we select our cases based on the estimated treatment effect rather than the values of public service deprivation (i.e., distance from public service hubs) we use in our quantitative analysis to explore HTEs for two reasons. First, while the measure of distance to public service hubs is a solid proxy for large-N cross-case analyses, it may not fully capture public service deprivation at the case level. For instance, prior work suggests that public service deprivation may also include limited access to services provided within the municipality (Cremaschi et al. Reference Cremaschi, Rettl, Cappelluti and De Vries2024).Footnote ¹¹ The second reason is that the extent to which public service deprivation translates into a narrative of having been left behind by the state in a specific case depends on historical and contextual factors that the measure of distance to public service hubs may not fully capture. We choose cases based on the estimated treatment effect to overcome some limitations of the quantitative measure of the moderator and explore the mechanism that we theorize, which highlights the importance of a preexisting community narrative of feeling left behind by the state rooted in previous experience of public service deprivation.

In April 2023, we conducted 3 weeks of fieldwork in the two selected municipalities (which we refer to as Querceto and Prosopo to protect the identities of our informants and town residents). Our primary data collection method was interviewing, supplemented by direct observation and examining local documents such as magazines and history books. We stayed in the positive case of Querceto and visited the negative case of Prosopo daily. We interviewed 30 people, including local and national politicians, farmers, entrepreneurs in the food service and agricultural products sectors, casual olive growers, as well as other community leaders and key informants (e.g., the town’s priest, the librarian, the local historian).Footnote ¹² We selected our interviewees by contacting the local administration, local party chapters, and farmers’ associations and asking for referrals to other relevant individuals.Footnote ¹³ Our semi-structured interviews focused on participants’ perspectives on how the outbreak affected their communities, the changes it brought about, and the role of national and local political elites.Footnote ¹⁴ To avoid priming participants, we did not include questions about public services in our interview guide, allowing the relevance of this moderator to be uncovered in the conversation.

Electoral Outcomes

Our statistical analysis consistently indicates that municipalities affected by the Xylella epidemic cast more votes for far-right parties after the outbreak. Table 1 summarizes the main results on the political consequences of the outbreak obtained through our DID electoral analysis. We present ATT estimates based on Equation 1 for four key political blocs and voter turnout.Footnote ¹⁵ The outbreak exhibits a significant and sizeable average effect on the percentage of valid votes cast for the far right of about 2.2 percentage points (column 4). The rest of the table addresses additional expectations.

Table 1. TWFE DID Estimates of the ATT on National Electoral Outcomes

Note: Outcomes measured in percentage-point shares. Estimates are based on 251 municipalities and 6 election years (N = 1,506). Standard errors are clustered at the municipality level. *** $ p<0.001 $ ; ** $ p<0.01 $ ; * $ p<0.05 $ ; $ {}^{+}p<0.1 $ .

There are two main take-aways from Table 1. First, in line with the notion that political incumbents might be punished for poorly managing disasters (Healy and Malhotra Reference Healy and Malhotra2013; see also Margalit Reference Margalit2019b), we would expect to find less support for the center-left Democratic Party. While our results indicate vote losses for this party, they are only significant at the 0.10 level and are comparable in magnitude to the losses experienced by other political blocs. We also explore Xylella’s impact on vote shares in regional elections.Footnote ¹⁶ The results presented in Figure B.5.1 in the Supplementary Material do not provide evidence that the epidemic affected the vote share of regional incumbent or far-right parties in regional elections. In Section B.5 of the Supplementary Material, we explain this result by stressing the second-order nature of regional elections (Reif and Schmitt Reference Reif and Schmitt1980) and voters’ inability to accurately attribute responsibility for various policies to different territorial levels of government (Arceneaux Reference Arceneaux2005; de Benedictis-Kessner Reference de Benedictis-Kessner2018; Wilson and Hobolt Reference Wilson and Hobolt2015).

Second, all parties flip-flopped on how to deal with the outbreak, which could have produced political disengagement, manifested in either reduced turnout or a protest vote for an anti-establishment challenger party (De Vries and Hobolt Reference De Vries and Hobolt2020). However, column 5 of Table 1 indicates that the outbreak increased turnout by an average of 3.8 percentage points.Footnote ¹⁷ While mainstream parties from both the center left and center right experienced electoral losses (−2.4 and−1.4 percentage points, respectively, significant at the 0.10 threshold), the far-left bloc (including far-left parties and the anti-establishment Five Star Movement) lost 1.9 percentage points (significant at the 0.05 threshold).Footnote ¹⁸ These results suggest that the increased far-right support we detect cannot be explained as simply the expression of a generic protest vote. Overall, our findings thus far support our argument that exposure to a plant disease epidemic boosted electoral support for far-right parties in Puglia.

Inspecting Parallel Trends

We employ a TWFE event-study specification to verify the parallel-trends assumption (Equation B.1.1 in the Supplementary Material). Figure 2 displays the results using two different measures of far-right vote share: (1) the share over the total number of valid votes (Figure 2a), which corresponds to the measure used in Table 1 and (2) the share over the number of votes cast for the right-wing bloc (Figure 2b). The figures, which plot the ATT on the share of votes in each year (taking the 2013 election as the baseline), confirm that treated and control municipalities had very similar voting behaviors before the outbreak. Figures B.2.1a, B.2.1b, and B.2.1c in the Supplementary Material also confirm the presence of parallel trends before the outbreak in turnout, vote for the left, and vote for the center right, respectively.

Figure 2. TWFE Event Study of Far-Right Vote Share, 2001–22

The event study also helps us explore the dynamic effect of the Xylella outbreak on far-right support. To ease interpretation, we complement classic event-study plots with a graph of the mean outcome value for the control group over time. The combined plot in Figure 2a establishes that while all municipalities in Puglia experienced an increasing time trend in vote shares for the far right starting in 2013, those affected by Xylella were disproportionately more likely to vote for these parties in 2022. While considering the share of valid votes is important for calibrating the outbreak’s electoral consequences, this outcome measure might obfuscate electoral movements that were already occurring within the right-wing electoral bloc at the start of the outbreak. Figure 2b indicates that the far-right vote share had already increased in the 2018 election at the expense of center-right parties. These results suggest that the Xylella outbreak first radicalized right-wing voters, and then generated a more general increase in far-right support.

Robustness Checks

In the Supplementary Material, we confirm the robustness of our key findings in eight steps. First, while not strictly necessary for the validity of our DID identification strategy, we show that treated and control municipalities are balanced on key covariates (Table B.19.1 in the Supplementary Material). Second, we increase our confidence that olive tree damage is driving the electoral effects we observe by demonstrating that the treatment effect increases with olive tree prevalence in two fully saturated interaction models (Figures B.4.1 and B.4.2 in the Supplementary Material). Third, we obtain consistent results when we replace the binary treatment indicator with a continuous variable measuring the length of exposure to Xylella in each election (Table B.4.1 in the Supplementary Material).

Fourth, we rule out the possibility that compositional effects are driving our results by simulating a very conservative counterfactual scenario in which young, potentially progressive citizens have not emigrated from affected municipalities (see Figure 3b), thus providing a lower bound to our ATT estimates (Section B.8 of the Supplementary Material). Fifth, we dismiss the possibility that our results are driven by Puglia-specific electoral dynamics, spatial spillovers, or olive market general equilibrium effects within the region by using municipalities in Calabria (another southern region that exhibited parallel trends to our treated municipalities) as an alternative control group (Figure B.9.1 in the Supplementary Material).

Figure 3. Staggered Event-Study Plots for Two Indicators of Economic and Sociocultural Hardship

Sixth, we conduct randomization inference tests, taking into account the randomness generated by treatment assignment rather than the sampling of municipalities in our study, by randomly simulating the treatment assignment (Figure B.16.1 in the Supplementary Material) or using an epidemic-like procedure (Figure B.16.2 in the Supplementary Material). Seventh, we account for potential spatial correlation in the error term by demonstrating that our estimates remain statistically significant even after clustering standard errors at the level of wide, spatially defined, clusters (Figure B.17.1 in the Supplementary Material). Finally, we perform appropriate checks on the statistical power ensured by our sample and on Type S(ign) as well as Type M(agnitude) errors in ATT estimation, and find strong evidence of the statistical soundness of our study (Table B.18.1 in the Supplementary Material).

Economic and Sociocultural Outcomes

The economic and sociocultural hardship that affected communities experienced as a result of Xylella is an important part of our explanation of why the outbreak increased far-right support. We quantitatively examine this intuition by replicating the DID analysis on two municipal-level indicators: (1) per capita pre-tax income (MEF 2022) and (2) youth emigration (measured as the number of Italian residents aged 20–35 using data by ISTAT 2022). Figure 3 plots estimates from staggered DID event-study models of these two yearly measures. Each graph depicts the annual ATT, taking the year before the municipality was infected as the reference.

Figure 3a documents the outbreak’s negative effects on pretax per capita income. It indicates a sharp and persistent economic deterioration following infection, with an average decrease of about 1% in the third year of the infection and beyond compared to baseline. The estimate captures the aggregate general equilibrium average impact on affected municipalities, and therefore also uncovers second-order effects on the local economy in addition to income losses incurred by those directly employed in the olive farming and processing sector. While these estimates confirm that the epidemic generated negative economic effects, the magnitude of these effects is moderate compared to the estimated electoral effects. The moderate effect on income is likely explained by the slow (but overall effective) implementation of compensation policies for large olive entrepreneurs, the redirection of labor to other sectors, and the relatively small income losses experienced by the large number of smallholders.

Figure 3b demonstrates that the population of young adults (aged 20–35) rapidly decreased in the years after the outbreak, with peak reductions of around 3% from baseline after a few years. This suggests that, faced with declining prospects, young people increasingly moved away from affected municipalities. Figure B.7.1 in the Supplementary Material reports the results of a staggered DID event study of an additional indicator of sociocultural hardship: the model provides suggestive evidence that the beginning of the outbreak coincided with an increase in the number of suicides. The results point to substantial social and psychological distress generated by the outbreak (cf. Case and Deaton Reference Case and Deaton2020). Overall, these results provide support, sustained by our in-depth case analysis, that the effects of the Xylella outbreak extend beyond the economic realm into the sociocultural sphere.

The Moderating Effect of Public Service Deprivation

Building on insights from geography (Barca Reference Barca2009; Barca, McCann, and Rodriguez-Pose Reference Barca, McCann and Rodriguez-Pose2012; Rodriguez-Pose Reference Rodriguez-Pose2018), we use a municipal-level measure of driving distance (in minutes) to public service hubs—municipalities or clusters of neighboring municipalities featuring (i) a nationally connected train station, (ii) a hospital offering services beyond an emergency room, and (iii) a high school—to proxy for access to public services before the Xylella outbreak.Footnote ¹⁹ Figure B.10.1 in the Supplementary Material depicts the classification of Italian municipalities based on our public service deprivation measure. It highlights variation in deprivation across municipalities in Puglia that is substantially orthogonal to the spread of Xylella.Footnote ²⁰

We estimate a HTE specification as in Equation B.1.2 in the Supplementary Material, using our indicator of public service provision as the moderator. Figure 4 plots the outbreak’s conditional average treatment effect (CATT) on far-right vote shares across different levels of access to local public services. It reveals substantial variation in the effects of the shock. The outbreak’s effect is not significantly different from zero in municipalities in the first tercile of our public service measure (shortest distance to local public services); it becomes statistically significant and sizeable in the second tercile (CATT = 1.8) and more than doubles in size in the third tercile (CATT = 4.3), when the distance to public services is greatest.

Figure 4. Xylella’s Conditional Effect on Far-Right Support, across Terciles of Distance from Public Service Hubs

We ran four robustness tests to check the relevance of the CATT. First, we use a residualized version of the baseline moderator (Figure B.11.1 in the Supplementary Material). The residuals come from regressing the distance from public services on a set of preinfection covariates which capture community dimensions that may be correlated with the moderator, such as income, rurality, tourism relevance, education, age composition, and social capital. The results remain substantively unchanged, which mitigates concerns that the moderating effect of public service provision could be entirely driven by its correlation with other potentially relevant dimensions of treatment effect heterogeneity. Second, in Section B.12 of the Supplementary Material, we estimate high-dimensional HTEs using new machine-learning methods (Wager and Athey Reference Wager and Athey2018). Again, we find that distance from public services has a moderating effect as well as a high degree of relevance in explaining treatment effect heterogeneity. Third, the results from an HTE model on average municipal income (Section B.13 of the Supplementary Material) rule out the possibility that greater access to local public services suppresses an increase in far-right support by providing an economic buffer. Fourth, in Section B.14 of the Supplementary Material, we replicate the HTE results using an alternative measure of public service deprivation. Overall, these results suggest that public service provision is a key dimension of heterogeneity in the effects of the outbreak. In the next section, we explore how this happens—and how a preexisting community narrative of being left behind by the state matters.

Exploring the Consequences of the Outbreak

Residents in affected areas described how the epidemic profoundly transformed their living environment. We conducted our fieldwork in the spring, when the landscape would have historically been verdant and alive; it is now a forest of withered trees interspersed with brambles and overgrown with weeds. A Querceto resident expressed how dead trees looked “like skeletons” and instilled “sadness” and “fear.”Footnote ²⁵ A Prosopo inhabitant explained that he now constantly felt like “walking in an open-air cemetery.”Footnote ²⁶

Other prevailing sentiments were nostalgia and affection. Residents of both towns recalled how their older acquaintances and relatives would tear up when discussing or seeing withered olive trees.Footnote ²⁷ A Querceto politician noted that for many, olive trees “were not just plants.”Footnote ²⁸ Having planted them many decades ago, elders were attached to the olive trees as if they were their children.Footnote ²⁹ Prosopo’s mayor shared that some even gave the trees human names.Footnote ³⁰ We encountered expressions of affection and nostalgia in three poems published in Querceto’s local newspaper, both before and after the outbreak.Footnote ³¹

Since many residents owned some land with trees (one Querceto resident jokingly referred to himself as “the sole person in town without any olive trees”),Footnote ³² community members perceived the outbreak as a loss that impacted everyone.Footnote ³³ In both towns, the outbreak disrupted established routines. In spring, many residents would traditionally have been busy preparing the land to plant new olive trees—now considered a custom of the past.Footnote ³⁴ An agricultural tools salesman, whose office was covered with photos of previous harvests, noted how traditional marriages were often based on families’ olive assets, as these formed part of the dowry.Footnote ³⁵

For many, one of the most cherished lost customs was the enjoyment of their own olive oil.Footnote ³⁶ The widespread tradition of producing oil for one’s family instilled a sense of pride and community beyond sheer economic value. According to Prosopo’s librarian, an essential aspect of Salento’s identity was the possession of at least 100 liters of self-produced oil for the year.Footnote ³⁷ A popular saying attests to past oil abundance: “olive oil could be used for washing one’s feet.”Footnote ³⁸ A Prosopo resident described how his mother would have felt dishonored to discover that he and the other town residents were buying oil from northern Puglia.Footnote ³⁹ A Querceto resident described witnessing their mothers “saving olive oil and frying with [cheaper] seed oil” as “a stab in the heart.”Footnote ⁴⁰

The epidemic also deeply impacted community life. Prosopo residents shared stories of the town emptying during the fall in the past, when families relocated to countryside homes to harvest olives.Footnote ⁴¹ Each family would take their harvested olives to the mill to produce their oil. Querceto residents recounted queuing at the olive mill as a lost moment of communal exchange.Footnote ⁴² According to a Querceto politician, the disappearance of these seasonal activities left a void in town meeting places.Footnote ⁴³ He also mentioned that the loss of vegetation had led to the closure of social spaces linked to other countryside activities like the hunting club, and explained that depopulation had led to the shutdown of clubs such as those of A.C. Milan and Juventus F.C. supporters.Footnote ⁴⁴ The loss of landscape and cultural and social heritage led the mayors of Querceto and Prosopo to state that the damage to the olive trees constituted the loss of a “symbol of the area.”Footnote ⁴⁵

Examining the Preexisting Community Narrative

While Querceto and Prosopo residents shared similar accounts of the economic and sociocultural hardships caused by the outbreak, their descriptions of local access to public services were starkly different. Shortly after our arrival, we were invited to a public assembly in Querceto, where the mayor inaugurated a modest new auditorium. Following a blessing by the town’s priest, the mayor outlined the achievements of his administration. During the question-and-answer session, an elderly woman took the floor to voice her grievances: “Excuse me, Mr. Mayor, you’ve listed a lot of money spent by this administration, but you haven’t mentioned the young people leaving the town or the many elderly people who remain. With all this money you’ve spent, how is it possible that we still don’t have a functioning elderly community center?”Footnote ⁴⁶ Some days later, an interviewee cited the woman’s speech as an example of how everyone noticed the neglect of the town.Footnote ⁴⁷ He emphasized that national politics had historically neglected Querceto.Footnote ⁴⁸ In interviews and casual conversations, other residents also related that they viewed Querceto as a neglected place.Footnote ⁴⁹

Local representatives also voiced discontent regarding access to public services in Querceto. A municipal councilor affiliated with the far-right Brothers of Italy party stated: “Opportunities here are not the same as in northern Italy. We suffer a gap in services and infrastructure.”Footnote ⁵⁰ He emphasized the malfunctioning of the local health and waste collection systems, describing them as “falling apart.”Footnote ⁵¹ The mayor, affiliated with the far-right League,Footnote ⁵² called Querceto a “godforsaken town,” citing the absence of a police station, hospital, and garbage collection center as well as a lack of government funds to restore the town’s archaeological site.Footnote ⁵³ He also complained about the absence of a reliable transport system: “we pay taxes like people in Milano, but we need to drive to Lecce to use the train, and it takes us half an hour to cover just 30 km for how roads are badly maintained. Most trains don’t even reach Lecce but only arrive in Bari [about 200 km away].”Footnote ⁵⁴

The mayor said his primary motivation for becoming politically engaged was to address Querceto’s enduring neglect. “I believe that we [Querceto’s politicians] must steal from politics what can benefit our community;”Footnote ⁵⁵ the robbery metaphor communicated the need to use every means available to secure the necessary state resources. He traced Querceto’s history of public service deprivation back nearly 80 years when, in the aftermath of World War II, a senator from Querceto was expelled from his party and lost his seat. The senator from the neighboring town exploited the opportunity to centralize public services in his own town, leaving Querceto behind. This long-standing narrative was consistently recounted by local politicians from different political parties and a historian; it also appeared in the local newspaper and a historical book.Footnote ⁵⁶ Many shared the mayor’s perspective; some referred to it as one of the “many trains that Querceto has missed.”Footnote ⁵⁷

Conversations and interviews in Prosopo rarely (if ever) mentioned a lack of public services or feelings of neglect. In fact, some elements of Prosopo residents’ community narrative were the opposite of those used by Querceto residents. This was somewhat surprising to us because while our theory predicts that a negative case like Prosopo would not feature a narrative of being left behind by the state, it would not necessarily expect the opposite narrative of positive experiences with public service provision. A local historian suggested that the presence of a railway since the nineteenth century had played a pivotal role in establishing a religious organization in Prosopo, which had bolstered the town’s development.Footnote ⁵⁸ The story of the monks’ arrival “by the railway” even featured in the title of a book sold at the town newsstand.Footnote ⁵⁹ Several participants recalled how a sentimental relationship between the mayor at the time and a prominent provincial politician had brought significant public resources to the town in the 1980s.Footnote ⁶⁰

Interviewees offered these accounts without prompting, which increases our confidence that in Prosopo a community narrative of feeling left behind by the state is not prominent. The divergent perceptions of citizens’ access to services in the two municipalities is also reflected by differences that we observed during our stay. In Prosopo, the town librarian received us in a castle owned by the municipal administration that had been carefully renovated to host public events. In Querceto, the library was closed throughout our stay, and we never received a response when we reached out using the contact information provided on the door. Querceto’s municipal administration held its public meetings in a modest auditorium that was used during the day as an elderly center.

The starkly contrasting accounts we uncovered in the two municipalities suggest that previous experiences of public service deprivation in Querceto have become ingrained in a community narrative of being left behind, and that this was not the case in Prosopo. This narrative serves as a framework for interpreting the outbreak as another example of being abandoned by the ruling elites.

Understanding the Community Narrative as an Interpretative Lens

The two towns’ distinct preexisting community narratives align with our argument that such accounts shape how residents interpret the outbreak. Indeed, we found that the towns differ significantly in how they interpret the outbreak. Section C.7 of the Supplementary Material provides additional qualitative evidence that residents of both Querceto and Prosopo were similarly dissatisfied with how authorities managed the epidemic. Here, we illustrate that many Querceto residents believe the ruling elites abandoned them during the outbreak due to a lack of interest in the affected areas. In our interviews and conversations, without any prompting, many attributed the sluggish containment measures to this geographically targeted neglect.Footnote ⁶¹ One participant expressed: “They really do not care about Salento, an insignificant peninsula for the nation.”Footnote ⁶² Residents stressed how a place’s political relevance shapes the state’s response: “Things would have been different if it had started in Bari [the capital of Puglia].”Footnote ⁶³ Another remarked: “Try driving up to Bari. You will notice they already have their [Xylella-resistant] plantations ready, and we do not. Have you ever seen anyone coming down here to check how things are?”Footnote ⁶⁴ Querceto’s mayor also highlighted this point: “Had it happened in another part of Italy, the attention would have been greater, but no one ever cared about this small land.”Footnote ⁶⁵ The widespread perception that the national government had allowed other olive oil-producing regions such as Toscana and Liguria to exploit the crisis in Salento helps explain why the outbreak generated electoral effects at the national, but not regional, level.Footnote ⁶⁶

While many Querceto residents attributed the disaster to neglect by the ruling elites, others interpreted it as a conspiracy against their community.Footnote ⁶⁷ The town priest noted: “Some people have even experienced the outbreak as if there was someone who wanted this and made it happen. With which objective? It is not clear. But people perceive it like this.”Footnote ⁶⁸ A municipal councilor mentioned: “The fact that the epidemic stopped before Bari gave rise to suspicions. Of course, I don’t believe in these suspicions.”Footnote ⁶⁹ The mayor of Querceto said: “Some say it was political because multinational companies had interests in replanting trees, but I am not the one who can tell.”Footnote ⁷⁰ In casual conversations, several people embraced elements of such conspiracy theories. For example, some said that planes had distributed the bacterium because the infection had spread along straight lines.Footnote ⁷¹ They blamed people from Bari, those from other oil-producing regions, or global elites seeking to profit from international trade.Footnote ⁷²

In Prosopo, where residents and authorities did not embrace a narrative of state neglect, the outbreak was not interpreted as a geographically targeted abandonment by elites or a conspiracy against their community. While some participants mentioned that conspiracy theories had circulated at the beginning of the outbreak, they quickly died down and were perceived as irrelevant in the longer term.Footnote ⁷³ As we show in Section C.7 of the Supplementary Material, citizens of both Prosopo and Querceto often blamed the hardship associated with the outbreak on politicians’ inability to adequately deal with it.Footnote ⁷⁴ However, residents of Prosopo did not maintain that this failure was because Salento was affected instead of a more politically important territory. When we tried to probe potential feelings of belonging to the periphery or of geographically concentrated abandonment by political elites, which we detected in Querceto, by asking if the outcome would have been different if the outbreak had occurred in another part of Italy, one Prosopo resident replied: “The problem was that it happened here in Salento. Because Salento is full of olive trees. In another place, it would have caused less damage.”Footnote ⁷⁵ Others saw the outbreak as an unfortunate event for which no one could ultimately be blamed; as one interviewee explained, “They [politicians] just could not anticipate this.”Footnote ⁷⁶

These differing interpretations of the outbreak in Querceto versus Prosopo suggest that residents relied on different community narratives to make sense of their experiences. In Querceto, residents interpreted the epidemic through the lens of being left behind by the state. In Prosopo, which did not feature a “left behind” community narrative, the outbreak was perceived as an unpreventable disaster that had to be managed, but not as a sign of targeted neglect of the community by the ruling political elite.

Querceto residents’ interpretation of the outbreak as a collective abandonment by political elites allowed far-right rhetoric to resonate more strongly within the community. We found little evidence that mobilizing grievances related to disaster management drove the electoral success of far-right parties. At the end of our interviews and conversations, we shared our quantitative results with local politicians from different parties; they unanimously responded that this evidence could not be caused by the Xylella outbreak.Footnote ⁷⁷ While the outbreak was salient in everyday life, local politicians did not consider it a relevant campaign issue.Footnote ⁷⁸ When we asked interviewees if any political party had behaved worse or better than others during the crisis, the most common reaction in both towns was that “they [politicians] all had done badly.”Footnote ⁷⁹ Overall, our evidence suggests that the outbreak increased the appeal of far-right politicians’ national messages highlighting place-based economic and social grievances with a promise to restore the status of peripheral areas.