research

My research agenda develops clean, testable theoretical frameworks—centered on incentives, information frictions, and selection—to clarify mechanisms before tracing their empirical implications. Empirically, I work with reproducible, end-to-end pipelines that often combine web scraping and NLP with geocoding and panel construction. My workflow typically relies on Python, Stata, and spatial tools such as QGIS (and, when appropriate, R).

Crime Perception and Voting Behavior: Evidence from Individual Data
joint with Giovanni Prarolo
Abstract
Keywords
JEL
Paper
This paper studies how exposure to geolocated crime-related news shapes individual voting behavior in Italian elections. Using a panel of non-relocating voters observed across multiple election rounds, we exploit within-individual variation in exposure to nearby crime news during the pre-election month, controlling for individual and district-by-election fixed effects. Aggregate exposure yields weak, unstable effects. Disaggregating by offender nationality reveals systematic patterns: immigrant-attributed crime news reduces support for parties with ambiguous immigration stances (e.g., Five Star Movement) and increases support for clear "law-and-order" parties, while Italian-attributed crime has negligible effects. Effects are stronger among high-skilled voters shifting from M5S and low-skilled voters abandoning Lega. In local elections, Italian crime punishes incumbents, while immigrant crime increases abstention. These asymmetric responses—absent in aggregate measures—indicate that crime salience operates primarily through identity-based framing rather than through generalized concerns about crime or security. The findings highlight how media attribution shapes electoral accountability.
crime; immigration; elections; news media; individual voting behavior.
D72; D83; K42; L82.
The Tipping Point of Temptation: Selection into the Public Sector
Abstract
Keywords
JEL
Paper
Does corruption attract low-integrity types into the public sector, or deter honest ones? Existing evidence is mixed: high-corruption settings show a negative association between honesty and public-sector preference, while low-corruption settings show a positive one. This paper proposes a behavioral mechanism that organizes both patterns. Building on self-control utility, we treat corruption rents as a short-run temptation. The rent attracts low-motivation types; but it also imposes a self-control cost on an intermediate range of honest-enough agents who resist temptation at a cost, generating a “missing mass” of intermediate-integrity types who select out. These opposing forces produce a tipping point: corruption rents expand the public sector when temptation is mild but contract it—improving average integrity—when temptation is strong. We show this tipping logic is not an artifact of the leading benchmark, where the threshold takes the closed form λ* = 1 + √2: it is the universal small-rent limit for a broad class of type distributions, and the sign of the selection response is governed by the joint distribution of honesty and ability, which the theory treats as a primitive of the institutional environment. A model-guided exercise using five lab and field experimental samples spanning the corruption distribution shows that cross-sample variation in this joint distribution is consistent with the mechanism. The selection frontier also carries a falsifiable signature that distinguishes self-control from reputation and Roy sorting, and we characterize the data a structural test of it would require—realized entry, incentivized integrity, and ability jointly observed—which existing samples do not yet provide.
occupational selection; self-control and temptation; corruption; reputation and self-image; public sector labor markets; institutional design.
D73; J45; D90; H83.
Parenthood, Age, and the Opportunity Cost of Voting: Evidence from Administrative Voter Records
Abstract
Keywords
JEL
Paper
This paper studies how parenthood and parental age are associated with voter turnout using a comprehensive administrative panel covering the universe of registered voters in Bologna across four municipal and national elections between 2004 and 2013. By linking individual turnout records to demographic, fiscal, and residential information, we identify parents, track the age of their children, and follow the same individuals over time. We estimate linear probability models with individual and election-year fixed effects, exploiting within-individual variation to account for permanent differences in civic engagement. On average, parenthood is not associated with lower turnout once individual fixed effects are included. However, substantial heterogeneity emerges over the parental life cycle. Parents of young children vote significantly less than comparable non-parents at younger ages: those with children aged 0–2 and 3–5 exhibit turnout penalties of approximately three to five percentage points. These gaps decline steadily—by about 0.2 percentage points per additional year of parental age—and disappear by around age forty. Parents of older children display no turnout deficit. The participation gap is driven almost entirely by mothers, while fathers’ turnout remains unaffected. The results are robust to alternative specifications and to controls for residential mobility, neighborhood characteristics, and distance to polling stations. Taken together, the findings highlight the importance of life-cycle factors in shaping political participation and suggest that periods of intensive childcare are associated with temporarily lower electoral engagement. More broadly, the analysis points to a channel through which demographic trends, such as delayed fertility, may have implications for democratic representation.
parenthood; age; voter turnout; opportunity cost; administrative data.
D72; J13; J22.
Public Goods and Political Participation: Childcare Access and Electoral Engagement
joint with Giorgio Bellettini, Carlotta Berti Ceroni, Martín Gonzalez-Eiras, and Giovanni Prarolo
Abstract
Keywords
JEL
Paper
Unequal political participation threatens democratic representation when groups with distinct policy preferences are systematically underrepresented at the ballot box. This paper studies how access to childcare-related public goods shapes electoral participation over the life cycle. We combine universe-level administrative turnout records from the city of Bologna with a newly constructed, geocoded panel of nurseries and preschools. Exploiting within-individual variation in proximity to age-appropriate childcare facilities—generated by residential mobility, school openings and closures, and administrative re-drawings of catchment boundaries—we show that greater distance to childcare infrastructure significantly reduces voter turnout among parents of young children. The effect is concentrated in municipal elections, where education policy is directly determined, and is driven by mothers, while effects for fathers are small and statistically insignificant. The magnitude of the effect is economically meaningful and comparable to other well-documented voting costs, despite automatic voter registration and high baseline turnout. These findings highlight a mechanism through which access to local public goods shapes political participation and, in turn, democratic representation.
voter turnout; public goods; childcare; gender; political participation; local elections.
D72; H41; J13; J16.
The Vanishing First Rung: Generative AI and Entry-Level Jobs
joint with Fabio Cerina, and Simone Nobili
Abstract
Keywords
JEL
Paper
Generative AI may not only reduce labor demand; it may also change where careers begin. Using 368 million U.S. online job postings from Lightcast (2016--2025) and a usage-anchored measure of LLM exposure, we study how the release of ChatGPT reshaped posted labor demand across occupations and seniority tiers. Top-quartile-exposed occupations experience a -9.0 log point decline in posting volume relative to less-exposed occupations. The contraction is concentrated at the entry-level margin: junior postings fall by -6.6 log points, compared with about -2 log points for mid- and senior-tier postings. Within exposed occupations, the junior share of postings falls by 3.3 percentage points, with most of the reallocation going to mid-level roles. Event-study estimates show flat pre-trends and a gradual post-2022 divergence, consistent with enterprise diffusion of generative AI. Robustness checks excluding tech-heavy sectors and computer occupations indicate that the results are not mechanically driven by the 2022--2023 technology-sector downturn. The evidence points to a narrowing of the first rung of exposed career ladders: posted demand shifts away from junior roles while demand for more experienced workers remains comparatively intact.
generative AI; ChatGPT; labor demand, seniority; job postings; difference-in-differences.
J23; J24; O33; M51.
State Capacity, Crime Reporting, and Local Development
Political Compensation under Criminal Governance
Costly Signaling and Populist Candidate Selection
Political Selection and Moral Frictions
Student Mobility and Political Participation
Farina E., Rosso M., Dansero L., et al. (2023). Short-term effect of colorectal cancer on income. Journal of Epidemiology & Community Health, 77:196–201.
Abstract
Introduction The ability to return to work after a cancer diagnosis is a key aspect of cancer survivorship and quality of life. Studies have reported a significant risk of income loss for cancer survivors; however, there is limited evidence of the Italian context.
Methods The Work Histories Italian Panel (WHIP)-Salute database was used to select a cohort of incident cases of colorectal cancer (CRC) among workers in the private sector, based on hospital discharges. A propensity score matching was used to find a balanced control group for several confounders. Ordinary least square and logistic regressions were used to estimate the effect of a CRC diagnosis on annual income and the probability of switching from a full-time contract to a part-time one considering 3 years after the diagnosis.
Results Overall, we identified 925 CRC incident cases from 2006 until 2012. Our results confirm a statistically significant reduction in survivors’ income compared with controls. This reduction was greater in the first year and then tend to decrease, with an average income loss over 3 years of about €12 000. Stratified analyses by sex and position confirmed the overall trend while indicating a strong effect modification. Regarding the switching from full-time to part-time employment, the results were never significant.
Conclusion Income loss does not seem to be related to an increase in part-time contracts, but rather to survivors’ reduced work capacity following the invasive treatments. Further research is needed to investigate the complex dynamics behind this association.
Ph.D. Thesis (2026), "Essays in Political Economy and Crime Economics"
Abstract
Chapters
This dissertation studies how non-monetary frictions—such as temptation, information salience, and time constraints—shape individual behavior in contexts where formal institutions are otherwise well-functioning. Combining theoretical analysis with micro-level empirical evidence, the three chapters examine how these frictions affect selection into public employment and participation in democratic processes.

The first chapter develops a behavioral theory of occupational selection in the public sector. Introducing self-control costs into a standard model of career choice, the analysis shows that corruption generates a dual selection effect: while illicit rents attract low-motivation individuals, the psychological costs of resisting temptation deter highly motivated agents. The model identifies an institutional tipping point at which corruption switches from expanding public employment while degrading workforce quality to acting as a severe screening device that improves average quality but reduces participation. These results clarify why similar anti-corruption environments can produce sharply different selection outcomes across countries.

The second chapter examines how crime-related information affects individual voting behavior. Using geolocated data that link retrospective voting choices to local crime news coverage in Bologna across multiple national and municipal elections, the analysis exploits within-individual variation in exposure to nearby crime reports. The results show that aggregate crime salience has weak and unstable effects, while crime attributed to immigrants generates systematic electoral responses: voters shift away from parties with ambiguous positions on immigration toward parties emphasizing law and order. In local elections, immigrant-related crime increases abstention, whereas crimes committed by natives lead to punishment of incumbents. These findings highlight the role of identity-based framing in mediating the political impact of crime.

The third chapter studies how parenthood and parental age shape electoral participation. Using administrative data covering the universe of registered voters in Bologna, the analysis follows individuals over time as their family circumstances evolve. Once permanent individual heterogeneity is accounted for, parenthood is not associated with lower turnout on average. However, substantial life-cycle heterogeneity emerges: parents of infants and preschoolers—especially mothers—exhibit sizable turnout penalties at younger ages, which decline steadily with parental age and disappear by around age forty. These results indicate that periods of intensive childcare impose temporary opportunity costs on political participation.

Taken together, the three chapters show how behavioral frictions and life-cycle constraints can generate distortions in selection and participation even in settings with low formal barriers to entry and participation. By emphasizing micro-level mechanisms rather than institutional failures alone, the dissertation contributes to the understanding of public-sector composition, electoral behavior, and democratic representation.
1. The Tipping Point of Temptation: Occupational Selection and Integrity in the Public Sector
2. Crime Perception and Voting Behavior: Evidence from Individual Data
3. Parenthood, Age, and the Opportunity Cost of Voting: Evidence from Administrative Voter Records
Master's Thesis (2018), "Effect of Breast and Colorectal Cancer on Earnings: Evidence from Italy"
Abstract
Using the Work History Italian Panel (WHIP) we estimate the causal effects of breast and colorectal cancer on earnings, on unemployment, and on the possibility to work part-time in the following three years after the cancer diagnosis. Since cancer patients differ from the rest of the population at socio-economic levels, we perform a propensity score matching to balance our observations. We analyse the effect on earnings throughout an OLS regression and on part-time and unemployment using a logistic regression. We observe that colorectal and breast cancer patients have different effects on earnings. A diagnosis of colorectal cancer appears to be more disabling, leading to a total reduction in earnings up to 10,000 € after three years. On the contrary, the negative effects of breast cancer are more concentrated in the short term and from the second year onwards these women are able to resume a normal working activity. This diversity is also supported by the results we achieved on unemployment and on part-time work.
Bachelor's Thesis (2014), "The Dual Simplex Method"
Abstract
This thesis provides an introduction to linear programming and to the dual simplex method. After presenting the basic concepts of optimization and the standard form of a linear programming problem, the thesis introduces Jordan exchanges as an algebraic tool for transforming systems of linear equations and inequalities. These transformations are then used to describe the simplex method, both from an algebraic and a geometric perspective, with particular attention to feasibility, optimality, degeneracy, and pivoting rules.
The second part of the thesis focuses on duality in linear programming. Starting from the relationship between primal and dual linear systems, the thesis develops the dual formulation of a linear programming problem and discusses the main theoretical results, including weak duality, strong duality, and the Karush-Kuhn-Tucker conditions. Finally, the dual simplex method is presented as a natural counterpart to the primal simplex method. While the primal simplex method preserves primal feasibility and progressively improves optimality, the dual simplex method starts from a dual-feasible table and restores primal feasibility through appropriate pivot operations. The thesis thus highlights the central role of duality in both the theoretical understanding and the algorithmic solution of linear programming problems.