Filetype
Single-use plastics (SUPs) impose substantial environmental costs. Following Directive (EU) 2019/904, Austria introduced producer charges and mandatory participation in collection and recycling systems. This thesis exploits a monthly aggregated and disaggregated panel of retail offer spells drawn from a price-comparison platform to estimate the extent to which compliance costs pass through to post
This paper investigates whether large information sets improve inflation forecasts for the Euro area by applying the machine learning framework of Medeiros et al. (2021) to monthly Harmonized Index of Consumer Prices (HICP) data over the period 2002 to 2024. Using a dataset of 118 macroeconomic series spanning national accounts, labour market conditions, credit aggregates, financial markets, indus
This thesis forecasts the Cboe Gold ETF Volatility Index (GVZ) using HAR-type models and machine learning methods, including Elastic Net, XGBoost, Random Forest, and a multilayer perceptron, evaluated out-of-sample over 2020–2026 across daily, weekly, and monthly horizons. GVZ's own persistence, captured through daily, weekly, and monthly HAR components, constitutes a strong and difficult-
Financial volatility is important for the correct functioning of the market and notably difficult to forecast. This thesis examines if Patch Encoder-based Mixture-of-Experts (MoE) framework with GARCH, Random Forest, Long Short-Term Memory and Transformer as experts can improve forecasting accuracy. The proposed framework is evaluated across multiple horizons using RMSE, MAE, MSFE, QLIKE and R2 an
This thesis estimates the causal effect of the EU Emissions Trading System on air passenger demand at the country-pair level, using an unbalanced panel of yearly passenger flows for 625 routes over the period 2008 to 2024. The empirical strategy uses a difference-in-differences design with two-way fixed effects, using the distinction between intra-EEA flights covered by the aviation ETS and routes
Forecasting models in financial markets are commonly trained using mean squared error (MSE), the dominant statistical loss function in supervised forecasting, even though portfolio decisions depend on the trade-off between return and risk rather than prediction error alone. This thesis compares Long Short-Term Memory (LSTM) models trained under alternative objectives within a unified forecasting a
This thesis investigates the predictability of returns on Research and Development (R&D) investments among leading European innovation-driven firms. R&D is the engine of long-term economic growth and corporate competitive advantage, yet its returns are notoriously uncertain due to high information asymmetry and long gestation periods. Traditionally, financial analysts and researchers have
This thesis examines how natural disaster exposure affects regional voting behavior in European Parliament (EP) elections between 1994 and 2019. Using a panel of 270 NUTS2 regions constructed from EM-DAT and EU-NED data, the analysis exploits staggered variation in disaster exposure and applies modern and robust difference-in-differences estimators that account for heterogeneous treatment timing (
Most research on remote work asks who can work from home, but rarely who ac- tually wants to. Without a direct measure of preferences, low WFH rates and un- met demand are indistinguishable. This paper uses German Socio-Economic Panel (SOEP) data for 2009–2014 to study both questions in a two-stage framework. The first stage estimates which characteristics predict WFH demand among full-time employ
This study compares a standard FX carry trade benchmark with Random Forest models that incorporate realized volatility and option-implied information extracted from the volatility surface, proxied by risk reversals and butterfly spreads. Investors continuously seek new ways to generate profits in global currency markets, making it important to assess whether machine learning models can extract for
This paper estimates the effect of the 2021-2022 European natural gas price shock on German export competitiveness. Exploiting the structural difference in gas dependence between Germany and Sweden, I apply a triple difference-in-differences design using monthly HS2-level export data from Eurostat. Consistent with a cost pass-through mechanism, the results suggest that German gas-intensive industr
This thesis examines the cohort-level differences in years of schooling and cognitive test scores associated with the introduction of Indonesia's Bantuan Operasional Sekolah (BOS) program, a near-universal per-student block grant launched in 2005. Using four waves of the Indonesia Family Life Survey (IFLS2-5, 1997-2014) and a cohort-based before-and-after design, this study compares children a
This study examines whether cross-border out-licensing announcements by Chinese listed biopharmaceutical firms generate positive abnormal returns and which mechanisms explain these reactions. Using 229 out-licensing events involving 62 Chinese listed firms from January 2020 to February 2026, the study conducted event-study and event-level cross-sectional regressions. The results show that announce
This thesis examines how the European Central Bank’s (ECB) return to positive policy rates, following eight years of negative interest rates, has affected the profitability differential between banks with high and low exposure to deposit funding. The study covers the period 2018-2024 and is based on a sample of 431 banks. To estimate the effect of the transition to positive policy rates on bank pr
Does sentiment in earnings call Q&A sessions help explain stock price reactions beyond reported earnings news? Using FinBERT to classify sentence-level sentiment in OMXS30 earnings calls from 2012 to 2025, we construct event level measures of overall tone and tone dispersion, where dispersion captures the coexistence of positive and negative language within a single discussion. The results sho
Using two cycles of PIAAC survey data collected approximately a decade apart across twelve countries, we study how the task content of work has evolved and what this means for wages. Rather than imposing a predefined taxonomy, we derive five task dimensions directly from the correlation structure of worker-reported survey responses, yielding measures that are empirically grounded and comparable ac
I examine whether the gender of external academic reviewers impacts how likely they are to recommend women for academic positions. As part of the Akademiska Anställningar project at Lund University, I have encoded 434 review letters from 216 Swedish recruitment processes across the years 2018-2020, for the largest scale econometric analysis of external academic reviewers yet. Results find that whi
This essay examines the relationship between environmental, social, and governance (ESG) performance and corporate borrowing costs in Europe, with particular focus on whether this relationship changed after the implementation of the Sustainable Finance Disclosure Regulation (SFDR). Using an unbalanced panel of 7,658 firm-year observations across 1,169 listed non-financial firms in the European Eco
The rapid digitalization of banking has reshaped operations, but its net effect on bank stability remains debated. Digital transformation may improve efficiency and credit screening, yet it may also erode deposit stickiness and amplify funding fragility, making its overall impact theoretically ambiguous. Using a panel of 435 publicly traded U.S. commercial banks (1999–2025), we proxy digitalizatio
