Research

In this project, I quantified the impact of large-scale economic shocks on different population segments following the 2008 financial crisis. I leveraged a longitudinal dataset encompassing 29 countries from 2005-2015. The project required significant data wrangling, merging disparate macroeconomic data from the IMF and OECD with individual-level survey data. I built cross-classified multilevel models to analyze the data's complex structure, using interaction terms to find differential impacts across education groups. My method involved group-mean centering to isolate within- and between-country effects. I validated all findings with extensive robustness tests, including three-way fixed effects, logistic regression, and sequential country-exclusion analyses to ensure durable results. The data revealed a counterintuitive pattern: rising unemployment widened health gaps (low-educated suffered disproportionately), but healthcare spending cuts actually narrowed them. This wasn't because cuts helped the poor. Rather, highly-educated individuals benefited more from robust healthcare systems, so when spending declined, their relative advantage diminished. This project demonstrates skills directly applicable to industry problems: handling large-scale longitudinal data, implementing causal inference techniques, conducting thorough sensitivity analyses, and extracting actionable insights that challenge conventional wisdom. The ability to find non-obvious patterns in complex data is exactly what drives product improvements and business strategy.

In this project, we challenged the popular narrative of the "middle-class squeeze" by leveraging four decades of complex household microdata from the Luxembourg Income Study across twelve major economies, including the US, UK, and Germany. Moving beyond simplistic income brackets, we developed a novel, occupation-based classification model to segment populations into more meaningful economic groups. This allowed for a more granular analysis of how employment and income trends truly evolved from 1980 to 2020. Our analysis revealed that, contrary to the prevailing myth, middle-class employment expanded significantly. The real economic story was the decline of the working class, which faced shrinking job shares and stagnant household income growth, particularly in the US and Germany where it was less than 0.5% per year. We conducted cohort analysis revealing that generational income mobility stalled for working-class families but continued for the middle class. This work highlights my core skills in managing large-scale longitudinal data, applying comparative econometrics, and using feature engineering to redefine a problem. It's a powerful example of my ability to derive clear conclusions from complex data and challenge prevailing narratives.

In this project, I tackled the question of how major economic downturns impact different population segments. Using a large-scale dataset covering 2.2 million people across 23 countries over 14 years, I investigated how the 2008 Great Recession and double dip recession affected earnings inequality. I deployed two robust analytical strategies: three-level multilevel modeling and multivariate time-series analysis. These methods test competing hypotheses and control for country-specific variations. This dual approach ensured the insights were statistically sound and captured the dynamic nature of the economic shock. My analysis delivered a clear, data-backed finding: the recession significantly widened the earnings gap, placing a disproportionate financial burden on the working class. This project showcases my ability to handle complex longitudinal data, perform comparative analysis across different national contexts, conduct extensive robustness testing, and translate results into a clear narrative. It demonstrates the skills needed to analyze how external events affect diverse user segments, providing the data-driven insights necessary to inform strategy and product decisions in a volatile global marketplace.

The 2008 European recession did not affect everyone equally. I used this event as a natural experiment. My goal was to measure how the downturn impacted different young professionals. I grouped them by their socioeconomic background. I combined 12 years of data from two major surveys across six countries. Then, I built regression models to identify which segments were affected most. The results were dramatically different for each segment. Lower-income professionals in Spain saw the largest drop in earnings. In contrast, higher-income groups in the UK and France faced a greater risk of unemployment. This is the same analytical framework used in business to understand how a product change or price adjustment affects different customer segments. My core skill is building models that find these varied effects in complex data. This directly answers questions like, "How did our price change affect power users versus casual users?" or "Which customer segments churned after our UI redesign?"