Essays on Empirical Macroeconomics: Hysteresis, Tail Risks, and Non-Gaussianity

This thesis develops three lines of research in applied macroeconomics: nonlinear hysteresis effects, tail risks in the climate--growth relationship, and non-Gaussian features of macro time series and their implications for macroeconomic models. The first chapter challenges the standard split between a supply-determined long-run trend and a demand-driven business cycle by studying macroeconomic hysteresis through the lens of asymmetry. While the debate has largely emphasized negative hysteresis - persistent scars from recessions - we ask whether positive demand shocks can also generate lasting gains. In this chapter, we identify a demand shock with potentially long-run effects (a hysteresis shock) using an SVAR with a mix of sign and long-run zero restrictions, and then test whether its effects differ between expansions and contractions using a nonlinear local projection framework. The results suggest a nuanced pattern: negative shocks dominate in the short run, but positive shocks build over the medium run and ultimately exert sizable effects on aggregate activity. To interpret these findings, we show that the key lies in the transmission channels: capital accumulation and innovation via R&D mainly transmit negative hysteresis, whereas positive hysteresis operates primarily through labor-market dynamics, with gradual medium-run increases in participation and sustained declines in unemployment, including long-term unemployment. While the first chapter studies persistent effects of demand shocks with sign-dependent nonlinearity, the second chapter examines weather shocks through a different nonlinear lens: heterogeneity and tail risks across the conditional distribution of GDP growth. This approach is motivated by the concern that climate change may affect macroeconomic vulnerability primarily in the tails rather than only "on average". To study this, we use Panel Quantile Local Projections to trace how local temperature shocks (chronic risk) and global temperature shocks (acute risk) affect the entire conditional distribution over time. We show that global temperature shocks are larger in magnitude than local shocks, but both primarily depress the right tail of the conditional GDP-growth distribution - acting mainly as a brake on growth opportunities by reducing upside potential more than amplifying downside risk. Importantly, these "at-risk" dynamics emerge at medium- to long-run horizons. In a third line of research, I return to business-cycle questions and study how non-Gaussian features of macroeconomic data affect the evaluation of structural macroeconomic models used for policy analysis. For these models to be credible, they must match key empirical regularities; yet many assume Gaussian shocks, despite strong evidence of fat tails in aggregate time series - consistent with lumpy boom-bust dynamics in which large shocks occur more often than under normality. Ignoring this non-Gaussianity can therefore lead to misleading inferences and policy conclusions. Motivated by this, in this chapter we propose a new impulse-response matching procedure that exploits non-Gaussianity in macro time series. We use an indirect inference approach in which the auxiliary model is an SVAR identified via independent component analysis, and we introduce a new minimum-distance criterion to match empirical and model-implied impulse responses. We characterize the estimator's asymptotic properties, assess its performance in Monte Carlo simulations, and illustrate the approach in a benchmark New Keynesian model.