Three essays on commodity markets

This dissertation is constituted of three distinct chapters on commodity markets, which cover different aspects of finance. The first chapter focuses on the consequences of the financialization of commodity markets, an important topic from a regulatory perspective. Using a multivariate change-point algorithm, I statistically date when the financialization materializes, with a narrow confidence interval. Using this date as a reference point, I use panel data methods to statistically assess the impact of the financialization on commodity futures prices. I find that it leads to a change in the risk-sharing structure, but without being detrimental neither for a typical investor nor for the functioning of commodity markets. The second chapter employs economically-motivated variables (time to maturity, inventories, and uncertainty resolution) and shows that they are significant determinants of the contemporaneous volatility. Moreover, for the long-term horizons, I find that these variables improve the out-of-sample forecasts of the most advanced high frequency-based, time-varying parameters, models. In a multi-quantile regression setting, I find that these variables do not yield more robust estimations for the out-of-sample expected shortfall. The third chapter explores whether the emergence of commodity index traders affects asset pricing properties (factors and characteristics). Using a Bayesian approach, I determine the optimal set of factors pricing commodity futures. I subsequently use this optimal risk adjustment to test how the financialization alters liquidity and insurance premia. Finally, I use a panel data method to test how these characteristics determine commodity returns, controlling for their potential correlation with factors. More broadly, this chapter also questions the existence of risk premia vs. pure characteristics (frictions) in the asset class of commodity derivatives, as advocated by e.g. Black (1976a), and calls for more research in this area.

Detecting and analyzing regional trends in sub-daily rainfall annual maxima by using the Meta-statistical Extreme Value Distribution

<p>Global warming is expected to modify the regime of extreme precipitation. Physical laws translate increasing atmospheric heat into increasing atmospheric water content that, together with changes in the atmospheric dynamics, drive precipitation changes. The literature generally agrees that extreme precipitation is changing. However, the study of observed annual maximum time series suggests that trends are highly variable in space and uncertain, also as a result of the inherent large stochastic uncertainty of rainfall maxima. In the present work, we exploit the Meta-statistical Extreme Value (MEV) Distribution to investigate the statistical processes behind these trends and understand how they can be related to changing meteorological conditions. The MEV framework was recently proposed for the frequency analyses of extremes under pre-asymptotic conditions and was shown to significantly improve estimation uncertainty for extreme events by using ordinary events. The narrow confidence interval characterizing MEV is a clear advantage for trend analysis, and its ability to separate storm intensity and yearly occurrence permits to better understand the statistical processes underlying extremes. We gathered data from 33 stations in the Trentino region (Eastern Italian Alps) with at least 25 years of 5-minute resolution records (average density 1/190 km<sup>-2</sup>) and computed the parameters describing the yearly intensity distribution of events at multiple durations ranging from 15 minutes to 24 hours as well as their yearly number. The Regional Mann-Kendall test is used for evaluating the presence of trend in the distribution parameters, number of events per year, estimated quantiles and recorded annual maxima. Results confirm the presence of significant trends in the annual maxima. Trends in the 2-year quantiles estimated yearly using MEV are consistent with the observed trends in annual maxima, which are more marked for 15min to 1 hr duration and less marked for 3hr to 24 hr duration. Conversely, trends in rare quantiles (10-year, 100-year) are significant for durations up to 1 hour and become not significant for longer durations. Analysis of the parameters shows that these trends are likely due to a combination of (i) increasing number of storm events per year and increasing intensity of the storms, and (ii) changes in the tail properties of the storms.</p>

The Two-Layer Hierarchical Distribution Model of Zeros of Riemann’s Zeta Function along the Critical Line

This article numerically analyzes the distribution of the zeros of Riemann’s zeta function along the critical line (CL). The zeros are distributed according to a hierarchical two-layered model, one deterministic, the other stochastic. Following a complex plane anamorphosis involving the Lambert function, the distribution of zeros along the transformed CL follows the realization of a stochastic process of regularly spaced independent Gaussian random variables, each linked to a zero. The value of the standard deviation allows the possible overlapping of adjacent realizations of the random variables, over a narrow confidence interval. The hierarchical model splits the ζ function into sequential equivalence classes, with the range of probability densities of realizations coinciding with the spectrum of behavioral styles of the classes. The model aims to express, on the CL, the coordinates of the alternating cancellations of the real and imaginary parts of the ζ function, to dissect the formula for the number of zeros below a threshold, to estimate the statistical laws of two consecutive zeros, of function maxima and moments. This also helps explain the absence of multiple roots.

Outcome of Home-Based Early Intervention for Autism in Sri Lanka: Follow-Up of a Cohort and Comparison with a Nonintervention Group

This paper presents the outcome of a home-based autism intervention program (HBAIP) in 18- to 40-month-old children newly diagnosed and treatment naïve. Intervention was exclusively implemented at home. Outcome was measured at 3 months and 6 months after intervention and compared with a group of newly diagnosed children with autism who were >40 months at intake but had not received any autism specific clinical management. Aim was also to estimate whether natural development would contribute to gain in skills and compare with the effect of intervention. Five selected parameters of behavior representing social interaction and social communication were used to assess outcome. Results showed a statistically significant improvement between preintervention and postintervention in all the measured parameters. The effect size was large when compared to preintervention and gains were indicated by changes in mean scores andpvalues within a narrow confidence interval. Highest gains were in first 3 months of postintervention which continued up to 6 months. Although the comparison group was more advanced in the measured skills at intake, they were significantly below the level reached by experimental group at 3 months and 6 months after intervention. This study was registered in the Sri Lanka Clinical Trials Registry (SLCTR/2009/011).

Assessment of the time constant of relaxation: insights from simulations and hemodynamic measurements

The objective of this study was to use high-fidelity animal data and numerical simulations to gain more insight into the reliability of the estimated relaxation constant derived from left ventricular pressure decays, assuming a monoexponential model with either a fixed zero or free moving pressure asymptote. Comparison of the experimental data with the results of the simulations demonstrated a trade off between the fixed zero and the free moving asymptote approach. The latter method more closely fits the pressure curves and has the advantage of producing an extra coefficient with potential diagnostic information. On the other hand, this method suffers from larger standard errors on the estimated coefficients. The method with fixed zero asymptote produces values of the time constant of isovolumetric relaxation (τ) within a narrow confidence interval. However, if the pressure curve is actually decaying to a nonzero pressure asymptote, this method results in an inferior fit of the pressure curve and a biased estimation of τ.