This thesis develops and validates a hybrid quantum–classical workflow for multi-asset lifecycle asset allocation. Building on the long-horizon portfolio-choice framework of Barberis (2000), the work replicates and extends it to a European, euro-denominated setting, in which equity and government-bond returns are partly predictable through a vector autoregression driven by the dividend yield and inflation. The central empirical finding is a reversal of sign: on euro-area data the dividend yield predicts Euro Stoxx 50 returns with a negative coefficient, so that equity displays mean aversion rather than the mean reversion documented on United States data. The optimal equity share consequently declines with the investment horizon, and euro-area equity is excluded altogether from the optimal five-asset portfolio. The core methodological challenge is computational: exhaustive optimisation becomes infeasible beyond a handful of assets. To address this, the Quantum Approximate Optimization Algorithm (QAOA) is employed as the portfolio optimiser. The quantum solver is first validated against an exact classical benchmark in low-dimensional cases, and then applied through a hierarchical quantum–classical scheme to a five-asset universe that lies beyond classical tractability. The lifecycle dimension is introduced via backward induction, and the pipeline is closed with a human-capital glide path derived from the Bodie–Merton–Samuelson framework, exploiting the wealth-homogeneity of CRRA preferences. Results are evaluated in certainty-equivalent terms throughout, isolating the welfare value of dynamic rebalancing and of human-capital-driven asset allocation. A commodity extension, in which gold replaces the excluded equity sleeve, shows that a large gold position survives Bayes–Stein shrinkage of the expected returns, marking it as a diversification result rather than an artefact of estimation error. The workflow is offered as a reproducible methodological contribution rather than as a predictive or operational tool: the results are in-sample, and the estimated predictive coefficient, while negative throughout most of the sample, varies in magnitude across sub-periods.
This thesis develops and validates a hybrid quantum–classical workflow for multi-asset lifecycle asset allocation. Building on the long-horizon portfolio-choice framework of Barberis (2000), the work replicates and extends it to a European, euro-denominated setting, in which equity and government-bond returns are partly predictable through a vector autoregression driven by the dividend yield and inflation. The central empirical finding is a reversal of sign: on euro-area data the dividend yield predicts Euro Stoxx 50 returns with a negative coefficient, so that equity displays mean aversion rather than the mean reversion documented on United States data. The optimal equity share consequently declines with the investment horizon, and euro-area equity is excluded altogether from the optimal five-asset portfolio. The core methodological challenge is computational: exhaustive optimisation becomes infeasible beyond a handful of assets. To address this, the Quantum Approximate Optimization Algorithm (QAOA) is employed as the portfolio optimiser. The quantum solver is first validated against an exact classical benchmark in low-dimensional cases, and then applied through a hierarchical quantum–classical scheme to a five-asset universe that lies beyond classical tractability. The lifecycle dimension is introduced via backward induction, and the pipeline is closed with a human-capital glide path derived from the Bodie–Merton–Samuelson framework, exploiting the wealth-homogeneity of CRRA preferences. Results are evaluated in certainty-equivalent terms throughout, isolating the welfare value of dynamic rebalancing and of human-capital-driven asset allocation. A commodity extension, in which gold replaces the excluded equity sleeve, shows that a large gold position survives Bayes–Stein shrinkage of the expected returns, marking it as a diversification result rather than an artefact of estimation error. The workflow is offered as a reproducible methodological contribution rather than as a predictive or operational tool: the results are in-sample, and the estimated predictive coefficient, while negative throughout most of the sample, varies in magnitude across sub-periods.
Quantum Optimization for Lifecycle Asset Allocation: A Hybrid QAOA-Classical Workflow for the European Pension Investor
ANDRETTO, DENIS
2025/2026
Abstract
This thesis develops and validates a hybrid quantum–classical workflow for multi-asset lifecycle asset allocation. Building on the long-horizon portfolio-choice framework of Barberis (2000), the work replicates and extends it to a European, euro-denominated setting, in which equity and government-bond returns are partly predictable through a vector autoregression driven by the dividend yield and inflation. The central empirical finding is a reversal of sign: on euro-area data the dividend yield predicts Euro Stoxx 50 returns with a negative coefficient, so that equity displays mean aversion rather than the mean reversion documented on United States data. The optimal equity share consequently declines with the investment horizon, and euro-area equity is excluded altogether from the optimal five-asset portfolio. The core methodological challenge is computational: exhaustive optimisation becomes infeasible beyond a handful of assets. To address this, the Quantum Approximate Optimization Algorithm (QAOA) is employed as the portfolio optimiser. The quantum solver is first validated against an exact classical benchmark in low-dimensional cases, and then applied through a hierarchical quantum–classical scheme to a five-asset universe that lies beyond classical tractability. The lifecycle dimension is introduced via backward induction, and the pipeline is closed with a human-capital glide path derived from the Bodie–Merton–Samuelson framework, exploiting the wealth-homogeneity of CRRA preferences. Results are evaluated in certainty-equivalent terms throughout, isolating the welfare value of dynamic rebalancing and of human-capital-driven asset allocation. A commodity extension, in which gold replaces the excluded equity sleeve, shows that a large gold position survives Bayes–Stein shrinkage of the expected returns, marking it as a diversification result rather than an artefact of estimation error. The workflow is offered as a reproducible methodological contribution rather than as a predictive or operational tool: the results are in-sample, and the estimated predictive coefficient, while negative throughout most of the sample, varies in magnitude across sub-periods.| File | Dimensione | Formato | |
|---|---|---|---|
|
tesi_final.pdf
accesso aperto
Dimensione
2.43 MB
Formato
Adobe PDF
|
2.43 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/115039