Non-parametric forecasting for conditional asset allocation

Since the dawn of the 21st century, major global economies have maintained prolonged periods of low interest rates to stimulate economic growth, establishing a stable low-interest-rate macroenvironment. However, in recent years, driven by factors such as elevated inflation and tightening monetary policies, the global interest rate landscape has embarked upon a structural transition from prolonged low rates towards higher rates. This shift has profoundly impacted the pricing logic of financial markets and investors' asset allocation frameworks. Traditional allocation strategies now face the risk of becoming obsolete, creating an urgent need for systematic research to provide theoretical and practical support. Focusing on the impact of evolving interest rate environments on investors' asset allocation decisions, this analysis examines asset performance across different interest rate cycles, with a specific emphasis on transition periods. By examining representative cases of high and low interest rates, it traces the historical performance of various asset classes during distinct interest rate phases and presents data analysis. Research findings indicate that bonds and growth equities demonstrate superior performance during low-interest-rate periods, while value equities, commodities, and cash-like assets yield more stable returns during high-interest-rate phases. Key signals identifying interest rate transition periodssuch as unexpected inflation surges and central bank policy shiftsare identified. Risk management strategies proposed include diversified allocation, dynamic duration adjustment, and increased allocation to inflation-hedging assets, offering practical guidance for investors to navigate interest rate cycles and optimise asset allocation.

Optimal Asset Location and Allocation with Taxable and Tax-Deferred Investing

Chapter 21 - Statistical forecasting—regression and time series analysis

The shift from defined benefit to defined contribution plans in the United States has drawn new attention to the effect of participants' asset allocation decisions on their financial resources for retirement. This paper develops a stochastic simulation algorithm to evaluate the effect of holding a broadly diversified portfolio of common stocks, or a portfolio of index bonds, on the distribution of 401(k) account balances at retirement. We compare the alternative distributions of retirement wealth both by showing the empirical distribution of potential wealth values, and by computing the expected utility of these outcomes under standard assumptions about the structure of household preferences. Our analysis highlights the critical role of other sources of wealth, such as Social Security, defined benefit pension annuities, and saving outside retirement plans in determining the expected utility cost of holding equities in the retirement account. Our findings also demonstrate the importance of the equity premium in affecting investors' utility from different retirement asset allocations. Viewed from the beginning of a working career, and given the historical pattern of returns on stocks and bonds, a household that does not have extremely high risk aversion would achieve a higher expected utility by holding a portfolio of stocks rather than bonds.

This paper sets out a basic framework for solving a stochastic portfolio problem using dynamic programming (DP). Dynamic portfolio decisions are concerned with simultaneous decisions on savings and asset allocation whereby asset returns, such as on equity and bonds, are stochastic as in Campbell and Viceira (Strategic asset allocation, portfolio choice for long-term investors, 2002). In contrast to CV (2002) we do not use a local approximation method to solve the stochastic model but rather use a global solution procedure such as DP. Whereas CV (2002) solve their model by assuming a constant consumption-wealth ratio and equity premium, we can allow both to be time varying. Different variances of equity and bond returns are explored in their impact on saving and asset allocation decisions and on the value function. The stochastic dynamic portfolio decision method proposed here allows for online decisions as data on asset returns are available in real time. The method is set up in a way such that it also helps to make fund decisions online for various types of investment opportunities.

Travel time prediction plays a significant role in traffic data analysis. In the area of travel time prediction, traditional methods such as linear regression and time series models have been widely applied to predict travel times using historical travel time data. However, with the development of artificial intelligence technologies, these traditional models may become replaceable. In recent years, various novel machine learning-based data prediction methods have been developed accordingly. These types of approaches aim to increase the efficiency and accuracy of the prediction. This study intends to employ ensemble learning-based models (including the XGBoost model and RF model) to predict freeway travel time using information such as time of day, day of the week, and weather conditions. The models are validated on a freeway corridor in the area of Charlotte, North Carolina. The temporal and spatial correlations between each segment are also considered in the model. The effects of different feature combinations in the model are investigated and quantified. The modelling results indicate that the XGBoost model has considerable advantages in terms of both prediction accuracy and efficiency. The research findings demonstrate the feasibility of applying XGBoost in travel-time prediction and prove that the XGBoost model is applicable and performs well.

The loosely-coupled microservices architecture has become increasingly popular due to the advantage of its modularity and elasticity in cloud applications. However, it also seriously complicates cloud management and degrades the performance of IT operations. Today, AI has been the locus of commerce and transactions, and transforming traditional IT operations for speed and growth. Inferring the dependencies among an application's microservices can greatly help SREs diagnose possible root causes of performance issues, which is a hard task due to the complex topology of microservices is often unknown in practice. Prior literature on detecting causal structure for cloud services requires significant application instrumentation, which rarely holds in reality. In this work, we leverage Granger causality models on just monitored log data of a microservice-based application to infer the impact of dependencies between microservices. We first describe the approach of modeling discrete log data as time series, and then formally define the Granger causality problem using both linear and nonlinear autoregressive models. Finally, we conduct an extensive comparative study to show the performance of the state-of-the-art linear and nonlinear (i.e., neural) Granger causality methods on both synthetic data and real-world log data from a publicly available benchmark microservice system. Our preliminary results indicate that neural Granger causality models outperform traditional Granger causality methods on both linear and nonlinear time series data, while for large linear time series, linear Granger causal models are more efficient with high accuracy. Using the real-world log data, we also demonstrate our interesting findings on inferred dependency graph of microservices by linear and neural Granger causality models.

We model the asset allocation decision of a stylised corporate defined benefit pension plan in the presence of hedgeable and unhedgeable risks. We assume that plan fiduciaries – who make the asset allocation decision – face non–linear payoffs linked to the plan's funding status because of the presence of pension insurance and a sponsoring employer who may share any shortfall or pension surplus. We find that even simple asymmetries in payoffs have large and highly persistent effects on asset allocation, while unhedgeable risks exert only a small effect. We conclude that institutional details are crucial in understanding DB pension asset allocation.

The popular finance literature describes the asset allocation decision as one of the most important factors in determining investment performance. This article reviews the implications of modern portfolio theory for the asset allocation decision and then examines the recommendations of some leading financial experts to see if they are consistent with theory.

This paper considers an asset allocation strategy over a finite period under investment uncertainty and short-sale constraints as a continuous-time stochastic control problem. Investment uncertainty is characterised by a stochastic interest rate and inflation risk. If there are no short-sale constraints, the optimal asset allocation strategy can be obtained analytically. We consider several kinds of short-sale constraints and employ the backward Markov chain approximation method to explore the impact of short-sale constraints on asset allocation decisions. Our results show that the short-sale constraints do indeed have a significant impact on these decisions.

PurposeThe paper is aimed at modelling time varying betas via a state space representation in order to decompose the marginal contribution to risk of downside and upside deviations of asset returns in portfolio optimisation.Design/methodology/approachThe approach enables to take into account the relationship between risk and excess returns in up‐side and down‐side markets and to arrange a flexible asset allocation model which directly incorporates the investor risk tolerance to positive or negative expected market moves. The model volatility through state space models and the Kalman filter, widely used to recursively and optimally estimate time varying betas.FindingsThe study shows that the application of an asset allocation model which splits beta in two parts, one related to Bear and the other to Bull markets, and reconciles them with a non negative risk aversion parameter may produce interesting financial results if compared with typical passive portfolios. The proposed model was tested by conducting extensive empirical evaluations on a set securities belonging to eight different markets. The outcomes show that active strategies can be developed and can lead to better performances.Research implicationsThe research affects optimisation models in particular considering the volatility indicators usually estimated not only by researchers but also by practitioners.Originality/valueIn financial literature we find empirical evidence that the constant beta model may be inaccurate and hazardous to use in asset allocation decisions and many statistical techniques have been developed to estimate time dependent betas. Rolling regression procedures allow to capture beta dynamics but require the definition of the estimation period. The paper provides an empirical analysis referred both to European and American market data which let us to allocate assets avoiding the usual limits of standard volatility indicators.

The effectiveness of complex, dynamic, and stochastic decisions require an understanding of the tradeoffs between decisions and the facility to perform what if analyses of alternatives. Set in the context of future combat systems, the current study explored features of a robotic asset allocation decision for a surveillance and reconnaissance task. Decision features explored included the method of interaction with plan data, possible decision bias, and robustness following a change in planning assumptions. Findings suggest that planning decisions made from a predetermined set of optimal plans may be faster and more effective than creating plans from scratch if the planning assumptions remain constant. However, plans created from scratch may more deeply engage the planner in the planning conditions, and consequently, result in plans that more effectively handle changes in the planning assumptions.

The EMU countries now share one central bank, one monetary policy, one short-term interest rate and one currency. This structural change has led to many proposed new opportunities to add value in Global Asset Allocation. We investigate many of these proposals, and conclude that most - but not all - are without merit. For example, we conclude that the sudden move from country-based models towards sector-based models is premature, as country factors are likely to increase in importance for the next few years. Sector factors will eventually dominate, but not until Euroland business cycles are synchronized. So the best opportunity to exploit EMU is to allocate across countries in the intermediate term, while preparing to allocate across sectors in the long run.

This report is a primer on the analysis of both linear and nonlinear time series with applications in nuclear safeguards and nonproliferation. We analyze eight simulated and two real time series using both linear and nonlinear modeling techniques. The theoretical treatment is brief but references to pertinent theory are provided. Forecasting is our main goal. However, because our most common approach is to fit models to the data, we also emphasize checking model adequacy by analyzing forecast errors for serial correlation or nonconstant variance.

Increasingly complex economic backgrounds and power consumption have created more difficulties for medium and long-term load forecasting. Traditional load forecasting models such as linear regression and time series models have a good effect on linear data, but the accuracy of nonlinear electric load data in Xinjiang is poor. The BP neural network model has better processing ability for nonlinear data, but it also has the problem of poor fitting and generalization ability. Aiming at this problem, this paper establishes a medium- and long-term power load forecasting model of Bayesian regularized BP neural network. Compared with the general gradient descent method BP neural network model, it is proved that the proposed model has a good effect on the electricity consumption forecast in Xinjiang.