Urban Traffic Dynamics Prediction—A Continuous Spatial-temporal Meta-learning Approach

Urban traffic status (e.g., traffic speed and volume) is highly dynamic in nature, namely, varying across space and evolving over time. Thus, predicting such traffic dynamics is of great importance to urban development and transportation management. However, it is very challenging to solve this problem due to spatial-temporal dependencies and traffic uncertainties. In this article, we solve the traffic dynamics prediction problem from Bayesian meta-learning perspective and propose a novel continuous spatial-temporal meta-learner (cST-ML), which is trained on a distribution of traffic prediction tasks segmented by historical traffic data with the goal of learning a strategy that can be quickly adapted to related but unseen traffic prediction tasks. cST-ML tackles the traffic dynamics prediction challenges by advancing the Bayesian black-box meta-learning framework through the following new points: (1) cST-ML captures the dynamics of traffic prediction tasks using variational inference, and to better capture the temporal uncertainties within tasks, cST-ML performs as a rolling window within each task; (2) cST-ML has novel designs in architecture, where CNN and LSTM are embedded to capture the spatial-temporal dependencies between traffic status and traffic-related features; (3) novel training and testing algorithms for cST-ML are designed. We also conduct experiments on two real-world traffic datasets (taxi inflow and traffic speed) to evaluate our proposed cST-ML. The experimental results verify that cST-ML can significantly improve the urban traffic prediction performance and outperform all baseline models especially when obvious traffic dynamics and temporal uncertainties are presented.

When The Last Tree Dies, The Last Man Dies: Do Forests Hold The Key To Survival In Ghana? A Critical Analysis Using The Bootstrapped Rolling Window Granger Causality Test Approach

This paper investigates the role of forests in the life expectancy of people in Ghana. We test whether the extinction of forests will inevitably lead to extinction of people in Ghana. We first examined the causal relationship between life expectancy and deforestation using the full sample bootstrap Granger causality test approach and find causality to run from deforestation to life expectancy with no feedback from life expectancy to deforestation. Testing for parameter stability, we found the short run and long run parameters of the estimated Vector Auto Regressive models to be unstable. A time-varying approach, the rolling window bootstrapped Granger causality test was then employed to investigate the causal relationship between life expectancy and deforestation. The results showed that deforestation has a negative effect on life expectancy, confirming the widely accepted saying that the health of forests is inextricably linked to the health of mankind. The empirical results further show that, on trend higher life expectancy increases the rate of deforestation in Ghana. Highlighting the importance of the role of forests in influencing life expectancy in Ghana, we recommend awareness creation on the role of forests in supporting human life and also extensive afforestation programs to reduce the rate of deforestation in Ghana. This, we believe, will reduce the spread of vector borne diseases such as malaria and reduce the surge in respiratory diseases which shorten the life span of Ghanaians.JEL codesQ23, Q50, Q53, Q58, Q58

Dynamic Conditional Bias-Adjusted Carry Cost Rate Futures Hedge Ratios

This paper proposes new dynamic conditional futures hedge ratios and compares their hedging performances along with those of common benchmark hedge ratios across three broad asset classes. Three of the hedge ratios are based on the upward-biased carry cost rate hedge ratio, where each is augmented in a different bias-mitigating way. The carry cost rate hedge ratio augmented with the dynamic conditional correlation between spot and futures price changes generally: (1) provides the highest hedging effectiveness and (2) has a statistically significantly higher hedging effectiveness than the other hedge ratios across assets, sub-periods, and rolling window sizes.

Tracking COVID-19 Decease Through Rolling Conditional Variance

The effects of COVID-19 have been devastating globally. However, countries have essential asymmetries regarding the disease spread dynamics and the respective mortality rates. In addition to containment strategies and boosting growth and economic development in the face of the COVID-19 pandemic, society calls for solutions that allow the development of vaccines, treatments for the disease, and especially, indicators or early warnings that anticipate the evolution of new infections and deaths. This research aims to track the total deaths caused by COVID-19 in the most affected countries by the pandemics after the approval, distribution, and implementation of vaccines from 2021. We proposed an Autoregressive Integrated Moving Average (ARIMA) specification as a first adjustment. Subsequently, we estimate the conditional variance of total deaths from an Exponential Generalized Autoregressive Conditional Heteroscedasticity (EGARCH). Finally, we compute a rolling density backtesting within a 7-day rolling window to demonstrate the robustness estimation for COVID-19 mortality. The work's main contribution lies in exhibiting a tracking indicator for volatility and COVID-19 direction, including a weekly window to observe its evolution.

Symmetric and Asymmetric Causal Relationship between Oil Prices and G7 Stock Markets: A Bootstrap Rolling-Window Granger Causality Test

This study examines the causal relationships between oil prices and the MSCI stock index of G7 countries between September 2004 and October 2020. This study is novel in implementing symmetric and asymmetric time-varying causality tests based on the bootstrap rolling-window approach. The results reveal that the causal link between oil prices and G7 stock markets is time-dependent. The periods of bidirectional causality roughly coincide with the global financial crisis and the ongoing COVID-19 pandemic. When asymmetry is accounted for, the results suggest an asymmetric causality between the two markets expressed by different patterns regarding positive and negative oil shocks. The results also indicate symmetric causality during the COVID-19 pandemic. These findings have implications for portfolio design and hedging strategies that are important to both policymakers and investors.

Policy turmoil in China: a barrier for FDI flows?

PurposeThis research examines the dynamic interrelationship between economic policy uncertainty (EPU) and the inflows of foreign direct investment (IFDI) in China.Design/methodology/approachThis research used the Granger causality and sub-sample time-varying rolling window causality method.FindingsThe empirical results reveal that EPU tends to have a negative impact on the IFDI in most periods that have been taken into consideration. However, there has been a positive relationship observed between the periods of the US subprime crisis. That is to say that the uncertainty of the Chinese economic policy does not always impede the IFDI. These results are supported by the general equilibrium model, which states that there are certain influences that come into play when moving from EPU to IFDI. On the other hand, the IFDI exert a positive influence on EPU during times of economic crisis and trade war, which indicates that the uncertainty in the economy may increase due to the sudden soar of foreign investment.Originality/valueDuring tense global trade situations and complicated economic scenarios, the results suggest the Chinese government should dedicate itself to expanding its initiatives to open up and improve the domestic business environment in order to increase the foreign investors' confidence and prevent the decline in the IFDI. In addition to this, it also suggests that multinational companies pay attention to the policy environment of the host country, especially when they decide to invest there.

Determining the Rolling Window Size of Deep Neural Network Based Models on Time Series Forecasting

Time series forecasting has always been a significant task in various domains. In this paper, we propose DeepARMA, a LSTM-based recurrent neural network to tackle this problem. DeepARMA is derived from an existing time series forecasting baseline, DeepAR, overcoming two of its weaknesses: (1) rolling window size determination: the way DeepAR determines rolling window size is casual and vulnerable, which may lead to the unnecessary computation and inefficiency of the model;(2) neglect of the noise: pure autoregressive model cannot deal with the condition where data are composed of various kinds of noise, neither do most of time series models including DeepAR. In order to solve these two problems, we first combine a classic information theoretic criterion, AIC, with the network to determine the proper rolling window size. Then, we propose a jointly-learned neural network fusing white Gaussian noise series given by ARIMA models to DeepAR’s input. That is exactly why we name the network ‘DeepARMA’. Our experiments on a real-world dataset demonstrate that our improvement settles those two problems put forward above.

Time-Invariance Coefficients Tests with the Adaptive Multi-Factor Model

This paper tests a multi-factor asset pricing model that does not assume that the return’s beta coefficients are constants. This is done by estimating the generalized arbitrage pricing theory (GAPT) using price differences. An implication of the GAPT is that when using price differences instead of returns, the beta coefficients are constant. We employ the adaptive multi-factor (AMF) model to test the GAPT utilizing a Groupwise Interpretable Basis Selection (GIBS) algorithm to identify the relevant factors from among all traded exchange-traded funds. We compare the performance of the AMF model with the Fama–French 5-factor (FF5) model. For nearly all time periods less than six years, the beta coefficients are time-invariant for the AMF model, but not for the FF5 model. This implies that the AMF model with a rolling window (such as five years) is more consistent with realized asset returns than is the FF5 model.