Search

In this paper, we document the importance of memory in machine learning (ML)-based models relying on firm characteristics for asset pricing. We come to three empirical conclusions. First, the pure out-of-sample fit of the models is often poor: we find that most R^2 measures are negative, especially when training samples are short. Second, we show that poor fit does not necessarily matter from an investment standpoint: what actually counts are measures of cross-sectional accuracy, which are seldom reported in the literature. Third, memory is key. The accuracy of models is maximal when both labels and features are highly autocorrelated. Relatedly, we show that investments are the most profitable when they are based on models driven by strong persistence. Average realized returns are the highest when the size of training samples is large and when the horizon of the predicted variable is also long.

We apply Factor-MIDAS (FaMIDAS) and Mixed-Frequency Vector Autoregression (MF-VAR and MF-Bayesian VAR) to nowcast quarterly GDP growth of Suriname. For this purpose, we use a set of 44 microeconomic indices over the sample period 2012Q1 - 2020Q2. In the target equation, we regress GDP growth upon its first lag and a beta coefficient. In the explanatory equations the first set of monthly regressors explain the variation of growth without lags while the second set of regressors are fitted with two-month lags. We apply three set of samples for model estimations: 2012Q1 – 2019Q3, 2012Q1 – 2020Q1 and 2012Q1 – 2020Q2. Model nowcast accuracy is benchmarked against GDP growth of 2019 and economic activity growth estimated by the monthly GDP indicator of March and June 2020. The models provide mixed results as compared to the benchmark indicators. We select the models with the lowest Root Mean Squared Error (RMSE) and based on own Judgment to nowcast. As the forecast horizon increases from 2019Q4 to 2020Q2, so do the RMSE. To hedge against high biases and variances, we combine the best nowcasts to produce a single nowcast. Furthermore, it appeared that the FaMIDAS and the MF-VAR models deliver adequate results for two nowcast horizons.Keywords. FaMIDAS; MF-VAR; MF-BVAR; Nowcasting.JEL. C22; C53; E37.

This study attempts to predict high growth firm (HGF) status with financial ratios. Measures related to the firm’s effectiveness in using assets to generate profits, EBITDA margin, debt ratio, equity-to-debt ratio and return on assets are associated with HGF status. While the financial ratios improve HGF prediction, prediction remains modest (AUC = 0.627). This study suggests it is difficult to assume a very good HGF forecast from only financial ratios; therefore, the recommendation for researchers and policymakers building models for predicting HGFs is to incorporate non-financial ratio variables, like the intangible innovation and team-related variables. Finally, study suggests a standardized reporting of prediction performance metrics in the out-of-sample and out-of-time simulation for HGF prediction studies.

This article examines the spatially varying effect of age on single-family house (SFH) prices. Age has been shown to be a key driver for house depreciation and is usually associated with a negative price effect. In practice, however, there exist deviations from this behavior which are referred to as vintage effects. We estimate a spatially varying coefficients (SVC) model to investigate the spatial structures of vintage effects on SFH pricing. For SFHs in the Canton of Zurich, Switzerland, we find substantial spatial variation in the age effect. In particular, we find a local, strong vintage effect primarily in urban areas compared to pure depreciative age effects in rural locations. Using cross validation, we assess the potential improvement in predictive performance by incorporating spatially varying vintage effects in hedonic models. We find a substantial improvement in out-of-sample predictive performance of SVC models over classical spatial hedonic models.

In the euro area, monetary policy is conducted by a single central bank for 20 member countries. However, countries are heterogeneous in their economic development, including their inflation rates. This paper combines a New Keynesian model and a neural network to assess whether the European Central Bank (ECB) conducted monetary policy between 2002 and 2022 according to the weighted average of the inflation rates within the European Monetary Union (EMU) or reacted more strongly to the inflation rate developments of certain EMU countries. The New Keynesian model first generates data which is used to train and evaluate several machine learning algorithms. They authors find that a neural network performs best out-of-sample. They use this algorithm to generally classify historical EMU data, and to determine the exact weight on the inflation rate of EMU members in each quarter of the past two decades. Their findings suggest disproportional emphasis of the ECB on the inflation rates of EMU members that exhibited high inflation rate volatility for the vast majority of the time frame considered (80%), with a median inflation weight of 67% on these countries. They show that these results stem from a tendency of the ECB to react more strongly to countries whose inflation rates exhibit greater deviations from their long-term trend.

Survey forecasts are prone to entry and exit of forecasters as well as forecasters not contributing every period leading to gaps. These gaps make it difficult to compare individual forecasters to each other and raises the question of how to deal with the missing observations. This is addressed for the variables GDP, CPI inflation and unemployment for the US. The theoretically optimal method of filling in missing observations is derived and compared to several competing methods. It is found that not filling in missing observations and taking the previous value do not perform particularly well. For the other methods assessed, there is no clear superior approach for all use cases, but the theoretically optimal one usually performs quite well.

This paper is concerned with the problem of variable selection when the marginal effects of signals on the target variable as well as the correlation of the covariates in the active set are allowed to vary over time, without committing to any particular model of parameter instabilities. It poses the issue of whether weighted or unweighted observations should be used at the variable selection stage in the presence of parameter instability, particularly when the number of potential covariates is large. Amongst the extant variable selection approaches, we focus on the One Covariate at a time Multiple Testing (OCMT) method. This procedure allows a natural distinction between the selection and forecasting stages. We establish three main theorems on selection, estimation post selection, and in-sample fit. These theorems provide justification for using unweighted observations at the selection stage of OCMT and down-weighting of observations only at the forecasting stage. The benefits of the proposed method as compared to Lasso, Adaptive Lasso and Boosting are illustrated by Monte Carlo studies and empirical applications to forecasting monthly stock market returns and quarterly output growths.

To date, only annual information on economic activity is published for the 16 German states. In this paper, we calculate quarterly regional GDP estimates for the period between 1995 to 2021, thereby improving the regional database for Germany. The new data set will regularly be updated when quarterly economic growth for Germany becomes available. We use the new data for an in-depth business cycle analysis and to estimate long-run growth dynamics. The business cycle analysis reveals large heterogeneities in the duration and amplitudes of state-specific fluctuations as well as in the degrees of cyclical concordance. Long-run trends are found to vary tremendously, with positive developments in economically strong regions and flat or even negative trends for economically much weaker states.

Policymakers, firms, and investors closely monitor traditional survey-based consumer confidence indicators and treat it as an important piece of economic information. We propose a latent factor model for the vector of monthly survey-based consumer confidence and daily sentiment embedded in economic media news articles. The proposed mixed-frequency dynamic factor model framework uses a novel covariance matrix specification. Model estimation and real-time filtering of the latent consumer confidence index are computationally simple. In a Monte Carlo simulation study and an empirical application concerning Belgian consumer confidence, we document the economically significant accuracy gains obtained by including daily news sentiment in the dynamic factor model for nowcasting consumer confidence.

This paper analyzes the performance of temporal fusion transformers in forecasting realized volatilities of stocks listed in the S&P 500 in volatile periods by comparing the predictions with those of state-of-the-art machine learning methods as well as GARCH models. The models are trained on weekly and monthly data based on three different feature sets using varying training approaches including pooling methods. I find that temporal fusion transformers show very good results in predicting financial volatility and outperform long short-term memory networks and random forests when using pooling methods. The use of sectoral pooling substantially improves the predictive performance of all machine learning approaches used. The results are robust to different ways of training the models.

We propose a novel risk measure that is built on comparing high-frequency time-varying volatility and low-frequency return spillover estimates. This measure permits to identify the markets that are epidemic in their complex interdependence. We conjecture that initially a highly volatile market experiences episodes of risk transmission, but only later absorbs risk and becomes an epidemic market. Moreover, we can detect newly emerging ‘contagion’ in the system. We examine the behaviour of 30 global equity markets and compare spillover measures, which encapsulate many large and small crises episodes. Instead of relying on ex post crisis information, our model identifies crises periods. An important implication of the proposed approach is that highly interrelated markets, such as China, are less likely to transmit a global economic crisis under the current interdependence setting.

Aggregate demand forecasting, also known as nowcasting when it applies to current quarter assessment, is of notable interest to policy makers. This paper concentrates on the empirical methods dealing with mixed-frequency data. In particular, it focuses on the MIDAS approach and its later extension, the Bayesian MFVAR. The two strategies are evaluated in terms of their accuracy to nowcast Macedonian GDP growth, using same monthly frequency data set. The results of this study indicate that the MIDAS regressions demonstrate comparable forecasting performance to that of MF-VAR model. Moreover, it is interesting to note that the two approaches are reciprocal, since in general, their combined forecast demonstrates clear superiority in predicting business cycle turning points. Additionally, the MF-VAR model showed higher precision in times of increased uncertainty.

Central Bank speeches usually function as aggregators of internal quantitative and qualitative analysis of the institutions regarding the macro economy, the monetary policy and the health of the financial systems. Speeches usually function as a summary of the current status of a countries economic health, the undergoing trends and some future perspectives of the global economy. In this study departing from classical econometrics we employ natural language processing technologies in combination with machine learning techniques in order to filter out the most important signals in the corpus of speeches and translate into a sentiment index for forecasting the future financial markets behaviour. In our analysis, it is evident that central banker's expectations on economy tend to exhibit a predictive ability for financial markets turmoil. Using a combination of dictionaries which are either predefined or build based on historical speeches of the corpus we train an Extreme Gradient Boosting model that generates a sentiment index which signals turmoil with acceptable accuracy when passing a specific threshold.

application/pdf, The natural rate of interest is an important concept since it is a kind of reference for economists in that this is a significant determinant of a variety of policy rules used to characterize monetary policy. Generally, the natural rate of interest is defined as the real short-term interest rate consistent with the output equaling its potential level or the one reflecting stable prices. Estimation of the natural rate of interest is required since it cannot be observed directly. One of the simple ways of estimating the natural rate of interest is to extract the trend of the real short-term interest rate. If the central bank is inclined to raise the policy interest rate when it acknowledges an inflationary pressure or excess demand and if such a policy implementation is smoothly transmitted into adjustment processes of prices and economic activities, observed short-term real interest rate should behave closely around the evolutionary pass of the natural rate of interest. In this sense, statistically extracted trend of real short-term interest rate can be presumed as the estimated value of the natural rate of interest. We utilize the Hodrick-Prescott filter for our extraction. On the other hand, we should have the estimation of expected inflation rate in order to acquire real short-term interest rate if we follow the Fisher equation. In this paper, Kanoh (2006) procedure is adopted to estimate the expected inflation rate. Our estimation result brings us to the conclusion that the natural rate of interest in Japan is likely to show a substantial decline in recent 20 years.

Football coaches must inevitably prepare their strategies not only for the upcoming match but also be prepared to restructure their strategies during the match itself. But there's more than that. The coach's duties extend beyond the four lines on weekends. The coach and his staff must plan a full week of training while considering everyone's current physical capabilities to avoid excessive fatigue or an injury that could prevent the player from being eligible to play for weeks. It is unquestionably in a coach's best interest to keep his players from being hurt or over-exhausted to the point where they are no longer suitable for the upcoming game. Fatigue or injuries are occurrences that can have a negative impact on every club stakeholder. Following investigations conducted by notable researchers like Bangsbo et al. (2006), Krustrup et al. (2010), Mohr et al. (2003), Rampini et al. (2009), Novak et al. (2021), and others, it was possible to achieve highly encouraging results that could actually be useful for coaches by providing them with knowledge of their players' degrees of fatigue in anticipation by developing a predictive model for three different time periods of anticipation (5-minutes; 10-minutes and 20-minutes) based on accurate monitoring of the players' GPS positions. Academically, it is believed that this work will open the door for more research initiatives of this kind as it was among the first, if not the first, to integrate real data to develop a realistic prediction model that could be used to evaluate fatigue. Os treinadores de futebol devem inevitavelmente preparar as suas estratégias não apenas para o próximo jogo, mas também estar preparados para redesenhar as mesmas durante a própria partida. Mas há mais do que isso. As funções do treinador vão para além das quatro linhas ao fim-de-semana. O treinador e o seu staff devem planear uma semana inteira de treinos, considerando as capacidades físicas, atuais, dos jogadores, com o objetivo de evitar fadiga excessiva ou lesões que possam impedir os jogadores de ser opção para jogar durante semanas. É inquestionavelmente do interesse de um treinador evitar que os seus jogadores se lesionem ou fiquem fatigados ao ponto de não serem opções válidas para o próximo jogo. Fadiga ou lesões são ocorrências que podem ter impacto negativo em todos os stakeholders do clube. Suportado em investigações conduzidas por investigadores notáveis como Bangsbo et al. (2006), Krustrup et al. (2010), Mohr et al. (2003), Rampini et al. (2009), Novak et al. (2021), e outros, foi possível obter resultados altamente encorajadores que podem ser realmente úteis para os treinadores, possibilitando fornecer informação relativamente ao grau de fadiga dos jogadores antecipadamente, através de um modelo preditivo desenvolvido para três diferentes períodos (5-minutos; 10-minutos e 20-minutos) com base na monitorização nas coordenadas GPS. Academicamente, acredita-se que este trabalho poderá abrir portas para mais iniciativas de pesquisa no âmbito, dado que foi uma das primeiras, se não a primeira, a integrar dados reais para desenvolver um modelo preditivo realista que pode ser usado para classificar fadiga.

This paper presents a new facet of the regional forecasts of the Institute for Employment Research (IAB). For the development in 2022, we report results at the level of administrative district types of the Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR). Overall, the forecasts for 2022 suggest a return to the regional disparities that already existed before the pandemic with regard to the development of unemployment and employment. While there are no major differences between the administrative district types in the unemployment forecast, the expected growth in employment increases in cities and urban areas compared to the previous year, but stagnates in the case of the two rural structural types.

India has been hard-hit by the Covid-19 pandemic. The virus has exacted a heavy toll in terms of lives lost and deteriorating health outcomes. The economic consequences of the pandemic have been similarly grim. In this paper we attempt an initial, interim, assessment of the impacts of the crisis on poverty. We review the growing literature that considers emerging poverty impacts, noting that there remain significant knowledge gaps due to limited evidence on current welfare outcomes. We analyze pre-Covid survey data to examine the incidence of chronic poverty and downward mobility during a period of rapid economic growth and declining poverty. A profile of poverty during such a period might offer a plausible, partial, window on population groups currently at risk. We suggest that, notwithstanding the severe initial impacts of the crisis on poverty, there are grounds for expecting further consequences going forward. As the virus has spread out of the relatively affluent cities, and as economic stagnation persists, rural areas, with historically higher rates of chronic poverty and vulnerability, may see particularly sharp increases in poverty. While recent vaccination developments offer some grounds for optimism, there remains an urgent need to identify, implement and amplify effective policy alleviation measures.

The value-at-risk (Va) method in market risk management is becoming a benchmark for measuring “market risk” for any financial instrument. The present study aims at examining which VaR model best describes the risk arising out of the Indian equity market (Bombay Stock Exchange (BSE) Sensex). Using data from 2006 to 2015, the VaR figures associated with parametric (variance–covariance, Exponentially Weighted Moving Average, Generalized Autoregressive Conditional Heteroskedasticity) and non-parametric (historical simulation and Monte Carlo simulation) methods have been calculated. The study concludes that VaR models based on the assumption of normality underestimate the risk when returns are non-normally distributed. Models that capture fat-tailed behaviour of financial returns (historical simulation) are better able to capture the risk arising out of the financial instrument.

Artificial intelligence and machine learning have increasing influence on the financial sector, but also on economy as a whole. The impact of artificial intelligence and machine learning on banking risk management has become particularly interesting after the global financial crisis. The research focus is on artificial intelligence and machine learning potential for further banking risk management improvement. The paper seeks to explore the possibility for successful implementation yet taking into account challenges and problems which might occur as well as potential solutions. Artificial intelligence and machine learning have potential to support the mitigation measures for the contemporary global economic and financial challenges, including those caused by the COVID-19 crisis. The main focus in this paper is on credit risk management, but also on analysing artificial intelligence and machine learning application in other risk management areas. It is concluded that a measured and well-prepared further application of artificial intelligence, machine learning, deep learning and big data analytics can have further positive impact, especially on the following risk management areas: credit, market, liquidity, operational risk, and other related areas.

Este trabalho tem como objetivo analisar documentos textuais submetidos por empresas portuguesas no momento de candidatura a programas de incentivos empresariais públicos. Com esta análise pretende-se extrair e selecionar variáveis relevantes, presentes nos textos, que possuam poder preditivo em relação a futuras ações das empresas candidatas aceites, no decorrer dos projetos. O objetivo concreto é a predição da anulação de projetos com fundos atribuídos, durante a sua duração prevista. Para realizar esta análise foi necessário criar uma cadeia de classificação de texto na qual são aplicadas variadas técnicas de processamento da língua natural, extração e seleção de variáveis, seleção e utilização de classificadores, e métricas de avaliação dos resultados. Foram utilizadas técnicas de referência de extração de variáveis como a extração de valores TF e TF-IDF e foram igualmente levadas a cabo experiências de extração de variáveis baseadas em geração de tópicos, análise de similaridade textual, análise de diversidade lexical, exploração de vocabulário específico, entre outros tipos de análise do conteúdo textual. A exploração de variáveis criadas a partir destas experiências mostra-nos características escondidas nos dados, como por exemplo, o facto de se verificar uma maior incidência de projetos com elevados níveis de similaridade em certos distritos do país. O principal objetivo foi alcançar o melhor desempenho possível nas métricas obtidas através da matriz de confusão (taxa de acerto; precisão; cobertura; F1-Score) na predição da anulação de projetos. Os melhores resultados da predição de anulação foram obtidos por um conjunto de variáveis provenientes de diversos métodos de extração e utilizando o algoritmo Classificador Naïve Bayes: 79% de taxa de acerto; 77% de precisão; 71% de cobertura; 74% de F1-Score. Neste trabalho é assim demonstrado o proveito da mistura de variáveis provenientes de diferentes métodos de extração de variáveis. This work aims to analyze the textual documents presented by Portuguese companies when applying for business incentive programs. This work intends to extract and select relevant features, present in the texts, which have predictive power in relation to future actions of the companies whose projects were accepted, during the projects. The concrete goal is the prediction of the cancellation of the projects with allocated funds, during their expected duration. It was necessary to create a text classification pipeline which applies natural language processing, various features extraction and selection techniques, classification algorithms and evaluation metrics. Many feature extraction techniques were used, such as classical techniques as TF and TF-IDF values generation, as also other experiments as topic generation, similarity analysis, lexical analysis, identification of specific vocabulary used, among other analysis of textual content that were also carried out. The feature analysis can show us hidden characteristics in the data, such as the fact that there is a preponderance of projects with high levels of similarity in certain districts of the country. The main objective, regarding the perdition of cancellation of the projects, was achieving the best possible performance, for that there were used the confusion matrix metrics (accuracy; precision; revocation; F1-Score). The best prediction results were obtained by a set of features from different extraction methods together with the use of the Naïve Bayes Classifier algorithm: 79% accuracy; 77% precision; 71% recall; 74% F1-Score. Therefore, it is shown the advantages of mixing features from different extraction methods on this text classification application.

This dissertation focus, firstly, on evaluating the forecast accuracy of the macroeconomic variables, GDP growth and inflation, issued for Portugal for the current year, 6 months ahead, 1 year ahead and 18 months ahead. Regarding the period of the analysis, this paper studies a longer period (2002-2021), which includes forecasts from Central Bank of Portugal, Ministry of Finance, European Commission, Organization for Economic Cooperation and Development and International Monetary Fund and a shorter period (2016-2021), that also includes the Portuguese Public Finance Council. To evaluate and compare the accuracy of the forecasts, summary statistics are computed, namely the mean forecast error, the mean absolute forecast error and root mean squared forecast error. Moreover, tests for unbiasedness and efficiency are performed to assess whether the forecasts are optimal. By using the Diebold Mariano test and a modified version proposed later, the difference in accuracy between institutions is evaluated. Secondly, the goal is to combine the forecasts and analyze if the combined forecasts are better than the individual institutions’ forecasts. The results suggest that, in general, the Central Bank of Portugal is the best institution, especially for the longer period. The results for the shorter period are more inconsistent, due to the small number of observations. In most cases, forecasts are unbiased and efficient. Regarding the combination of forecasts, the average of all institutions has a better performance than the majority of single institutions. A presente dissertação, numa primeira fase, avalia a qualidade de previsões, crescimento do PIB e inflação, realizadas para Portugal para o ano corrente, e para os 6 meses, 1 ano e 18 meses seguintes. São considerados dois períodos de análise: um período mais longo (2002-2021) que inclui as previsões do Banco de Portugal, Ministério das Finanças, Comissão Europeia, a Organização para a Cooperação e Desenvolvimento Económico e o Fundo Monetário Internacional e um período mais curto (2016-2021) que inclui, para além das referidas, o Conselho das Finanças Públicas. Para avaliar e comparar a qualidade das previsões são usadas estatísticas descritivas, como o erro de previsão médio, o erro de previsão absoluto médio e a raiz quadrada do erro de previsão. Adicionalmente são realizados testes de enviesamento e eficiência, que são dois requisitos de previsões ótimas. De forma a avaliar se os erros de previsão são significativamente diferentes entre as instituições, usa-se o teste de Diebold Mariano e uma versão modificada proposta posteriormente. Numa segunda fase, o objetivo passa por combinar as previsões e analisar se apresentam uma melhor qualidade comparativamente com as instituições. No período mais longo, os resultados sugerem que o Banco de Portugal é a instituição com melhor desempenho. Relativamente ao período mais curto, os resultados são mais inconsistentes devido ao pequeno número de observações. Maioritariamente, as previsões são não enviesadas e eficientes. Em relação à combinação de previsões, a média de todas as instituições demonstra ter um desempenho melhor que muitas das instituições individualmente.

O presente trabalho tem como principal objetivo estudar séries temporais financeiras, nomeadamente as taxas de câmbio, e identificar o método que apresenta melhores resultados nas suas previsões. Atualmente a análise das taxas de câmbio tem sido um constante desafio, devido à imprevisibilidade das mesmas. Para contornar este problema foram escolhidos instrumentos econométricos específicos com a intenção de encontrar o modelo mais eficiente em prever valores de taxas de câmbio. Neste estudo foram utilizadas as taxas de câmbio EUR/CNY e EUR/USD, ambas com um total de 242 observações entre janeiro de 2002 e fevereiro de 2022 e ambas com frequência mensal, e os modelos econométricos ARIMA, ARIMA-GARCH e Random Walk, incluindo a utilização do algoritmo Prophet. Foram também selecionadas algumas medidas de erro para comparar as previsões obtidas com os valores reais das séries temporais financeiras. Da análise feita, concluiu-se que o modelo ARIMA(0,1,1)-GARCH(1,1) apresentou as melhores previsões in-sample, para ambas as séries originais. As previsões out-of-sample para estas séries não foram as melhores, sendo que o modelo que apresentou a melhor previsão out-of-sample foi o ARIMA(0,1,1), para ambos os casos. Em relação aos retornos, os modelos que apresentaram as melhores previsões in-sample para as séries RetLogEURCNY e RetLogEURUSD foram o Prophet e o MA(1), respetivamente. Nas previsões out-of-sample, foi o Prophet que apresentou o menor erro de previsão, ainda que elevado, para ambos os casos. The main objective of this work is to study financial time series, namely exchange rates, and to identify the method that presents the best results in its forecasts. Currently, the analysis of exchange rates has been a constant challenge, due to their unpredictability. To overcome this problem, specific econometric instruments were chosen with the intention of finding the most efficient model to predict exchange rate values. In this study, the EUR/CNY and EUR/USD exchange rates were used, both with a total of 242 observations between January 2002 and February 2022 and both at a monthly frequency, and the econometric models used were the ARIMA, ARIMA-GARCH and Random Walk, including the use of the Prophet algorithm. Some error measures were also selected to compare the forecasts with the actual values of both financial time series. From the analysis, it was concluded that the ARIMA(0,1,1)-GARCH(1,1) model presented the best in-sample forecasts, for both original series. The out-of-sample forecasts for the original series were not the best, but the model that presented the best out-ofsample forecasts was the ARIMA(0,1,1), for both cases. Regarding the returns, the models that presented the best in-sample forecasts for the RetLogEURCNY and RetLogEURUSD series were the Prophet and MA(1), respectively. In the out-of-sample forecasts, Prophet presented the smallest forecast error, although high, for both cases.

The following thesis presents an empirical analysis of how Covid-19 affected the real economic activity, credit, and the Fed funds rate in the US between March 4th, 2020, and 9th, 2022. Another question of interest is whether the credit helped boost real economic activity in this period. To achieve the objective, VAR models are employed. To estimate the impact of Covid-19 on these variables, we use OIRF’s. The results indicate that, there will be a negative response in the real economic activity index to the new confirmed cases or deaths growth rate Covid-19 one-standard deviation shock, between 0.3% and 0.16%. Also, the total credit growth rate and the Fed Funds rate are not considerably affected, with an effect close to zero. Moreover, in terms of impact duration, there is a negative effect for fifteen weeks on the real economic activity index caused by the Covid-19 deaths growth rate with an average effect of 0.3%. In the models with the discriminated credit, we see which types of credit most contributed to the short-term economic activity: consumer and commercial and industrial loans, which create a positive effect of about 0.2% and 0.15%, on average one month after the initial shock. Regarding the accumulated impact, consumer loans are the only type of credit that seems effective in boosting real economic activity, with the peak occurring two months after the initial shock with an average effect of 0.24%. Finally, we conclude that our VAR models are not suited to predict future variables values. A seguinte tese apresenta uma análise empírica de como o Covid-19 afetou a atividade económica, o crédito e a taxa de juro da Federal Reserve nos EUA, entre 4 de março de 2020 e 9 de março de 2022. Outra questão é se o crédito ajudou a impulsionar a atividade económica. Para atingir tal objetivo, são aplicados modelos VAR. Para estimar o impacto do Covid-19 nestas variáveis, utilizamos funções de Resposta a Impulsos Ortogonais. Os resultados indicam que, após um choque de um desvio-padrão na taxa de crescimento do número de casos ou mortes do Covid-19 haverá uma resposta negativa na variação do índice de atividade económica, entre 0,3% e 0,16%. A taxa de crescimento do crédito total e a taxa de juro do Federal Reserve apresentam um efeito perto de zero. Sobre duração a do impacto, existe um efeito negativo médio de 0,3% até quinze semanas na variação do índice de atividade económica, causado pela taxa de crescimento de mortes por Covid-19. Nos modelos com o crédito discriminado, verificamos os seguintes tipos de crédito que mais contribuíram para a atividade económica: crédito ao consumidor, e crédito comercial e industrial, que geram um efeito semanal de 0,2% e 0,15% respetivamente, quatro semanas após o choque inicial de um desvio-padrão. Sobre o efeito acumulado, o crédito ao consumidor é o único tipo de crédito que é eficaz para impulsionar a atividade económica, com o pico dois meses após o choque inicial, com um efeito médio semanal de 0,24%. Finalmente, concluímos que nossos modelos VAR inadequados para prever futuros valores das variáveis.

The importance of using the right statistical, mathematical and computational tools can highly influence the decision-making process. With the recent computational progress, Deep Learning methodologies based on Artificial Intelligence seem to be pointed out as a promising tool to study financial time series, characterised by out-of-the-ordinary patterns. Cryptocurrencies are a new asset class with several specially interesting characteristics that still lack deep study and differ from the traditional time series. Bitcoin in particular is characterised by extraordinary high volatility, high number of structural breaks and other identified characteristics that might further difficult the study and forecasting of the time series using classical models. The goal of this study is to critically compare the forecasting properties of classic methodologies (ARCH and GARCH) with Deep Learning Techniques (with MLP, RNN and LSTM architectures) when forecasting Bitcoin’s Volatility. The empirical study focuses on the forecasting of Bitcoin’s Volatility using such models and comparing its forecasting quality using MAE and MAPE for one, three- and seven-day’s forecasting horizons. The Deep learning methodologies show advantages in terms of forecasting quality (when we take in consideration the MAPE) but also require huge computational costs. Diebold-Mariano tests were also performed to compare the forecasts concluding the superiority of Deep Learning Methodologies. A importância de usar as ferramentas estatísticas, matemáticas e computacionais certas pode certamente influenciar o processo de decisão. Com os recentes avanços computacionais, as metodologias Deep-Learning, baseadas em Inteligência Artificial apontam para uma ferramenta promissora para o estudo de séries temporais de dados financeiros, caracterizadas por padrões que são fora do normal. As criptomoedas são uma nova classe de ativos que são caracterizados por alta volatilidade, elevado número de quebras de estrutura e outras características que podem dificultar o estudo e previsão por parte de modelos clássicos. O objetivo deste trabalho é analisar de forma crítica as capacidades de previsão das metodologias clássicas (ARCH e GARCH) comparativamente a metodologias de Deep-Learning (nomeadamente arquiteturas de redes neuronais: MLP, RNN e LSTM) para a previsão da volatilidade da bitcoin. O estudo empírico deste trabalho foca-se na previsão da volatilidade da bitcoin com os modelos supramencionados e comparar a sua qualidade preditiva usando as medidas de erro MAE e MAPE para horizontes de previsão de um, três e sete dias. As metodologias de Deep-Learning apresentam algumas vantagens no que respeita à qualidade de previsão (pela análise da métrica de erro MAPE) mas apresentam um custo computacional superior. Também foram realizados Testes de Diebold-Mariano para comparar as previsões, concluindo-se a superioridade das metodologias de Deep-Learning.

In this paper, we investigate the downside and upside risk spillovers from three kinds of commercial banks (state-owned commercial banks (SOCBs), joint-stock commercial banks (JSCBs) and city commercial banks (CCBs)) to China’s financial system by proposing a new copula quantile regression-based CoVaR model. We find that (i) the dynamic risk spillovers show heterogeneity over time, specifically that its downward trend is significant after the stock market disaster in 2015; (ii) JSCBs display the largest risk spillovers, indicating that JSCBs are the main contributors to systemic risk in China’s financial system; and (iii) the risk spillovers are not symmetrical, as the upside risk spillovers are smaller than the downside risk spillovers. Our results have crucial implications for financial regulators and investors who want to measure and prevent systemic financial risk and optimise their investment strategies.

This study examines the connectedness between the US yield curve components (i.e., level, slope, and curvature), exchange rates, and the historical volatility of the exchange rates of the main safe-haven fiat currencies (Canada, Switzerland, EURO, Japan, and the UK) and the leading cryptocurrency, the Bitcoin. Results of the static analysis show that the level and slope of the yield curve are net transmitters of shocks to both the exchange rate and its volatility. The exchange rate of the Euro and the volatility of the Euro and the Canadian dollar exchange rate are net transmitters of shocks. Meanwhile, the curvature of the yield curve and the Japanese Yen, Swiss Franc, and British Pound act mainly as net receivers. Our static connectedness analysis shows that Bitcoin is mainly independent of shocks from the yield curve’s level, slope, and curvature, and from any main currency investigated. These findings hint that Bitcoin might provide hedging benefits. However, similar to the static analysis, our dynamic analysis shows that during different periods and particularly in stressful times, Bitcoin is far from being isolated from other currencies or the yield curve components. The dynamic analysis allows us to observe Bitcoin’s connectedness in times of stress. Evidence supporting this contention is the substantially increased connectedness due to policy shocks, political uncertainty, and systemic crisis, implying no empirical support for Bitcoin’s safe-haven property during stress times. The increased connectedness in the dynamic analysis compared with the static approach implies that in normal times and especially in stressful times, Bitcoin has the property of a diversifier. The results may have important implications for investors and policymakers regarding their risk monitoring and their assets allocation and investment strategies.

This paper builds a short-term inflation projections (STIP) model for Latvia. The model is designed to forecast highly disaggregated consumer prices using cointegrated ARDL approach of [Pesaran, M., & Shin, Y. (1998). An Autoregressive Distributed Lag Modelling Approach to Cointegration Analysis. Econometric Society Monographs, 31, 371-413.]. We assess the forecast accuracy of STIP model using out-of-sample forecast exercise and show that our model outperforms both aggregated and disaggregated AR(1) benchmarks. Across inflation components, the forecast accuracy gains are 20-30% forecasting 3 months ahead and 15-55% forecasting 12 months ahead.

The objective of this study is to compare the different methods which are effective in predicting data of the short-term effect of COVID-19 confirmed cases and DJI closed stock market in the US. Data for confirmed cases of COVID-19 has been obtained from Worldometer, the database of Johns Hopkins University and the US stock market data (DJI) was obtained from Yahoo Finance. The data starts from 20 January 2020 (first confirmed COVID-19 case the US) to 06 December 2020 and DJI data covers 21 January 2019 to 04 December 2020. COVID-19 data was tested for the period 30 November to 06 December and DJI from 25 November 2020 to 04 December. From the result, we find that the method SutteARIMA was found more suitable to calculate the daily forecasts of COVID-29 confirmed cases and DJI in the US and this method has been used in this study. For the evaluation of the prediction methods, the accuracy measure means absolute percentage error (MAPE) has been used. The MAPE value with the SutteARIMA of 0.56 and 0.60 for COVID-19 and DJI stock respectively was found to be smaller than the MAPE value with ARIMA method.

While the use of simple deterministic models to calculate rental value growth (RVGrowth) rate of reversionary freeholds across epochs prior to upward rent review appears illusive, literature evidence of the synthesis between short-cut DCF valuation and Solver tools in a spreadsheet does not constitute an exhaustive list of solutions. This study examined alternative spreadsheet and iteration tools that can determine RVGrowth rate of freehold investment properties across rent review epochs. With recourse to a hypothetical case of a freehold investment property, this experimental study identified the mathematical composition of RVGrowth in an explicit DCF framework, performed short-cut DCF Valuation and equivalent yield calculation at specific epochs prior to and including the full reversion; as well as using Goal Seek to calculate RVGrowth across all epochs prior to- and including the full reversion. Excel® Solver and Goal Seek, as well as the graphing/root-solving tool in Kyplot® were found to feasibly produce identical results for RVGrowth rate. This is among the limited studies that identified and researched the veracity of alternative tools for RVGrowth rate iteration. The value of this study is the awareness of alternative analytical tools avail freehold investors who desire knowledge of RVGrowth rate when making purchase-, hold-, and sales decisions.

Research purpose. The study is focused on the expected market dynamics of global cobotics markets. This study investigates the current market structure of the global cobotics market. The scientific aim of the research is to report the first data-based market structure analysis of the global cobotics market with the HHI index and with the LKI index analysis. With analysis we are able to show the diversification rate of the global cobotics market. Design / Methodology / Approach. The methodology is based on conventional statistical index theory and statistics. The methodology is the calculation of the Herfindahl-Hirchman Index and the Lauraéus-Kaivo-oja Index. The authors compare the results of these two methodologies. Findings. The Herfindahl-Hirschman Index (HHI) and the Lauraéus-Kaivo-oja Index are statistical measures of market concentration, and they can be used to determine market competitiveness. This paper demonstrates novel data analytics possibilities of new market data collected by the Statzon, Ltd with various comparative analytical results and findings. By our analyses we can help multiple industrial stakeholders make faster decisions and better strategic plans with the easiest and fastest access to accurate, reliable, and up-to-date cobotics industry statistics, forecasts, and insights. The finding is that this study reveals the current market structure of global cobotics. It is a novel finding and result. Originality / Value / Practical implications. This paper demonstrates the novel and exciting possibilities of transparent index calculation tools. The authors provide original results. Authors underline that extra value added to stakeholders and customers will be provided by joint data pooling strategy of various data sources, which is a key approach of this paper. Real-time market structure analyses create reliable and knowledge-based information for decision-makers and stakeholders of the global cobotics industry.

This paper proposes a novel forecast reconciliation framework using Bayesian state-space methods. It allows for the joint reconciliation at all forecast horizons and uses predictive distributions rather than past variation of forecast errors. Informative priors are used to assign weights to specific predictions, which makes it possible to reconcile forecasts such that they accommodate specific judgmental predictions or managerial decisions. The reconciled forecasts adhere to hierarchical constraints, which facilitates communication and supports aligned decision-making at all levels of complex hierarchical structures. An extensive forecasting study is conducted on a large collection of 13,118 time series that measure Swiss merchandise exports, grouped hierarchically by export destination and product category. We find strong evidence that in addition to producing coherent forecasts, reconciliation also leads to substantial improvements in forecast accuracy. The use of state-space methods is particularly promising for optimal decision-making under conditions with increased model uncertainty and data volatility. © 2020 Elsevier B.V., European Journal of Operational Research, 291 (2), ISSN:0377-2217, ISSN:1872-6860

Given the tremendous growth of factor allocation strategies in active and passive fund management, we investigate whether factor or sector asset allocation strategies provide investors with a superior performance. Our focus is on comparing factor versus sector allocations as some recent empirical evidence indicates the dominance of sector over country portfolios. We analyze the performance and performance differences of sector and factor portfolios for various weighting and portfolio optimization approaches, including “equal-weighting” (1/N), “risk parity,” minimum-variance, mean-variance, Bayes–Stein and Black–Litterman. We employ a sample-based approach in which the sample moments are the input parameters for the allocation model. For the period from May 2007 to November 2020, our results clearly reveal that, over longer investment horizons, factor portfolios provide relative superior performances. For shorter periods, however, we observe time-varying and alternating performance dominances as the relative advantage of one over the other strategy depends on the economic cycle. One important insight is that during “normal” times factor portfolios clearly dominate sector portfolios, whereas during crisis periods sector portfolios are superior offering better diversification opportunities. Supplementary Information The online version contains supplementary material available at 10.1057/s41260-021-00225-1.

Following (Almeida, Ardison, Kubudi, Simonsen, & Vicente, 2018) we implement a segmented three factor Nelson-Siegel model for the yield curve using daily observable bond prices and short term interbank rates for Colombia. The flexible estimation for each segment (short, medium, and long) provides an improvement over the classical Nelson-Siegel approach in particular in terms of in-sample and out-of-sample forecasting performance. A segmented term structure model based on observable bond prices provides a tool closer to the needs of practitioners in terms of reproducing the market quotes and allowing for independent local shocks in the different segments of the curve.

This paper forecasts the response of Islamic banks’ dynamics (size, profitability, nonperforming financing, and stability) to the COVID-19 pandemic over the period ranging from 2019Q4 to 2021Q4. Nine jurisdictions are considered based on their Islamic banks’ systemic importance, namely Bahrain, Brunei, Indonesia, Kuwait, Malaysia, Pakistan, Saudi Arabia, Turkey, and UAE. Using the bivariate VARX model, our forecasting exercise shows that the Islamic banks’ response to the COVID-19 pandemic is not uniform across jurisdictions. While the Islamic banks’ dynamics in Saudi Arabia, UAE, and Kuwait are less likely to be impaired, Bahrain, Brunei, Malaysia, Pakistan, and Turkey are expected to be relatively more affected especially in terms of their size growth. Saudi Arabia will continue leading the growth momentum of the global Islamic banking sector, and its Islamic banks’ assets are expected to reach at least $185.4 billion by the end of the fourth quarter of 2021. This paper recommends a prioritization approach for the implementation of the policy measures by the jurisdictions based on their banks-specific responses to the COVID-19 pandemic.

We use daily data from Lombardy, the Italian region most affected by the COVID-19 outbreak, to calibrate a SIR model individually on each municipality. These are all covered by the same health system and, in the post-lockdown phase we focus on, all subject to the same social distancing regulations. We find that municipalities with a higher number of cases at the beginning of the period analyzed have a lower rate of diffusion, which cannot be imputed to herd immunity. In particular, there is a robust and strongly significant negative correlation between the estimated basic reproduction number ($R_0$) and the initial outbreak size, in contrast with the role of $R_0$ as a \emph{predictor} of outbreak size. We explore different possible explanations for this phenomenon and conclude that a higher number of cases causes changes of behavior, such as a more strict adoption of social distancing measures among the population, that reduce the spread. This result calls for a transparent, real-time distribution of detailed epidemiological data, as such data affects the behavior of populations in areas affected by the outbreak., Comment: Data and code are available upon request

In this paper, we examine the ability of Fisher effect to describe the subjective behaviour of monetary policy responses for nations constrained by global factors. We developed and estimated a simple DSGE model for appraising the consequence of an integrated financial market predictor on national monetary policy response in Africa’s largest economies – Nigeria and South Africa. The paper integrated the theoretical intuition of the famous Fisher effect on the New Keynesian DSGE model with global predictors to describe national monetary policy response as it influence domestic financial variables and macroeconomic fundamentals. Simulations show that the existence of global factors threatens the abilities of national monetary policy to predict financial variables and macroeconomic fundamentals in their economies.

We propose to add ranking restrictions on impulse‐responses to sign restrictions to narrow the identified set in vector autoregressions (VARs). Ranking restrictions come from micro data on heterogeneous industries in VARs, bounds on elasticities, or restrictions on dynamics. Using both a fully Bayesian conditional uniform prior and prior‐robust inference, we show that these restrictions help to identify productivity news shocks in the data. In the prior‐robust paradigm, ranking restrictions, but not sign restrictions alone, imply that news shocks raise output temporarily, but significantly. This holds both in an application with rankings in the form of heterogeneity restrictions and in another applications with slope restrictions as rankings. Ranking restrictions also narrow bounds on variance decompositions. For example, the bound of the contribution of news shocks to the forecast error variance of output narrows by about 30 pp at the one‐year horizon. While misspecification can be a concern with added restrictions, they are consistent with the data in our applications. Structural VAR set‐identification sign restrictions ranking restrictions heterogeneity posterior bounds Bayesian inference sampling methods productivity news C32 C53 E32

Change-point (CP) processes are one flexible approach to model long time series. We propose a method to uncover which model parameters truly vary when a CP is detected. Given a set of breakpoints, we use a penalized likelihood approach to select the best set of parameters that changes over time and we prove that the penalty function leads to a consistent selection of the true model. Estimation is carried out via the deterministic annealing expectation-maximization algorithm. Our method accounts for model selection uncertainty and associates a probability to all the possible time-varying parameter specifications. Monte Carlo simulations highlight that the method works well for many time series models including heteroskedastic processes. For a sample of fourteen hedge fund (HF) strategies, using an asset-based style pricing model, we shed light on the promising ability of our method to detect the time-varying dynamics of risk exposures as well as to forecast HF returns.

In this study, we examine the response of emerging stock markets due to the uncertainty of pandemics and epidemics (UPE), including the COVID-19 pandemic. We demonstrate this by evaluating the stock return predictability of 24 emerging market stocks using the new datasets on uncertainty due to pandemics as well as the global fear index for the COVID-19 pandemic. We partition the data sample into periods before and after the announcement of the COVID-19 pandemic and employ panel data techniques that account for salient features of both the series and predictive model. We found that emerging stock markets are more vulnerable to UPE than developed market stocks. Put differently, developed stock markets provide a better hedge against UPE than emerging stock markets. We also find that incorporating the UPE indicator in the valuation of stocks, particularly during pandemics, is crucial for investment decisions.

We develop novel forecasting methods for panel data with heterogeneous parameters and examine them together with existing approaches. We conduct a systematic comparison of their predictive accuracy in settings with different cross-sectional (N) and time (T) dimensions and varying degrees of parameter heterogeneity. We investigate conditions under which panel forecasting methods can perform better than forecasts based on individual estimates and demonstrate how gains in predictive accuracy depend on the degree of parameter heterogeneity, whether heterogeneity is correlated with the regressors, the goodness of fit of the model, and, particularly, the time dimension of the data set. We propose optimal combination weights for forecasts based on pooled and individual estimates and develop a novel forecast poolability test that can be used as a pretesting tool. Through a set of Monte Carlo simulations and three empirical applications to house prices, CPI inflation, and stock returns, we show that no single forecasting approach dominates uniformly. However, forecast combination and shrinkage methods provide better overall forecasting performance and offer more attractive risk profiles compared to individual, pooled, and random effects methods.

In view of the sector-wide effect of the nCOVID-19 pandemic in the USA and the probable effect on certain over-the-counter (OTC) pharmaceutical products, the current study examined potential inflation in the pharmaceutical industry arising from the pandemic-related uncertainty. In this case, the USA's producer price indexes vis-à-vis inflation of the immune-related pharmaceutical items: multivitamin, vitamins nutrients and hematinic (V-N-H), other vitamins (other-V), antidepressant, and antidiabetic were examined alongside the uncertainties arising from the world pandemic and economic policy. Thus, the (Diebold and Yilmaz in Int J Forecast 28(1): 57-66, 2012) result implied that the world pandemic uncertainty contributed a significantly huge shock to the examined pharmaceutical compounds, thus affirming the vulnerability of certain pharmaceuticals to pandemic-related uncertainty. The total spillover increased from 34.2% (with economic policy uncertainty) to 47.6% (when pandemic uncertainty is incorporated). In specific, there are negative net spillovers from the multivitamins, other vitamins, antidiabetic, and antidepressant especially due to high pandemic and economic policy uncertainties. However, the statistical evidence implied that higher uncertainty arising from the pandemic is responsible for the severity of shock received by the indicated pharmaceutical products as against economic policy uncertainty. Thus, a relevant policy inference is posited from the result of the study.

The number of short-time workers from January to April 2020 is used to now- and forecast quarterly GDP growth. We purge the monthly log level series from the systematic component to extract unexpected changes or shocks to log short-time workers. These monthly shocks are included in a univariate model for quarterly GDP growth to capture timely, current-quarter unexpected changes in growth dynamics. Included shocks additionally explain 24% in GDP growth variation. The model is able to forecast quite precisely the decrease in GDP during the financial crisis. It predicts a mean decline in GDP of 5.7% over the next two quarters. Without additional growth stimulus, the GDP level forecast remains persistently 4% lower in the long run. The uncertainty is large, as the 95% highest forecast density interval includes a decrease in GDP as large as 9%. A recovery to pre-crisis GDP level in 2021 lies only in the upper tail of the 95% highest forecast density interval.

Research background: Even though in recent decades, a lot of new techniques were developed, there is still a lack of studies aimed at comparing the performance of variable selection methods. Bankruptcy prediction is an excellent example of the conservative research field with the tendency to use classical approaches. Although the results of studies in this field are directly applied in banks and other financial institutions, variables selected for these models can be biased by the author’s preference for one technique. Purpose: This work aims to compare different variable selection approaches and introduce a new methodology of sequential variable selection that can be applied when the low-dimensional model is preferred. Research methodology: This study has been conducted on Polish companies’ insolvency data from the period of 2007–2013. The risk has been modeled with logistic regression; hence variables have been selected with approaches suitable for linear models. Results: The one-step methods did not lead to sufficient dimensionality reduction, while the sequential approach provided compact models keeping the high-performance level. Also, this method allowed us to identify the main financial determinants of insolvency for studied companies, which are the volume of total assets and the ratio of profit to total assets. Novelty: This paper compares different variable selection methods and demonstrates the effectiveness of their sequential application for dimensionality reduction.

In the present article, we deal with a generalization of the logistic function. Starting from the Riccati differential equation with constant coefficients, we find its analytical form and describe basic properties. Then we use the generalized logistic function for modeling some economic phenomena.

NK models are agent-based simulations of market evolution generated through new entry and firm innovation, and are often focused on better understanding complex interdependencies in organizational phenomena. We provide a counterpoint to the mostly optimistic descriptions of the advantages and of the requirements for these fitness-landscape-based analyses. We do so by offering a comprehensive list of limitations of that modelling technique as well as a critical analysis of two recent applications of the NK model—one of a theory-testing application and one of a theory-building application. Our analysis reveals that when care is not taken to capture the essential parts of a phenomenon, the NK approach may be unnecessary at best, and misleading at worst. We discuss the implications of these analyses and update past suggestions for future uses of NK simulations in organizational research.

This paper proposes an Exponential HEAVY (EHEAVY) model. The model specifies the dynamics of returns and realized measures of volatility in an exponential form, which guarantees the positivity of volatility without restrictions on parameters and naturally allows the asymmetric effects. It provides a more flexible modelling of the volatility than the HEAVY models. A joint quasi-maximum likelihood estimation and closed form multi-step ahead forecasting is derived. The model is applied to 31 assets extracted from the Oxford-Man Institute's realized library. The empirical results show that the dynamic of return volatility is driven by the realized measure, while the asymmetric effect is captured by the return shock (not by the realized return shock). Hence, both return and realized measure are included in the return volatility equation. Out-of-sample forecast and portfolio exercise further shows the superior forecasting performance of the EHEAVY model, in both statistical and economic sense.

We propose a novel copula approach to producing density forecasts of economic aggregates combining models using disaggregate data. Our copula approach is more flexible compared to existing techniques, because it is applicable to any econometric model that produces density forecasts. We construct a set of Monte Carlo studies to investigate the properties of the suggested approach. In our empirical application, we use the Norwegian index for goods consumption (VKI) and the Norwegian consumer price index for underlying inflation (CPI-ATE). We find that the copula approach compares well to alternative methods using recursive out-of-sample estimation.

We use a long history of global temperature and atmospheric carbon dioxide (CO2) concentration to estimate the conditional joint evolution of temperature and CO2 at a millennial frequency. We document three basic facts. First, the temperature-CO2 dynamics are non-linear, so that large deviations in either temperature or CO2 concentrations take a long time to correct-on the scale of multiple millennia. Second, the joint dynamics of temperature and CO2 concentrations exhibit multimodality around historical turning points in temperature and concentration cycles, so that prior to the start of cooling periods, there is a noticeable probability that temperature and CO2 concentrations may continue to increase. Finally, evaluating the future evolution of temperature and CO2 concentration conditional on alternative scenarios realizing, we document that, even conditional on the net-zero 2050 scenario, there remains a significant risk of elevated temperatures for at least a further five millennia.

The COVID-19 pandemic has led to enormous data movements that strongly affect parameters and forecasts from standard VARs. To address these issues, we propose VAR models with outlier-augmented stochastic volatility (SV) that combine transitory and persistent changes in volatility. The resulting density forecasts are much less sensitive to outliers in the data than standard VARs. Predictive Bayes factors indicate that our outlier-augmented SV model provides the best data fit for the pandemic period, as well as for earlier subsamples of relatively high volatility. In historical forecasting, outlier-augmented SV schemes fare at least as well as a conventional SV model.