Nowcasting unemployment insurance claims in the time of COVID-19

Near-term forecasts, also called nowcasts, are most challenging but also most important when the economy experiences an abrupt change. In this paper, we explore the performance of models with different information sets and data structures in order to best nowcast US initial unemployment claims in spring of 2020 in the midst of the COVID-19 pandemic. We show that the best model, particularly near the structural break in claims, is a state-level panel model that includes dummy variables to capture the variation in timing of state-of-emergency declarations. Autoregressive models perform poorly at first but catch up relatively quickly. The state-level panel model, exploiting the variation in timing of state-of-emergency declarations, also performs better than models including Google Trends. Our results suggest that in times of structural change there is a bias–variance tradeoff. Early on, simple approaches to exploit relevant information in the cross sectional dimension improve forecasts, but in later periods the efficiency of autoregressive models dominates.     

Nowcasting in real time using popularity priors

We construct a “Google Recession Index” (GRI) using Google Trends data on internet search popularity, which tracks the public’s attention to recession-related keywords in real time. We then compare nowcasts made with and without this index using both a standard dynamic factor model and a Bayesian approach with alternative prior setups. Our results indicate that using the Bayesian model with GRI-based “popularity priors,” we could identify the 2008Q3 turning point in real time, without sacrificing the accuracy of the nowcasts over the rest of the sample periods.     

Forecasting unemployment insurance claims in realtime with Google Trends

Leveraging the increasing availability of ”big data” to inform forecasts of labor market activity is an active, yet challenging, area of research. Often, the primary difficulty is finding credible ways with which to consistently identify key elasticities necessary for prediction. To illustrate, we utilize a state-level event-study focused on the costliest hurricanes to hit the U.S. mainland since 2004 in order to estimate the elasticity of initial unemployment insurance (UI) claims with respect to search intensity, as measured by Google Trends. We show that our hurricane-driven Google Trends elasticity leads to superior real-time forecasts of initial UI claims relative to other commonly used models. Our approach is also amenable to forecasting both at the state and national levels, and is shown to be well-calibrated in its assessment of the level of uncertainty for its out-of-sample predictions during the Covid-19 pandemic.     

Financial nowcasts and their usefulness in macroeconomic forecasting

Financial data often contain information that is helpful for macroeconomic forecasting, while multi-step forecast accuracy benefits from incorporating good nowcasts of macroeconomic variables. This paper considers the usefulness of financial nowcasts for making conditional forecasts of macroeconomic variables with quarterly Bayesian vector autoregressions (BVARs). When nowcasting quarterly financial variables’ values, we find that taking the average of the available daily data and a daily random walk forecast to complete the quarter typically outperforms other nowcasting approaches. Using real-time data, we find gains in out-of-sample forecast accuracy from the inclusion of financial nowcasts relative to unconditional forecasts, with further gains from the incorporation of nowcasts of macroeconomic variables. Conditional forecasts from quarterly BVARs augmented with financial nowcasts rival the forecast accuracy of mixed-frequency dynamic factor models and mixed-data sampling (MIDAS) models.     

Nowcasting growth using Google Trends data: A Bayesian Structural Time Series model

This paper investigates the benefits of internet search data in the form of Google Trends for nowcasting real U.S. GDP growth in real time through the lens of mixed frequency Bayesian Structural Time Series (BSTS) models. We augment and enhance both model and methodology to make these better amenable to nowcasting with large number of potential covariates. Specifically, we allow shrinking state variances towards zero to avoid overfitting, extend the SSVS (spike and slab variable selection) prior to the more flexible normal-inverse-gamma prior which stays agnostic about the underlying model size, as well as adapt the horseshoe prior to the BSTS. The application to nowcasting GDP growth as well as a simulation study demonstrate that the horseshoe prior BSTS improves markedly upon the SSVS and the original BSTS model with the largest gains in dense data-generating-processes. Our application also shows that a large dimensional set of search terms is able to improve nowcasts early in a specific quarter before other macroeconomic data become available. Search terms with high inclusion probability have good economic interpretation, reflecting leading signals of economic anxiety and wealth effects.     

Kaggle forecasting competitions: An overlooked learning opportunity

We review the results of six forecasting competitions based on the online data science platform Kaggle, which have been largely overlooked by the forecasting community. In contrast to the M competitions, the competitions reviewed in this study feature daily and weekly time series with exogenous variables, business hierarchy information, or both. Furthermore, the Kaggle data sets all exhibit higher entropy than the M3 and M4 competitions, and they are intermittent.In this review, we confirm the conclusion of the M4 competition that ensemble models using cross-learning tend to outperform local time series models and that gradient boosted decision trees and neural networks are strong forecast methods. Moreover, we present insights regarding the use of external information and validation strategies, and discuss the impacts of data characteristics on the choice of statistics or machine learning methods. Based on these insights, we construct nine ex-ante hypotheses for the outcome of the M5 competition to allow empirical validation of our findings.     

Mixed random forest,cointegration, and forecasting gasoline prices

One of the most successful forecasting machine learning (ML) procedures is random forest (RF). In this paper, we propose a new mixed RF approach for modeling departures from linearity that helps identify (i) explanatory variables with nonlinear impacts, (ii) threshold values, and (iii) the closest parametric approximation. The methodology is applied to weekly forecasts of gasoline prices, cointegrated with international oil prices and exchange rates. Recent specifications for nonlinear error correction (NEC) models include threshold autoregressive models (TAR) and double-threshold smooth transition autoregressive (STAR) models. We propose a new mixed RF model specification strategy and apply it to the determinants of weekly prices of the Spanish gasoline market from 2010 to 2019. In particular, the mixed RF is able to identify nonlinearities in both the error correction term and the rate of change of oil prices. It provides the best weekly gasoline price forecasting performance and supports the logistic error correction model (ECM) approximation.     

Forecasting volatility via stock return,range, trading volume and spillover effects: The case of Brazil

For the purpose of developing alternative approach for forecasting volatility, we consider heterogeneous VAR (HVAR) model which accommodates the market effects of different horizons, namely, daily, weekly and monthly effects, and examine the interdependence of stock markets in Brazil and the US, based on information of daily return, range and trading volume. To compare with the new approach, we also work with the univariate and multivariate GARCH models with asymmetric effects, trading volumes and fat-tails. The heteroskedasticity-corrected Granger causality tests based on the HVAR show the strong evidence of such spillover effects. We assess the value-at-risk thresholds for Brazil, based on the out-of-sample forecasts of the HVAR model, finding the new approach works satisfactory for the periods including the global financial crisis, without assuming heavy-tailed conditional distributions.     

Google data in bridge equation models for German GDP

Interest in the use of “big data” when it comes to forecasting macroeconomic time series such as private consumption or unemployment has increased; however, applications to the forecasting of GDP remain rather rare. This paper incorporates Google search data into a bridge equation model, a version of which usually belongs to the suite of forecasting models at central banks. We show how such big data information can be integrated, with an emphasis on the appeal of the underlying model in this respect. As the decision as to which Google search terms should be added to which equation is crucial —- both for the forecasting performance itself and for the economic consistency of the implied relationships —- we compare different (ad-hoc, factor and shrinkage) approaches in terms of their pseudo real time out-of-sample forecast performances for GDP, various GDP components and monthly activity indicators. We find that sizeable gains can indeed be obtained by using Google search data, where the best-performing Google variable selection approach varies according to the target variable. Thus, assigning the selection methods flexibly to the targets leads to the most robust outcomes overall in all layers of the system.     

Testing for linear and nonlinear granger causality in the stock price-volume relation: Korean evidence

In this paper, linear and nonlinear Granger causality tests are used to examine the dynamic relationship between daily Korean stock returns and trading volume. We find evidence of significant bidirectional linear and nonlinear causality between these two series. ARCH-ype models are used to examine whether the nonlinear causal relations can be explained by stock returns and volume serving as proxies for information flow in the stochastic process generating volume and stock returns respectively. After controlling for volatility persistent in both series and filtering for linear dependence, we find evidence of nonlinear bidirectional causality between stock returns and volume series. The finding of strong bidirectional stock price-volume causal relationships implies that knowledge of current trading volume improves the ability to forecast stock prices. This evidence is not supportive of the efficient market hypothesis. Another finding is that the nonlinear relationship is sensitive to institutional, organizational, and structural factors. The results of this study should be useful to regulators, practitioners and derivative market participants whose success precariously depends on the ability to forecast stock price movements.     

Dynamic volatility modelling of Bitcoin using time-varying transition probability Markov-switching GARCH model

Bitcoin (BTC), as the dominant cryptocurrency, has attracted tremendous attention lately due to its excessive volatility. This paper proposes the time-varying transition probability Markov-switching GARCH (TV-MSGARCH) models incorporated with BTC daily trading volume and daily Google searches singly and jointly as exogenous variables to model the volatility dynamics of BTC return series. Extensive comparisons are carried out to evaluate the modelling performances of the proposed model with the benchmark models such as GARCH, GJRGARCH, threshold GARCH, constant transition probability MSGARCH and MSGJRGARCH. Results reveal that the TV-MSGARCH models with skewed and fat-tailed distribution predominate other models for the in-sample model fitting based on Akaike information criterion and other benchmark criteria. Furthermore, it is found that the TV-MSGARCH model with BTC daily trading volume and student-t error distribution offers the best out-of-sample forecast evaluated based on the mean square error loss function using Hansen’s model confidence set. Filardo’s weighted transition probabilities are also computed and the results show the existence of time-varying effect on transition probabilities. Lastly, different levels of long and short positions of value-at-risk and the expected shortfall forecasts based on MSGARCH, MSGJRGARCH and TV-MSGARCH models are also examined.     

America's decoupling from China: A perspective from stock markets

America's decoupling-from-China debate started after July 2018, reached its peak in August 2020, and is likely to continue even if it may not be a high priority for the Biden administration. Many studies have examined various aspects of this issue, especially the potential economic impacts on the US economy. Unlike previous research, this study looks at the response of stock markets. Using Google Trends data, this study created a weekly dataset from January 2020 to June 2021 to measure investor sentiment towards the US decoupling from China. Employing the generalised autoregressive conditional heteroskedasticity (GARCH) models, the study finds that concern over decoupling is associated with significant variations in stock market prices. From this we can infer that the overall effects of decoupling on the US economy are likely to be considerable.     

Density Nowcasts and Model Combination: Nowcasting Euro‐Area GDP Growth over the 2008–09 Recession*

Combined density nowcasts for quarterly Euro‐area GDP growth are produced based on the real‐time performance of component models. Components are distinguished by their use of ‘hard’ and ‘soft’, aggregate and disaggregate, indicators. We consider the accuracy of the density nowcasts as within‐quarter indicator data accumulate. We find that the relative utility of ‘soft’ indicators surged during the recession. But as this instability was hard to detect in real‐time it helps, when producing density nowcasts unknowing any within‐quarter ‘hard’ data, to weight the different indicators equally. On receipt of ‘hard’ data for the second month in the quarter better calibrated densities are obtained by giving a higher weight in the combination to ‘hard’ indicators.     

Machine learning advances for time series forecasting

In this paper, we survey the most recent advances in supervised machine learning (ML) and high-dimensional models for time-series forecasting. We consider both linear and nonlinear alternatives. Among the linear methods, we pay special attention to penalized regressions and ensemble of models. The nonlinear methods considered in the paper include shallow and deep neural networks, in their feedforward and recurrent versions, and tree-based methods, such as random forests and boosted trees. We also consider ensemble and hybrid models by combining ingredients from different alternatives. Tests for superior predictive ability are briefly reviewed. Finally, we discuss application of ML in economics and finance and provide an illustration with high-frequency financial data.     

Forecasting the Covid-19 recession and recovery: Lessons from the financial crisis

We consider simple methods to improve the growth nowcasts and forecasts obtained by mixed-frequency MIDAS and UMIDAS models with a variety of indicators during the Covid-19 crisis and recovery period, such as combining forecasts across various specifications for the same model and/or across different models, extending the model specification by adding MA terms, enhancing the estimation method by taking a similarity approach, and adjusting the forecasts to put them back on track using a specific form of intercept correction. Among these methods, adjusting the original nowcasts and forecasts by an amount similar to the nowcast and forecast errors made during the financial crisis and subsequent recovery seems to produce the best results for the US, notwithstanding the different source and characteristics of the financial crisis. In particular, the adjusted growth nowcasts for 2020Q1 get closer to the actual value, and the adjusted forecasts based on alternative indicators become much more similar, all unfortunately indicating a much slower recovery than without adjustment, and very persistent negative effects on trend growth. Similar findings also emerge for forecasts by institutions, for survey forecasts, and for the other G7 countries.     

Measuring output gap nowcast uncertainty

We propose a methodology for gauging the uncertainty in output gap nowcasts across a large number of commonly-deployed vector autoregressive (VAR) specifications for inflation and the output gap. Our approach utilises many output gap measures to construct ensemble nowcasts for inflation using a linear opinion pool. The predictive densities for the latent output gap utilise weights based on the ability of each specification to provide accurate probabilistic forecasts of inflation. In an application based on US real-time data, nowcasting over the out-of-sample evaluation period from 1991q2 to 2010q1, we demonstrate that a system of bivariate VARs produces well-calibrated ensemble densities for inflation, in contrast to univariate autoregressive benchmarks. The implied nowcast densities for the output gap are multimodal and indicate a considerable degree of uncertainty. For example, we assess the probability of a negative output gap at around 45% between 2004 and 2007. Despite the Greenspan policy regime, there still remained a substantial risk that the nowcast for output was below potential in real time. We extend our methodology to include distinct output gap measures, based on alternative filters, and show that, in our application, the nowcast density for the output gap is sensitive to the detrending method.     

Linear and nonlinear dynamic relationships between housing prices and trading volumes

This study applied linear and nonlinear causality tests and estimation models to investigate the efficiency of housing prices and volumes in the United States and its four major regions. The results of this study confirm that housing volumes can function as a price-discovery indicator. According to the nonlinear volatility of housing prices, this study verified numerous hypotheses. Housing returns can also influence housing volume. The results of this study imply that housing price efficiency can vary based on market conditions. Consequently, estimating the behavior of housing prices through a linear model can result in underestimating the information reflected by housing returns.     

On the globalization of stock markets: An application of Vector Error Correction Model,Mutual Information and Singular Spectrum Analysis to the G7 countries

This paper analyzes stock market relationships among the G7 countries between 1973 and 2009 using three different approaches: (i) a linear approach based on cointegration, Vector Error Correction (VECM) and Granger Causality; (ii) a nonlinear approach based on Mutual Information and the Global Correlation Coefficient; and (iii) a nonlinear approach based on Singular Spectrum Analysis (SSA). While the cointegration tests are based on regression models and capture linearities in the data, Mutual Information and Singular Spectrum Analysis capture nonlinear relationships in a non-parametric way. The framework of this paper is based on the notion of market integration and uses stock market correlations and linkages both in price levels and returns. The main results show that significant co-movements occur among most of the G7 countries over the period analyzed and that Mutual Information and the Global Correlation Coefficient actually seem to provide more information about the market relationships than the Vector Error Correction Model and Granger Causality. However, unlike the latter, the direction of causality is difficult to distinguish in Mutual Information and the Global Correlation Coefficient. In this respect, the nonlinear Singular Spectrum Analysis technique displays several advantages, since it enabled us to capture nonlinear causality in both directions, while Granger Causality only captures causality in a linear way. The results also show that stock markets are closely linked both in terms of price levels and returns (as well as lagged returns) over the 36 years analyzed.     

The accuracy of IMF crises nowcasts

The International Monetary Fund (IMF) provides loans to countries in economic crises as a lender of last resort. IMF loan approvals are tied to policy reforms and quantitative targets that reflect the IMF’s crisis assessment. An extensive literature scrutinizes the efficacy of IMF loan programs, instead, we examine the accuracy of the IMF’s crisis assessments (nowcasts) that predicate program designs. Analyzing an unprecedented 602 IMF loan programs from 1992 to 2019, we contradict previous findings that IMF nowcasts are generally optimistic. Disentangling the structure of the IMF’s nowcast bias, we find the IMF systematically overestimates high-growth recoveries GDPs, while low-growth recoveries for low-income countries (LICs) are underestimated. In contrast, non-LICs’ nowcasts exhibit no statistically significant optimistic and pessimistic bias. Interestingly, shorter nowcast horizons do not improve accuracy, and GDP growth nowcasts improved substantially since 2013, while inflation nowcasts remain inefficient. We also isolate the sources of IMF nowcast inefficiencies according to ((i) program objectives, ((ii) program conditionality type, ((iii) geographic regions, ((iv) global crises, and ((v) geopolitics (elections, conflicts, and disasters).     

The impact of sentiment and attention measures on stock market volatility

We analyze the impact of sentiment and attention variables on the stock market volatility by using a novel and extensive dataset that combines social media, news articles, information consumption, and search engine data. We apply a state-of-the-art sentiment classification technique in order to investigate the question of whether sentiment and attention measures contain additional predictive power for realized volatility when controlling for a wide range of economic and financial predictors. Using a penalized regression framework, we identify the most relevant variables to be investors’ attention, as measured by the number of Google searches on financial keywords (e.g. “financial market” and “stock market”), and the daily volume of company-specific short messages posted on StockTwits. In addition, our study shows that attention and sentiment variables are able to improve volatility forecasts significantly, although the magnitudes of the improvements are relatively small from an economic point of view.