Autoregressive Benchmark Research Articles

Labor market forecasting in unprecedented times: A machine learning approach

AbstractThe COVID‐19 pandemic ushered in unprecedented social and economic conditions, alongside unexpected policy responses, challenging the effectiveness of traditional labor market forecasting approaches. This article presents a novel approach that integrates macroeconomic variables, traditional labor market metrics, and Google search data to develop a machine learning‐based indicator for the Colombian labor market. We employ support vector machine for regression and neural networks models to forecast monthly employment and unemployment rates, explicitly focusing on the third wave of COVID‐19 in the first half of 2021. Our study's findings reveal that the proposed models outperform the autoregressive benchmark regarding forecast accuracy, demonstrating a rapid adaptation to labor market shifts.

Forecasting GDP with many predictors in a small open economy: forecast or information pooling?

The online version contains supplementary material available at 10.1007/s00181-022-02356-9.

Out-of-sample prediction of Bitcoin realized volatility: Do other cryptocurrencies help?

Inspired by cross-market information flows among international stock markets, we incorporate external predictive information from other cryptocurrency markets to forecast the realized volatility (RV) of Bitcoin. To make the most of such external information, we employ six widely accepted approaches to construct predictive models based on multivariate information. Our results suggest that the scaled principal component analysis (SPCA) approach steadily improves the predictive ability of the prevailing heterogeneous autoregressive (HAR) benchmark model considering both the model confidence set (MCS) test and the Diebold–Mariano (DM) test based on three widely accepted loss functions. The forecasting performance is persistent to various robustness checks and extensions. Notably, a mean–variance investor can obtain steady positive economic gains if the investment portfolio is constructed on the basis of the forecasts from the HAR-SPCA model. The results of this study show that external predictive information is statistically and economically important in forecasting Bitcoin RV.

Forecasting the European Monetary Union equity risk premium with regression trees

This paper investigates whether classification and regression trees ensemble algorithms such as bagging, random forests and boosting improve on traditional parametric models for forecasting the equity risk premium. In particular, we work with European Monetary Union (EMU) data for the period from its foundation in 2000 to 2020. The paper first compares the monthly out-of-sample forecasting ability of multiple economic and technical variables using univariate linear regression models and regression tree techniques. The results obtained suggest that regression trees do not show better forecasting ability than a first-order autoregressive benchmark model and univariate linear regressions. The paper then analyses asset allocation strategies with regression trees and checks whether these can select the best economic predictors to form dynamic portfolios composed of two assets: a risk-free asset and an equity index. The results indicate that trading strategies built with two or three economic predictors selected with boosting and random forest algorithms can generate economic value for a risk-averse investor with a quadratic utility function.

Regional heterogeneity and U.S. presidential elections: Real-time 2020 forecasts and evaluation

This paper exploits cross-sectional variation at the level of U.S. counties to generate real-time forecasts for the 2020 U.S. presidential election. The forecasting models are trained on data covering the period 2000–2016, using high-dimensional variable selection techniques. Our county-based approach contrasts the literature that focuses on national and state level data but uses longer time periods to train their models. The paper reports forecasts of popular and electoral college vote outcomes and provides a detailed ex-post evaluation of the forecasts released in real time before the election. It is shown that all of these forecasts outperform autoregressive benchmarks. A pooled national model using One-Covariate-at-a-time-Multiple-Testing (OCMT) variable selection significantly outperformed all models in forecasting the U.S. mainland national vote share and electoral college outcomes (forecasting 236 electoral votes for the Republican party compared to 232 realized). This paper also shows that key determinants of voting outcomes at the county level include incumbency effects, unemployment, poverty, educational attainment, house price changes, and international competitiveness. The results are also supportive of myopic voting: economic fluctuations realized a few months before the election tend to be more powerful predictors of voting outcomes than their long-horizon analogs.

Assessing the usefulness of survey‐based data in forecasting firms' capital formation: Evidence from Italy

AbstractWhile there is a vast macroeconomic literature that singles out the main drivers of capital accumulation in advanced economies during and after the global financial and sovereign debt crises' recessionary phase, there is much less research seeking to identify both models and variables that possess out‐of‐sample forecasting ability for gross fixed capital formation. Moreover, micro‐founded variables are scarcely employed in macroeconomic forecasting of real investment. We fill this gap by considering a battery of univariate and multivariate time series models to forecast investment of nonfinancial corporations in Italy, an interesting case study due to its steep downturn during the two aforementioned crises. We find that a vector error correction model augmented with firm survey‐based variables accounting for business confidence, demand uncertainty and financing constraints generally outperforms the autoregressive benchmark, and a series of competing multivariate time series models in various, alternative, evaluation samples that take into account the impact of both the global financial crisis and the sovereign debt crisis on forecast accuracy.

Economic drivers of commodity volatility: The case of copper

This paper examines whether economic variables provide useful information with which to forecast monthly copper price volatility. Prior literature regarding equity markets has discussed this question extensively, but less is known about these variables’ predictive power in the context of commodity markets. We focus on copper, which is a non-renewable commodity that plays a significant role in several economies and markets. To shed new light on this topic, we employ a Bayesian Model Averaging (BMA) approach to account for both parameter and model uncertainty. Our empirical results show that several economic variables have significant forecasting power when compared against an autoregressive benchmark model, and that predictability varies across the business cycle.

Forecasting stock return volatility using a robust regression model

AbstractThis paper aims to accurately forecast stock return volatility based on a robust regression model. The robust regression model is developed by replacing the mean squared error (MSE) in the autoregressive (AR) model with the Huber loss function, and the resulting model is called the ARH model. The empirical results show that the ARH model displays significantly stronger predictive power than the AR benchmark model for different evaluation periods and forecasting horizons. From an asset allocation perspective, a mean–variance investor can obtain sizeable utility gains based on the volatility forecasts produced by the ARH model. Furthermore, we find that the superior performance of the ARH model comes from assigning small weights for the extreme values, which are mainly found during recessions and periods of high volatility. Finally, our results are robust to various settings.

Predictive Regressions for Aggregate Stock Market Volatility with Machine Learning

We investigate whether machine learning techniques and a large set of financial and macroeconomic variables can be used to predict future S&P realized volatility. We evaluate the aggregate volatility predictions of regularization methods (Ridge, Lasso, and Elastic Net), tree-based methods (Random forest and Gradient boosting), and forecast combination methods. We find that the machine learning algorithms outperform autoregressive benchmark models, both statistically and economically, and that the tree-based methods perform the best. In addition to its past realizations, our analysis reveals that the main drivers of aggregate volatility are several financial and macroeconomic uncertainty proxies.

A comparison of monthly global indicators for forecasting growth

This paper evaluates the predictive content of a set of alternative monthly indicators of global economic activity for nowcasting and forecasting quarterly world real GDP growth using mixed-frequency models. It shows that a recently proposed indicator that covers multiple dimensions of the global economy consistently produces substantial improvements in forecasting accuracy, while other monthly measures have more mixed success. Specifically, the best-performing model yields impressive gains with MSPE reductions of up to 34% at short horizons and up to 13% at long horizons relative to an autoregressive benchmark. The global economic conditions indicator also contains valuable information for assessing the current and future state of the economy for a set of individual countries and groups of countries. This indicator is used to track the evolution of the nowcasts for the U.S., the OECD area, and the world economy during the COVID-19 pandemic and the main factors that drive the nowcasts are quantified.

Backcasting, Nowcasting, and Forecasting Residential Repeat-Sales Returns: Big Data meets Mixed Frequency

The Case-Shiller is the reference repeat-sales index for the U.S. residential real estate market, yet it is released with a two-month delay. We find that incorporating recent information from 71 financial and macro predictors improves backcasts, now-casts, and short-term forecasts of the index returns. Combining individual forecasts with recently-proposed weighting schemes delivers large improvements in forecast accuracy at all horizons. Additional gains obtain with mixed-data sampling methods that exploit the daily frequency of financial variables, reducing the mean square forecast error by as much as 13% compared to a simple autoregressive benchmark. The forecast improvements are largest during economic turmoils, throughout the 2020 COVID-19 pandemic period, and in more populous metropolitan areas.

On the Predictability of China Macro Indicator with Carbon Emissions Trading

Accurate and timely macro forecasting requires new and powerful predictors. Carbon emissions data with high trading frequency and short releasing lag could play such a role under the framework of mixed data sampling regression techniques. This paper explores the China case in this regard. We find that our multiple autoregressive distributed lag model with mixed data sampling method setup outperforms either the auto-regressive or autoregressive distributed lag benchmark in both in-sample and out-of-sample nowcasting for not only the monthly changes of the purchasing managers’ index in China but also the Chinese quarterly GDP growth. Moreover, it is demonstrated that such capability operates better in nowcasting than h-step ahead forecasting, and remains prominent even after we account for commonly-used macroeconomic predictive factors. The underlying mechanism lies in the critical connection between the demand for carbon emission in excess of the expected quota and the production expansion decision of manufacturers.

Macroeconomic forecasting through news, emotions and narrative

This study proposes a new method of incorporating emotions from newspaper articles into macroeconomic forecasts, attempting to forecast industrial production and consumer prices leveraging narrative and sentiment from global newspapers. For the most part, existing research includes positive and negative tone only to improve macroeconomic forecasts, focusing predominantly on large economies such as the US. These works use mainly anglophone sources of narrative, thus not capturing the entire complexity of the multitude of emotions contained in global news articles. This study expands the existing body of research by incorporating a wide array of emotions from newspapers around the world – extracted from the Global Database of Events, Language and Tone (GDELT) – into macroeconomic forecasts. We present a thematic data filtering methodology based on a bi-directional long short term memory neural network (Bi-LSTM) for extracting emotion scores from GDELT and demonstrate its effectiveness by comparing results for filtered and unfiltered data. We model industrial production and consumer prices across a diverse range of economies using an autoregressive framework, and find that including emotions from global newspapers significantly improves forecasts compared to three autoregressive benchmark models. We complement our forecasts with an interpretability analysis on distinct groups of emotions and find that emotions associated with happiness and anger have the strongest predictive power for the variables we predict.

Forecasting UK inflation bottom up

We forecast CPI inflation in the United Kingdom up to one year ahead using a large set of monthly disaggregated CPI item series combined with a wide set of forecasting tools, including dimensionality reduction techniques, shrinkage methods and non-linear machine learning models. We find that exploiting CPI item series over the period 2011–19 yields strong improvements in forecasting UK inflation against an autoregressive benchmark, above and beyond the gains from macroeconomic predictors. Ridge regression and other shrinkage methods perform best across specifications that include item-level data, yielding gains in relative forecast accuracy of up to 70% at the one-year horizon. Our results suggests that the combination of a large and relevant information set combined with efficient penalisation is key for good forecasting performance for this problem. We also provide a model-agnostic approach to address the general problem of model interpretability in high-dimensional settings based on model Shapley values, partial re-aggregation and statistical testing. This allows us to identify CPI divisions that consistently drive aggregate inflation forecasts across models and specifications, as well as to assess model differences going beyond forecast accuracy.

Forecasting Corporate Capital Accumulation in Italy: The Role of Survey-Based Information

While there is a vast macroeconomic literature that singles out the main drivers of capital accumulation in advanced economies during and after the global financial and sovereign debt crises’ recessionary phase, there is much less research seeking to identify both models and variables that possess out-of-sample forecasting ability for gross fixed capital formation. Moreover, micro-founded variables are scarcely employed in macroeconomic forecasting of real investment. We fill this gap by considering a battery of univariate and multivariate time-series models to forecast investment of non-financial corporations in Italy, an interesting case study due to its steep downturn during the two afore-mentioned crises. We find that a vector error correction model augmented with firm survey-based variables accounting for business confidence, demand uncertainty and financing constraints generally outperforms the autoregressive benchmark and a series of competing multivariate time-series models in various, alternative, evaluation samples that take into account the impact of both the global financial crisis and the sovereign debt crisis on forecast accuracy.

Should crude oil price volatility receive more attention than the price of crude oil? An empirical investigation via a large‐scale out‐of‐sample forecast evaluation of US macroeconomic data

AbstractApart from the percentage change in the price of crude oil, there is a growing tradition of using various nonlinear transformations of the price of crude oil to forecast real gross domestic product growth rates, equity returns, inflation and other macroeconomic variables. This study attempts to quantify the additional potential predictive power afforded by crude oil price volatility relative to widely used crude oil price‐based variables for more than 300 US macroeconomic time series at the monthly and the quarterly sampling frequency. We observe that regressions employing crude oil price realized volatility and crude oil price realized semivolatilities tend to afford a more consistent pattern of out‐of‐sample prediction gains relative to competitors using well‐known crude oil price measures and the autoregressive benchmark at the quarterly and monthly sampling frequency. While it is somewhat harder to find evidence of finite‐sample predictive gains relative to the benchmark, the evidence is stronger with respect to population‐level predictability 1 quarter (1 month) ahead for the model with crude oil price realized semivolatilities across the considered data and models. Furthermore, point (density) forecasts employing crude oil price realized volatility tend to be more accurate than corresponding forecasts produced under the crude oil price‐based predictive regressions in a horse race.

Forecasting US stock market volatility: How to use international volatility information

AbstractThis paper aims to accurately forecast US stock market volatility by using international market volatility information flows. The results show the significant ability of the combined international volatility information to predict US stock volatility. The predictability is found to be both statistically and economically significant. Furthermore, in this framework, we compare the performance of a large set of approaches dealing with multivariate information. Dynamic model averaging (DMA) and dynamic model selection (DMS) perform better than a wide variety of competing strategies, including the heterogeneous autoregressive (HAR) benchmark, kitchen sink model, popular forecast combinations, principal component analysis (PCA), partial least squares (PLS), and the ridge, lasso, and elastic net shrinkage methods. A wide range of extensions and robustness checks reduce the concern regarding data mining. DMA and DMS are also able to significantly forecast international stock market volatilities.

Nowcasting GDP using machine-learning algorithms: A real-time assessment

Can machine-learning algorithms help central banks understand the current state of the economy? Our results say yes! We contribute to the emerging literature on forecasting macroeconomic variables using machine-learning algorithms by testing the nowcast performance of common algorithms in a full ‘real-time’ setting—that is, with real-time vintages of New Zealand GDP growth (our target variable) and real-time vintages of around 600 predictors. Our results show that machine-learning algorithms are able to significantly improve over a simple autoregressive benchmark and a dynamic factor model. We also show that machine-learning algorithms have the potential to add value to, and in one case improve on, the official forecasts of the Reserve Bank of New Zealand.

Forecasting financial vulnerability in the USA: A factor model approach

AbstractThis paper presents a factor‐based forecasting model for the financial market vulnerability, measured by changes in the Cleveland Financial Stress Index (CFSI). We estimate latent common factors via the method of the principal components from 170 monthly frequency macroeconomic data in order to forecast the CFSI out‐of‐sample. Our factor models outperform both the random walk and the autoregressive benchmark models in out‐of‐sample predictability at least for the short‐term forecast horizons, which is a desirable feature since financial crises often come to a surprise realization. Interestingly, the first common factor, which plays a key role in predicting the financial vulnerability index, seems to be more closely related with to activity variables rather than nominal variables. We also present a binary‐choice version factor model that estimates the probability of the high stress regime successfully.

A new indicator for nowcasting employment subject to social security contributions in Germany

Contrary to the number of unemployed or vacancies, the number of employees subject to social security contributions (SSC) for Germany is published after a time lag of 2 months. Furthermore, there is a waiting period of 6 months until the values are not revised any more. This paper uses monthly data on the number of people subject to compulsory health insurance (CHI) as auxiliary variable to better nowcast SSC. Statistical evaluation tests using real-time data show that CHI significantly improves nowcast accuracy compared to purely autoregressive benchmark models. The mean squared prediction error for nowcasts of SSC can be reduced by approximately 20%. In addition, CHI outperforms alternative candidate variables such as unemployment, vacancies and industrial production.