We provide a comprehensive examination of the predictive performance of panel forecasting methods based on individual, pooling, fixed effects, and empirical Bayes estimation, and propose optimal weights for forecast combination schemes. We consider linear panel data models, allowing for weakly exogenous regressors and correlated heterogeneity. We quantify the gains from exploiting panel data and demonstrate how forecasting performance depends on the degree of parameter heterogeneity, whether such heterogeneity is correlated with the regressors, the goodness of fit of the model, and the dimensions of the data. Monte Carlo simulations and empirical applications to house prices and CPI inflation show that empirical Bayes and forecast combination methods perform best overall and rarely produce the least accurate forecasts for individual series.

This article presents a method for selecting variables and determining parameter heterogeneity in Bayesian hierarchical panel data models. Mixture distributions are used as priors for the mean and the variance of the individuals' parameters. Selection indicators determine the best-fitting component of each mixture distribution and indicate whether the mean parameter is non zero and whether the parameters are heterogeneous. The method is applied to two panel data sets. The first is on inflation of US CPI sub-indices, and the results suggest that a heterogeneous panel AR model with a lagged, first principal component is the preferred model. A second application to house price inflation across US metropolitan statistical areas shows that the model includes either the autoregressive component or the lagged spatial components, but not both at the same time.

This paper compares the ability of several econometric and machine learning methods to nowcast GDP in (pseudo) real-time. The analysis takes the example of Dutch GDP over the years 1992-2018 using a broad data set of monthly indicators. It discusses the forecast accuracy but also analyzes the use of information from the large data set of regressors. We find that the random forest forecast provides the most accurate nowcasts while using the different variables in a relative stable and equal manner.

Time-series forecasting is a fundamental problem associated with a wide range of engineering, financial, and social applications.The challenge arises from the complexity due to the time-variant property of time series and the inevitable diminishing utility of predictive models.Therefore, it is generally difficult to accurately predict values, especially in a multi-step ahead setting.However, in domains such as financial time series forecasting, an ex-ante prediction of the relative order of values in the near future is sufficient; i.e., the next 100 days can help make meaningful investment decisions.In this paper, we propose a dynamic prediction framework that makes it possible to make an ex-ante forecast of time series with a special focus on the relative ordering of the forecast within a forward-looking time horizon.Through the lens of the concordance index (CI), we compare the proposed method with conventional regression-based time-series forecasting methods, discriminative learning methods and hybrid methods.Moreover, we discuss the use of the proposed framework for different types of time series and under a variety of conditions.Extensive experimental results on financial time series across a majority of liquid asset classes show that the proposed framework outperforms the benchmark methods significantly.

Mean square forecast error loss implies a bias–variance trade-off that suggests that structural breaks of small magnitude should be ignored. In this paper, we provide a test to determine whether modeling a structural break improves forecast accuracy. The test is near optimal even when the date of a local-to-zero break is not consistently estimable. The results extend to forecast combinations that weight the post-break sample and the full sample forecasts by our test statistic. In a large number of macroeconomic time series, we find that structural breaks that are relevant for forecasting occur much less frequently than existing tests indicate.

We derive forecasts for Markov switching models that are optimal in the mean square forecast error (MSFE) sense by means of weighting observations. We provide analytic expressions of the weights conditional on the Markov states and conditional on state probabilities. This allows us to study the effect of uncertainty around states on forecasts. It emerges that, even in large samples, forecasting performance increases substantially when the construction of optimal weights takes uncertainty around states into account. Performance of the optimal weights is shown through simulations and an application to U.S. GNP, where using optimal weights leads to significant reductions in MSFE. Supplementary materials for this article are available online.

This paper investigates the ability of the adaptive learning approach to replicate the expectations of professional forecasters. For a range of macroeconomic and financial variables, we compare constant and decreasing gain learning models to simple, yet powerful benchmark models. We find that constant gain models provide a better fit for the expectations of professional forecasters. For macroeconomic series they usually perform significantly better than a naive random walk forecast. In contrast, we find it difficult to beat the no-change benchmark using the adaptive learning models to forecast financial variables.

This paper considers the problem of forecasting under continuous and discrete structural breaks and proposes weighting observations to obtain optimal forecasts in the MSFE sense. We derive optimal weights for one step ahead forecasts. Under continuous breaks, our approach largely recovers exponential smoothing weights. Under discrete breaks, we provide analytical expressions for optimal weights in models with a single regressor, and asymptotically valid weights for models with more than one regressor. It is shown that in these cases the optimal weight is the same across observations within a given regime and differs only across regimes. In practice, where information on structural breaks is uncertain, a forecasting procedure based on robust optimal weights is proposed. The relative performance of our proposed approach is investigated using Monte Carlo experiments and an empirical application to forecasting real GDP using the yield curve across nine industrial economies.

This article considers the problem of testing for cross‐section independence in limited dependent variable panel data models. It derives a Lagrangian multiplier (LM) test and shows that in terms of generalized residuals of Gourieroux et al. (1987) it reduces to the LM test of Breusch and Pagan (1980). Because of the tendency of the LM test to over‐reject in panels with large N (cross‐section dimension), we also consider the application of the cross‐section dependence test (CD) proposed by Pesaran (2004). In Monte Carlo experiments it emerges that for most combinations of N and T the CD test is correctly sized, whereas the validity of the LM test requires T (time series dimension) to be quite large relative to N. We illustrate the cross‐sectional independence tests with an application to a probit panel data model of roll‐call votes in the US Congress and find that the votes display a significant degree of cross‐section dependence.

Linkages between banks and insurance companies are important when forecasting the fragility of the banking and insurance sectors. We propose a novel empirical framework that allows us to estimate unobserved linkages in panel data sets that contain observed regressors. We find that taking unobserved common factors into account reduces the root mean square forecasts error of firm specific forecasts by up to 9%, of system forecasts by up to 14%, and by up to 39% for systemic forecasts of more distressed firms relative to a model based on observed variables only. Estimates of the factor loadings suggest that the correlation of financial institutions has been relatively stable over the forecast period.

In this article we consider combining forecasts generated from the same model but over different estimation windows. We develop theoretical results for random walks with breaks in the drift and volatility and for a linear regression model with a break in the slope parameter. Averaging forecasts over different estimation windows leads to a lower bias and root mean square forecast error (RMSFE) compared with forecasts based on a single estimation window for all but the smallest breaks. An application to weekly returns on 20 equity index futures shows that averaging forecasts over estimation windows leads to a smaller RMSFE than some competing methods.

This paper conducts a broad-based comparison of iterated and direct multi-step forecasting approaches applied to both univariate and multivariate models. Theoretical results and Monte Carlo simulations suggest that iterated forecasts dominate direct forecasts when estimation error is a first-order concern, i.e. in small samples and for long forecast horizons. Conversely, direct forecasts may dominate in the presence of dynamic model misspecification. Empirical analysis of the set of 170 variables studied by Marcellino, Stock and Watson (2006) shows that multivariate information, introduced through a parsimonious factor-augmented vector autoregression approach, improves forecasting performance for many variables, particularly at short horizons.

This paper offers a solution to the puzzle that economic activity recovers after a financial crisis without a rebound in credit. These credit-less recoveries, known as "Phoenix Miracles", question the importance of credit. We argue that these recoveries appear credit-less because GDP is compared to the stock of credit. We show in a theoretical model that recoveries in GDP coincide with recoveries in the flow of credit and this can occur even as the stock of credit declines. Data from emerging and developed economies confirm this finding.

This paper considers the problem of forecast combination when forecasts are generated from the same model but use different estimation windows. It develops theoretical results for random walks when their drift and/or volatility are subject to one or more structural breaks. The analysis is then extended to a linear regression model with an exogenous regressor. It is shown that compared to forecasts based on a single estimation window, averaging of forecasts over different estimation windows leads to a lower bias and to a lower root mean square forecast error for all but the smallest of breaks. Similar results are also obtained when observations are exponentially down-weighted, although in this case the forecast performance critically depends on the choice of the weighting coefficient. An application to weekly returns on 20 equity index futures shows that averaging forecasts over estimation windows generally leads to a smaller RMSFE compared to a range of competing methods. JEL classifications C22, C53 * This paper was previously circulated under the title " Forecasting Random Walks under Drift Instability. " We are grateful to University of Cambridge. He acknowledges financial support from Sinopia, quantitative specialist of HSBC Global Asset Management. The opinions expressed in this paper do not necessarily reflect those of De Nederlandsche Bank.

It has become almost a stylized fact that after flnancial crises, eco- nomic activity recovers without a rebound in credit. We investigate the relationship between credit and economic activity over the busi- ness cycle. In a simple model we show that a rebound in the ∞ow of credit has closer relationship with economic recovery than a rebound in the stock of credit. Using data from developed and emerging market countries we flnd that the ∞ow of credit has a higher correlation with GDP than the stock of credit, in particular during recovery periods from flnancial crises.

This paper considers forecast averaging when the same model is used but estimation is carried out over different estimation windows. It develops theoretical results for random walks when their drift and/or volatility are subject to one or more structural breaks. It is shown that compared to using forecasts based on a single estimation window, averaging over estimation windows leads to a lower bias and to a lower root mean square forecast error for all but the smallest of breaks. Similar results are also obtained when observations are exponentially down-weighted, although in this case the performance of forecasts based on exponential down-weighting critically depends on the choice of the weighting coefficient. The forecasting techniques are applied to monthly inflation series of 21 OECD countries and it is found that average forecasting methods in general perform better than using forecasts based on a single estimation window.

This paper presents a canonical, econometric model of contagion and investigates the conditions under which contagion can be distinguished from inter-dependence. In a two-country (market) setup it is shown that for a range of fundamentals the solution is not unique, and for sufficiently large values of the contagion coefficients it has interesting bifurcation properties with bimodal density functions. The extension of the model to herding behaviour is also briefly discussed. To identify contagion effects in the presence of inter-dependencies the equations for the individual markets or countries must contain country (market) specific forcing variables. This sheds doubt on the general validity of the correlation-based tests of contagions recently proposed in the literature which do not involve any country (market) specific fundamentals. Finally, we show that ignoring inter-dependence can introduce an upward bias in the estimate of the contagion coefficient, and using Monte Carlo experiments we further show that this bias could be substantial.

In this paper we discuss tests for residual cross section dependence in nonlinear panel data models. The tests are based on average pair-wise residual correlation coefficients. In nonlinear models, the definition of the residual is ambiguous and we consider two approaches: deviations of the observed dependent variable from its expected value and generalized residuals. We show the asymptotic consistency of the cross section dependence (CD) test of Pesaran (2004). In Monte Carlo experiments it emerges that the CD test has the correct size for any combination of N and T whereas the LM test relies on T large relative to N. We then analyze the roll-call votes of the 104th U.S. Congress and find considerable dependence between the votes of the members of Congress.