C53 Forecasting and Other Model Applications
Refine
Year of publication
Document Type
- Working Paper (24)
- Article (2)
- Report (1)
Language
- English (27)
Has Fulltext
- yes (27)
Is part of the Bibliography
- no (27)
Keywords
- real-time data (7)
- forecasting (6)
- forecast combination (5)
- DSGE models (4)
- model uncertainty (4)
- oil price (4)
- Greenbook (3)
- density forecasts (3)
- Bayesian VAR (2)
- Brent (2)
- Machine learning (2)
- Neural Networks (2)
- Prognose (2)
- Transfer Learning (2)
- WTI (2)
- Zinsfuß (2)
- futures (2)
- oil market (2)
- regime-switching (2)
- term structure (2)
- ABS (1)
- ARCH-Prozess (1)
- Angebot (1)
- Apache Spark (1)
- Artificial Intelligence; (1)
- Bayesian inference (1)
- Börsenkurs (1)
- CBO (1)
- CDO (1)
- CLO (1)
- Central Bank Communication (1)
- Conditional Forecasts (1)
- Credit (1)
- Crises Forecasting (1)
- DSGE (1)
- Deutschland (1)
- Distributed Computing (1)
- Econometrics (1)
- Energy Performance Certificate (1)
- Energy efficiency (1)
- European Monetary Union (1)
- Exchange rate regime (1)
- Exponential smoothing (1)
- Factor Structure (1)
- Forecast Comparison/ Competition (1)
- GARCH (1)
- GARCH model (1)
- GARCH-Prozess (1)
- Geschichte 2002 (1)
- Group shrinkage (1)
- Growth-at-Risk (1)
- Index-Futures (1)
- Inflation Forecasting (1)
- Interest Rate Forecasting (1)
- Interval prediction (1)
- Kalman filter (1)
- Kapitalmarkteffizienz (1)
- Labor cost adjustments (1)
- Limit Order Book (1)
- Liquidity Risk (1)
- Liquidität (1)
- Local projection (1)
- MBS (1)
- Machine Learning (1)
- Macroeconomic Forecasting (1)
- Mitgliedsstaaten (1)
- Monetary Policy (1)
- Monte Carlo integration (1)
- Multiple hypothesis testing (1)
- Nachfrage (1)
- Natural Language Processing (1)
- Neural Network (1)
- New Keynesian Models (1)
- Nichtlineare Zeitreihenanalyse (1)
- OECD (1)
- OPEC (1)
- Pfandbrief (1)
- Prediction (1)
- Production, Saving, Consumption and Investment Forecasting (1)
- Prognoseverfahren (1)
- Quantile regression (1)
- Regularized regression (1)
- Retail gasoline price (1)
- Rezession (1)
- Securitisation (1)
- Semiparametric Model (1)
- Structural policies (1)
- Supervised hierarchical clustering (1)
- Sydney Stock Exchange (1)
- Tree-based models (1)
- Volatilität (1)
- Weak moment inequalities (1)
- Wild bootstrap (1)
- Word Embedding (1)
- Zeitreihenanalyse (1)
- acquisition cost (1)
- big data (1)
- business cycles (1)
- combined forecasting (1)
- conditional forecasts (1)
- crack spread (1)
- cross-equation restrictions of rational expectations (1)
- density forecasting (1)
- economic fluctuations (1)
- expectations (1)
- expert forecasts (1)
- factorization of matrix polynomials (1)
- fiscal policy (1)
- forecast (1)
- forecast accuracy (1)
- forecasts (1)
- heterogeneous price expectations (1)
- implied volatility (1)
- interest rates (1)
- market efficiency (1)
- market expectation (1)
- missing data (1)
- mixed frequency (1)
- model misspecification (1)
- monetary policy (1)
- oil price shock (1)
- peak oil (1)
- peso problem (1)
- policy rules (1)
- predictive likelihood (1)
- reconciliation of Lucas's advocacy of rational-expectations modelling and policy predictions and Sims's advocacy of VAR modelling (1)
- refined products (1)
- risk premium (1)
- robust policy (1)
- structural change (1)
- structured finance (1)
- survey expectations (1)
- unconventional oil (1)
This paper contributes a multivariate forecasting comparison between structural models and Machine-Learning-based tools. Specifically, a fully connected feed forward non-linear autoregressive neural network (ANN) is contrasted to a well established dynamic stochastic general equilibrium (DSGE) model, a Bayesian vector autoregression (BVAR) using optimized priors as well as Greenbook and SPF forecasts. Model estimation and forecasting is based on an expanding window scheme using quarterly U.S. real-time data (1964Q2:2020Q3) for 8 macroeconomic time series (GDP, inflation, federal funds rate, spread, consumption, investment, wage, hours worked), allowing for up to 8 quarter ahead forecasts. The results show that the BVAR improves forecasts compared to the DSGE model, however there is evidence for an overall improvement of predictions when relying on ANN, or including them in a weighted average. Especially, ANN-based inflation forecasts improve other predictions by up to 50%. These results indicate that nonlinear data-driven ANNs are a useful method when it comes to macroeconomic forecasting.
The hierarchical feature regression (HFR) is a novel graph-based regularized regression estimator, which mobilizes insights from the domains of machine learning and graph theory to estimate robust parameters for a linear regression. The estimator constructs a supervised feature graph that decomposes parameters along its edges, adjusting first for common variation and successively incorporating idiosyncratic patterns into the fitting process. The graph structure has the effect of shrinking parameters towards group targets, where the extent of shrinkage is governed by a hyperparameter, and group compositions as well as shrinkage targets are determined endogenously. The method offers rich resources for the visual exploration of the latent effect structure in the data, and demonstrates good predictive accuracy and versatility when compared to a panel of commonly used regularization techniques across a range of empirical and simulated regression tasks.
Whatever it takes to understand a central banker : embedding their words using neural networks
(2023)
Dictionary approaches are at the forefront of current techniques for quantifying central bank communication. In this paper, the author propose a novel language model that is able to capture subtleties of messages such as one of the most famous sentences in central bank communications when ECB President Mario Draghi stated that "within [its] mandate, the ECB is ready to do whatever it takes to preserve the euro".
The authors utilize a text corpus that is unparalleled in size and diversity in the central bank communication literature, as well as introduce a novel approach to text quantication from computational linguistics. This allows them to provide high-quality central bank-specific textual representations and demonstrate their applicability by developing an index that tracks deviations in the Fed's communication towards inflation targeting. Their findings indicate that these deviations in communication significantly impact monetary policy actions, substantially reducing the reaction towards inflation deviation in the US.
This paper proposes tests for out-of-sample comparisons of interval forecasts based on parametric conditional quantile models. The tests rank the distance between actual and nominal conditional coverage with respect to the set of conditioning variables from all models, for a given loss function. We propose a pairwise test to compare two models for a single predictive interval. The set-up is then extended to a comparison across multiple models and/or intervals. The limiting distribution varies depending on whether models are strictly non-nested or overlapping. In the latter case, degeneracy may occur. We establish the asymptotic validity of wild bootstrap based critical values across all cases. An empirical application to Growth-at-Risk (GaR) uncovers situations in which a richer set of financial indicators are found to outperform a commonly-used benchmark model when predicting downside risk to economic activity.
Measuring and reducing energy consumption constitutes a crucial concern in public policies aimed at mitigating global warming. The real estate sector faces the challenge of enhancing building efficiency, where insights from experts play a pivotal role in the evaluation process. This research employs a machine learning approach to analyze expert opinions, seeking to extract the key determinants influencing potential residential building efficiency and establishing an efficient prediction framework. The study leverages open Energy Performance Certificate databases from two countries with distinct latitudes, namely the UK and Italy, to investigate whether enhancing energy efficiency necessitates different intervention approaches. The findings reveal the existence of non-linear relationships between efficiency and building characteristics, which cannot be captured by conventional linear modeling frameworks. By offering insights into the determinants of residential building efficiency, this study provides guidance to policymakers and stakeholders in formulating effective and sustainable strategies for energy efficiency improvement.
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.
Linear rational-expectations models (LREMs) are conventionally "forwardly" estimated as follows. Structural coefficients are restricted by economic restrictions in terms of deep parameters. For given deep parameters, structural equations are solved for "rational-expectations solution" (RES) equations that determine endogenous variables. For given vector autoregressive (VAR) equations that determine exogenous variables, RES equations reduce to reduced-form VAR equations for endogenous variables with exogenous variables (VARX). The combined endogenous-VARX and exogenous-VAR equations comprise the reduced-form overall VAR (OVAR) equations of all variables in a LREM. The sequence of specified, solved, and combined equations defines a mapping from deep parameters to OVAR coefficients that is used to forwardly estimate a LREM in terms of deep parameters. Forwardly-estimated deep parameters determine forwardly-estimated RES equations that Lucas (1976) advocated for making policy predictions in his critique of policy predictions made with reduced-form equations.
Sims (1980) called economic identifying restrictions on deep parameters of forwardly-estimated LREMs "incredible", because he considered in-sample fits of forwardly-estimated OVAR equations inadequate and out-of-sample policy predictions of forwardly-estimated RES equations inaccurate. Sims (1980, 1986) instead advocated directly estimating OVAR equations restricted by statistical shrinkage restrictions and directly using the directly-estimated OVAR equations to make policy predictions. However, if assumed or predicted out-of-sample policy variables in directly-made policy predictions differ significantly from in-sample values, then, the out-of-sample policy predictions won't satisfy Lucas's critique.
If directly-estimated OVAR equations are reduced-form equations of underlying RES and LREM-structural equations, then, identification 2 derived in the paper can linearly "inversely" estimate the underlying RES equations from the directly-estimated OVAR equations and the inversely-estimated RES equations can be used to make policy predictions that satisfy Lucas's critique. If Sims considered directly-estimated OVAR equations to fit in-sample data adequately (credibly) and their inversely-estimated RES equations to make accurate (credible) out-of-sample policy predictions, then, he should consider the inversely-estimated RES equations to be credible. Thus, inversely-estimated RES equations by identification 2 can reconcile Lucas's advocacy for making policy predictions with RES equations and Sims's advocacy for directly estimating OVAR equations.
The paper also derives identification 1 of structural coefficients from RES coefficients that contributes mainly by showing that directly estimated reduced-form OVAR equations can have underlying LREM-structural equations.
The authors examine the effectiveness of labor cost reductions as a means to stimulate economic activity and assesses the differences which may occur with the prevailing exchange rate regime. They develop a medium-scale three-region DSGE model and show that the impact of a cut in the employers’ social security contributions rate does not vary significantly under different exchange rate regimes. They find that both the interest rate and the exchange rate channel matters. Furthermore, the measure appears to be effective even if it comes along with a consumption tax increase to preserve long-term fiscal sustainability.
Finally, they assess whether obtained theoretical results hold up empirically by applying the local projection method. Regression results suggest that changes in employers’ social security contributions rates have statistically significant real effects – a one percentage point reduction leads to an average cumulative rise in output of around 1.3 percent in the medium term. Moreover, the outcome does not differ significantly across the different exchange rate regimes.
This paper provides an overview of how to use "big data" for economic research. We investigate the performance and ease of use of different Spark applications running on a distributed file system to enable the handling and analysis of data sets which were previously not usable due to their size. More specifically, we explain how to use Spark to (i) explore big data sets which exceed retail grade computers memory size and (ii) run typical econometric tasks including microeconometric, panel data and time series regression models which are prohibitively expensive to evaluate on stand-alone machines. By bridging the gap between the abstract concept of Spark and ready-to-use examples which can easily be altered to suite the researchers need, we provide economists and social scientists more generally with the theory and practice to handle the ever growing datasets available. The ease of reproducing the examples in this paper makes this guide a useful reference for researchers with a limited background in data handling and distributed computing.
It has been forty years since the oil crisis of 1973/74. This crisis has been one of the defining economic events of the 1970s and has shaped how many economists think about oil price shocks. In recent years, a large literature on the economic determinants of oil price fluctuations has emerged. Drawing on this literature, we first provide an overview of the causes of all major oil price fluctuations between 1973 and 2014. We then discuss why oil price fluctuations remain difficult to predict, despite economists’ improved understanding of oil markets. Unexpected oil price fluctuations are commonly referred to as oil price shocks. We document that, in practice, consumers, policymakers, financial market participants and economists may have different oil price expectations, and that, what may be surprising to some, need not be equally surprising to others.
Some observers have conjectured that oil supply shocks in the United States and in other countries are behind the plunge in the price of oil since June 2014. Others have suggested that a major shock to oil price expectations occurred when in late November 2014 OPEC announced that it would maintain current production levels despite the steady increase in non-OPEC oil production. Both conjectures are perfectly reasonable ex ante, yet we provide quantitative evidence that neither explanation appears supported by the data. We show that more than half of the decline in the price of oil was predictable in real time as of June 2014 and therefore must have reflected the cumulative effects of earlier oil demand and supply shocks. Among the shocks that occurred after June 2014, the most influential shock resembles a negative shock to the demand for oil associated with a weakening economy in December 2014. In contrast, there is no evidence of any large positive oil supply shocks between June and December. We conclude that the difference in the evolution of the price of oil, which declined by 44% over this period, compared with other commodity prices, which on average only declined by about 5%-15%, reflects oil-market specific developments that took place prior to June 2014.
Although there is much interest in the future retail price of gasoline among consumers, industry analysts, and policymakers, it is widely believed that changes in the price of gasoline are essentially unforecastable given publicly available information. We explore a range of new forecasting approaches for the retail price of gasoline and compare their accuracy with the no-change forecast. Our key finding is that substantial reductions in the mean-squared prediction error (MSPE) of gasoline price forecasts are feasible in real time at horizons up to two years, as are substantial increases in directional accuracy. The most accurate individual model is a VAR(1) model for real retail gasoline and Brent crude oil prices. Even greater reductions in MSPEs are possible by constructing a pooled forecast that assigns equal weight to five of the most successful forecasting models. Pooled forecasts have lower MSPE than the EIA gasoline price forecasts and the gasoline price expectations in the Michigan Survey of Consumers. We also show that as much as 39% of the decline in gas prices between June and December 2014 was predictable.
he predictive likelihood is of particular relevance in a Bayesian setting when the purpose is to rank models in a forecast comparison exercise. This paper discusses how the predictive likelihood can be estimated for any subset of the observable variables in linear Gaussian state-space models with Bayesian methods, and proposes to utilize a missing observations consistent Kalman filter in the process of achieving this objective. As an empirical application, we analyze euro area data and compare the density forecast performance of a DSGE model to DSGE-VARs and reduced-form linear Gaussian models.
Futures markets are a potentially valuable source of information about market expectations. Exploiting this information has proved difficult in practice, because the presence of a time-varying risk premium often renders the futures price a poor measure of the market expectation of the price of the underlying asset. Even though the expectation in principle may be recovered by adjusting the futures price by the estimated risk premium, a common problem in applied work is that there are as many measures of market expectations as there are estimates of the risk premium. We propose a general solution to this problem that allows us to uniquely pin down the best possible estimate of the market expectation for any set of risk premium estimates. We illustrate this approach by solving the long-standing problem of how to recover the market expectation of the price of crude oil. We provide a new measure of oil price expectations that is considerably more accurate than the alternatives and more economically plausible. We discuss implications of our analysis for the estimation of economic models of energy-intensive durables, for the debate on speculation in oil markets, and for oil price forecasting.
This paper investigates the accuracy of point and density forecasts of four DSGE models for inflation, output growth and the federal funds rate. Model parameters are estimated and forecasts are derived successively from historical U.S. data vintages synchronized with the Fed’s Greenbook projections. Point forecasts of some models are of similar accuracy as the forecasts of nonstructural large dataset methods. Despite their common underlying New Keynesian modeling philosophy, forecasts of different DSGE models turn out to be quite distinct. Weighted forecasts are more precise than forecasts from individual models. The accuracy of a simple average of DSGE model forecasts is comparable to Greenbook projections for medium term horizons. Comparing density forecasts of DSGE models with the actual distribution of observations shows that the models overestimate uncertainty around point forecasts.
This paper investigates the accuracy of forecasts from four DSGE models for inflation, output growth and the federal funds rate using a real-time dataset synchronized with the Fed’s Greenbook projections. Conditioning the model forecasts on the Greenbook nowcasts leads to forecasts that are as accurate as the Greenbook projections for output growth and the federal funds rate. Only for inflation the model forecasts are dominated by the Greenbook projections. A comparison with forecasts from Bayesian VARs shows that the economic structure of the DSGE models which is useful for the interpretation of forecasts does not lower the accuracy of forecasts. Combining forecasts of several DSGE models increases precision in comparison to individual model forecasts. Comparing density forecasts with the actual distribution of observations shows that DSGE models overestimate uncertainty around point forecasts.
The substantial variation in the real price of oil since 2003 has renewed interest in the question of how to forecast monthly and quarterly oil prices. There also has been increased interest in the link between financial markets and oil markets, including the question of whether financial market information helps forecast the real price of oil in physical markets. An obvious advantage of financial data in forecasting oil prices is their availability in real time on a daily or weekly basis. We investigate whether mixed-frequency models may be used to take advantage of these rich data sets. We show that, among a range of alternative high-frequency predictors, especially changes in U.S. crude oil inventories produce substantial and statistically significant real-time improvements in forecast accuracy. The preferred MIDAS model reduces the MSPE by as much as 16 percent compared with the no-change forecast and has statistically significant directional accuracy as high as 82 percent. This MIDAS forecast also is more accurate than a mixed-frequency realtime VAR forecast, but not systematically more accurate than the corresponding forecast based on monthly inventories. We conclude that typically not much is lost by ignoring high-frequency financial data in forecasting the monthly real price of oil.
The U.S. Energy Information Administration (EIA) regularly publishes monthly and quarterly forecasts of the price of crude oil for horizons up to two years, which are widely used by practitioners. Traditionally, such out-of-sample forecasts have been largely judgmental, making them difficult to replicate and justify. An alternative is the use of real-time econometric oil price forecasting models. We investigate the merits of constructing combinations of six such models. Forecast combinations have received little attention in the oil price forecasting literature to date. We demonstrate that over the last 20 years suitably constructed real-time forecast combinations would have been systematically more accurate than the no-change forecast at horizons up to 6 quarters or 18 months. MSPE reduction may be as high as 12% and directional accuracy as high as 72%. The gains in accuracy are robust over time. In contrast, the EIA oil price forecasts not only tend to be less accurate than no-change forecasts, but are much less accurate than our preferred forecast combination. Moreover, including EIA forecasts in the forecast combination systematically lowers the accuracy of the combination forecast. We conclude that suitably constructed forecast combinations should replace traditional judgmental forecasts of the price of oil.
Are product spreads useful for forecasting? An empirical evaluation of the Verleger hypothesis
(2013)
Notwithstanding a resurgence in research on out-of-sample forecasts of the price of oil in recent years, there is one important approach to forecasting the real price of oil which has not been studied systematically to date. This approach is based on the premise that demand for crude oil derives from the demand for refined products such as gasoline or heating oil. Oil industry analysts such as Philip Verleger and financial analysts widely believe that there is predictive power in the product spread, defined as the difference between suitably weighted refined product market prices and the price of crude oil. Our objective is to evaluate this proposition. We derive from first principles a number of alternative forecasting model specifications involving product spreads and compare these models to the no-change forecast of the real price of oil. We show that not all product spread models are useful for out-of-sample forecasting, but some models are, even at horizons between one and two years. The most accurate model is a time-varying parameter model of gasoline and heating oil spot spreads that allows the marginal product market to change over time. We document MSPE reductions as high as 20% and directional accuracy as high as 63% at the two-year horizon, making product spread models a good complement to forecasting models based on economic fundamentals, which work best at short horizons.
Credit boom detection methodologies (such as threshold method) lack robustness as they are based on univariate detrending analysis and resort to ratios of credit to real activity. I propose a quantitative indicator to detect atypical behavior of credit from a multivariate system - a monetary VAR. This methodology explicitly accounts for endogenous interactions between credit, asset prices and real activity and detects atypical credit expansions and contractions in the Euro Area, Japan and the U.S. robustly and timely. The analysis also proves useful in real time.