nep-for New Economics Papers
on Forecasting
Issue of 2013‒04‒13
nineteen papers chosen by
Rob J Hyndman
Monash University

  1. The ins and outs of forecasting unemployment: Using labor force flows to forecast the labor market By Regis Barnichon; Christopher J. Nekarda
  2. Combining disaggregate forecasts for inflation: The SNB's ARIMA model By Marco Huwiler; Daniel Kaufmann
  3. Real-Time Nowcasting with a Bayesian Mixed Frequency Model with Stochastic Volatility By Carriero, Andrea; Clark, Todd; Marcellino, Massimiliano
  4. Short-term GDP forecasting with a mixed frequency dynamic factor model with stochastic volatility By Marcellino, Massimiliano; Porqueddu, Mario; Venditti, Fabrizio
  5. Forecasting Stock Returns under Economic Constraints By Pettenuzzo, Davide; Timmermann, Allan G; Valkanov, Rossen
  6. Short-Run Forecasting of Argentine GDP Growth By Maximo Camacho; Marcos Dal Bianco; Jaime Martinez-Martin
  7. Are Forecast Updates Progressive? By Chia-Lin Chang; Philip Hans Franses; Michael McAleer
  8. Forecasting U.S. Recessions with Macro Factors By Fossati, Sebastian
  9. On the Predictive Power of Yield Spread for Future Growth and Recession: The Turkish Case? By Huseyin Kaya
  10. Do Dynamic Neural Networks Stand a Better Chance in Fractionally Integrated Process Forecasting? By Delavari, Majid; Gandali Alikhani, Nadiya; Naderi, Esmaeil
  11. Macroeconomic forecasting during the Great Recession: The return of non-linearity? By Ferrara, Laurent; Marcellino, Massimiliano; Mogliani, Matteo
  12. Financial Time Series Forecasting by Developing a Hybrid Intelligent System By Abounoori, Abbas Ali; Naderi, Esmaeil; Gandali Alikhani, Nadiya; Amiri, Ashkan
  13. Imputing Individual Effects in Dynamic Microsimulation Models. An application of the Rank Method By Richiardi Matteo; Poggi Ambra
  14. Can Google Data Help Predict French Youth Unemployment? By Frédéric Karamé; Yannick Fondeur
  15. Confidence Bands for ROC Curves with Serially Dependent Data By Kajal Lahiri; Liu Yang
  16. Truncated Product Methods for Panel Unit Root Tests By Xuguang Sheng; Jingyun Yang
  17. Distilling the Macroeconomic News Flow By Beber, Alessandro; Brandt, Michael; Luisi, Maurizio
  18. Forecasting Fines for Health and Safety in Employment Offences in New Zealand By Alan Woodfield; Stephen Hickson; Andrea Menclova
  19. The transformative effect of unscheduled generation by solar PV and wind generation on net electricity demand By Bell, William Paul; Wild, Phillip; Foster, John

  1. By: Regis Barnichon; Christopher J. Nekarda
    Abstract: This paper presents a forecasting model of unemployment based on labor force ows data that, in real time, dramatically outperforms the Survey of Professional Forecasters, historical forecasts from the Federal Reserve Board's Greenbook, and basic time-series models. Our model's forecast has a root-mean-squared error about 30 percent below that of the next-best forecast in the near term and performs especially well surrounding large recessions and cyclical turning points. Further, because our model uses information on labor force ows that is likely not incorporated by other forecasts, a combined forecast including our model's forecast and the SPF forecast yields an improvement over the latter alone of about 35 percent for current-quarter forecasts, and 15 percent for next-quarter forecasts, as well as improvements at longer horizons.
    Date: 2013
  2. By: Marco Huwiler; Daniel Kaufmann
    Abstract: This study documents the SNB's ARIMA model based on disaggregated CPI data used to produce inflation forecasts over the short-term horizon, and evaluates its forecasting performance. Our findings suggest that the disaggregate ARIMA model for the Swiss CPI performed better than relevant benchmarks. In particular, estimating ARIMA models for individual CPI expenditure items and aggregating the forecasts from these models gives better results than directly applying the ARIMA methodto the total CPI. We then extend the model to factor in changes in the collection frequency of the Swiss CPI data and show that this extension further improves the forecasting performance.
    Keywords: Swiss CPI inflation, Forecast combination, Forecast aggregation, Disaggregateinformation, ARIMA models, Missing data, Kalman filter
    JEL: C22 C52 C53 E37
    Date: 2013
  3. By: Carriero, Andrea; Clark, Todd; Marcellino, Massimiliano
    Abstract: This paper develops a method for producing current-quarter forecasts of GDP growth with a (possibly large) range of available within-the-quarter monthly observations of economic indicators, such as employment and industrial production, and financial indicators, such as stock prices and interest rates. In light of existing evidence of time variation in the variances of shocks to GDP, we consider versions of the model with both constant variances and stochastic volatility. We also evaluate models with either constant or time-varying regression coefficients. We use Bayesian methods to estimate the model, in order to facilitate providing shrinkage on the (possibly large) set of model parameters and conveniently generate predictive densities. We provide results on the accuracy of nowcasts of real-time GDP growth in the U.S. from 1985 through 2011. In terms of point forecasts, our proposal is comparable to alternative econometric methods and survey forecasts. In addition, it provides reliable density forecasts, for which the stochastic volatility specification is quite useful, while parameter time-variation does not seem to matter.
    Keywords: Bayesian methods; forecasting; mixed frequency models; Prediction
    JEL: C22 C53 E37
    Date: 2013–01
  4. By: Marcellino, Massimiliano; Porqueddu, Mario; Venditti, Fabrizio
    Abstract: In this paper we develop a mixed frequency dynamic factor model featuring stochastic shifts in the volatility of both the latent common factor and the idiosyncratic components. We take a Bayesian perspective and derive a Gibbs sampler to obtain the posterior density of the model parameters. This new tool is then used to investigate business cycle dynamics and for forecasting GDP growth at short-term horizons in the euro area. We discuss three sets of empirical results. First we use the model to evaluate the impact of macroeconomic releases on point and density forecast accuracy and on the width of forecast intervals. Second, we show how our setup allows to make a probabilistic assessment of the contribution of releases to forecast revisions. Third we design a pseudo out of sample forecasting exercise and examine point and density forecast accuracy. In line with findings in the Bayesian Vector Autoregressions (BVAR) literature we find that stochastic volatility contributes to an improvement in density forecast accuracy.
    Keywords: Business cycle; Forecasting; Mixed-frequency data; Nonlinear models; Nowcasting
    JEL: C22 E27 E32
    Date: 2013–02
  5. By: Pettenuzzo, Davide; Timmermann, Allan G; Valkanov, Rossen
    Abstract: We propose a new approach to imposing economic constraints on time-series forecasts of the equity premium. Economic constraints are used to modify the posterior distribution of the parameters of the predictive return regression in a way that better allows the model to learn from the data. We consider two types of constraints: Non-negative equity premia and bounds on the conditional Sharpe ratio, the latter of which incorporates timevarying volatility in the predictive regression framework. Empirically, we find that economic constraints systematically reduce uncertainty about model parameters, reduce the risk of selecting a poor forecasting model, and improve both statistical and economic measures of out-of-sample forecast performance. The Sharpe ratio constraint, in particular, results in considerable economic gains.
    Keywords: Bayesian analysis; Economic constraints; Sharpe Ratio; Stock return predictability
    JEL: C11 C22 G11 G12
    Date: 2013–03
  6. By: Maximo Camacho; Marcos Dal Bianco; Jaime Martinez-Martin
    Abstract: In this paper, we propose a small-scale dynamic factor model for monitoring Argentine GDP in real time using economic data at mixed frequencies (monthly and quarterly). Our model not only produces a coincident index of the Argentine business cycle in striking accordance with professional consensus and the history of the Argentine business cycle, but also generates accurate short-run forecasts of Argentine GDP growth. By using a simulated real-time empirical evaluation, we are able to demonstrate that our model produces reliable backcasts, nowcasts and forecasts well before the official data is released.
    Keywords: Real-time forecasting, Argentine GDP, business cycles, state-space models, mixed frequencies
    JEL: C22 C53 E27 E32 E37
    Date: 2013–04
  7. By: Chia-Lin Chang (National Chung Hsing University Taichung); Philip Hans Franses (Erasmus University Rotterdam); Michael McAleer (Erasmus University Rotterdam, Complutense University of Madrid, Kyoto University)
    Abstract: Many macroeconomic forecasts and forecast updates like those from IMF and OECD typically involve both a model component, which is replicable, as well as intuition, which is non-replicable. Intuition is expert knowledge possessed by a forecaster. If forecast updates are progressive, forecast updates should become more accurate, on average, as the actual value is approached. Otherwise, forecast updates would be neutral. The paper proposes a methodology to test whether macroeconomic forecast updates are progressive, where the interaction between model and intuition is explicitly taken into account. The data set for the empirical analysis is for Taiwan, where we have three decades of quarterly data available of forecasts and their updates of the inflation rate and real GDP growth rate. Our empirical results suggest that the forecast updates for Taiwan are progressive, and that progress can be explained predominantly by improved intuition.
    Keywords: Macroeconomic forecasts; econometric models; intuition; progressive forecast updates; forecast errors
    JEL: C53 C22 E27 E37
    Date: 2013–03–25
  8. By: Fossati, Sebastian (University of Alberta, Department of Economics)
    Abstract: Dynamic factors estimated from panels of macroeconomic indicators are used to predict future recessions using probit models. Three factors are considered: a bond and exchange rates factor; a stock market factor; a real activity factor. Three results emerge. First, models that use only financial indicators exhibit a large deterioration in fit after 2005. Second, models that use factors yield better fit than models that use indicators directly. Out-of-sample forecasting exercises confirm these results for 3-, 6-, and 12-month horizons using both ex-post revised data and real-time data. Third, results show evidence that data revisions affect factors less than individual indicators.
    Keywords: recession; forecasting; factors; probit model
    JEL: C22 C25 E32
    Date: 2013–04–03
  9. By: Huseyin Kaya (Bahcesehir University; Bahcesehir University Faculty of Economics and Administrative Sciences)
    Abstract: This paper investigates the predictive power of the yield spread on future industrial production growth and recession in Turkey. Employing the linear regression model we find that the yield spread has predictive power when forecasting industrial production growth. The results also suggest that in the inflation targeting monetary policy period, predictive power of the yield spread has increased. Furthermore, we investigate whether the yield spread predicts recession by employing a probit model. Since no official recessions are available in Turkey, we determine the recessions using the BBQ methodology. The findings suggest that the yield spread predicts the recessions about one year ahead.
    Keywords: yield curve, recession, growth, forecast
    JEL: C22 E37 E43
    Date: 2013–03
  10. By: Delavari, Majid; Gandali Alikhani, Nadiya; Naderi, Esmaeil
    Abstract: The main purpose of the present study was to investigate the capabilities of two generations of models such as those based on dynamic neural network (e.g., Nonlinear Neural network Auto Regressive or NNAR model) and a regressive (Auto Regressive Fractionally Integrated Moving Average model which is based on Fractional Integration Approach) in forecasting daily data related to the return index of Tehran Stock Exchange (TSE). In order to compare these models under similar conditions, Mean Square Error (MSE) and also Root Mean Square Error (RMSE) were selected as criteria for the models’ simulated out-of-sample forecasting performance. Besides, fractal markets hypothesis was examined and according to the findings, fractal structure was confirmed to exist in the time series under investigation. Another finding of the study was that dynamic artificial neural network model had the best performance in out-of-sample forecasting based on the criteria introduced for calculating forecasting error in comparison with the ARFIMA model.
    Keywords: Stock Return, Forecasting, Long Memory, NNAR, ARFIMA
    JEL: C14 C22 C45 C53
    Date: 2012–09–11
  11. By: Ferrara, Laurent; Marcellino, Massimiliano; Mogliani, Matteo
    Abstract: The debate on the forecasting ability of non-linear models has a long history, and the Great Recession episode provides us with an interesting opportunity for a reassessment of the forecasting performance of several classes of non-linear models. We conduct an extensive analysis over a large quarterly database consisting of major macroeconomic variables for a large panel of countries. It turns out that, on average, non-linear models cannot outperform standard linear specifications, even during the Great Recession. However, non-linear models lead to an improvement of the predictive accuracy in almost 40% of cases, and interesting specific patterns emerge among models, variables and countries. These results suggest that this specific episode seems to be characterized by a sequence of shocks with unusual large magnitude, rather than by an increase in the degree of non-linearity of the stochastic processes underlying the main macroeconomic time series.
    Keywords: Great Recession; Macroeconomic forecasting; Non-linear models
    JEL: C22 C53 E37
    Date: 2013–01
  12. By: Abounoori, Abbas Ali; Naderi, Esmaeil; Gandali Alikhani, Nadiya; Amiri, Ashkan
    Abstract: The design of models for time series forecasting has found a solid foundation on statistics and mathematics. On this basis, in recent years, using intelligence-based techniques for forecasting has proved to be extremely successful and also is an appropriate choice as approximators to model and forecast time series, but designing a neural network model which provides a desirable forecasting is the main concern of researchers. For this purpose, the present study tries to examine the capabilities of two sets of models, i.e., those based on artificial intelligence and regressive models. In addition, fractal markets hypothesis investigates in daily data of the Tehran Stock Exchange (TSE) index. Finally, in order to introduce a complete design of a neural network for modeling and forecasting of stock return series, the long memory feature and dynamic neural network model were combined. Our results showed that fractal markets hypothesis was confirmed in TSE; therefore, it can be concluded that the fractal structure exists in the return of the TSE series. The results further indicate that although dynamic artificial neural network model have a stronger performance compared to ARFIMA model, taking into consideration the inherent features of a market and combining it with neural network models can yield much better results.
    Keywords: Stock Return, Long Memory, NNAR, ARFIMA, Hybrid Models
    JEL: C22 C45 C53 G10
    Date: 2013–01–17
  13. By: Richiardi Matteo; Poggi Ambra (University of Turin)
    Abstract: Dynamic microsimulation modeling involves two stages: estimation and forecasting. Unobserved heterogeneity is often considered in estimation, but not in forecasting,beyond trivial cases. Non-trivial cases involve individuals that enter the simulation with a history of previous outcomes. We show that the simple solutions of attributing to these individuals a null effect or a random draw from the estimat ed unconditional distributions lead to biased forecasts, which are often worse than those obtained neglect ing unobserved heterogeneity altogether. We then present a first implementation of the Rank method, a new algorithm for attributing the individual effects to the simulation sample which greatly simplifies those already known in the literature. Out - of - sample validation of our model shows that correctly imputing unobserved heterogeneity significantly improves the quality of the forecasts.
    Date: 2012–09
  14. By: Frédéric Karamé (EPEE (Université d’Evry-Val-d’Essonne), TEPP (FR CNRS n°3126), DYNARE Team (Cepremap) and Centre d’Etudes de l‘Emploi); Yannick Fondeur (Centre d’Etudes de l’Emploi)
    Abstract: According to the rising “Google econometrics” literature, Google queries may help predict economic activity. The aim of our paper is to test if these data can enhance predictions for youth unemployment in France. As we have on the one hand weekly series on web search queries and on the other hand monthly series on unemployment for the 15 to 24-year-olds, we use the unobserved components approach in order to exploit all available information. Our model is estimated with a modified version of the Kalman filter taking into account the twofold issues of non-stationarity and multiple frequencies in our data. We find that including Google data improves unemployment predictions relatively to a competing model without search data queries.
    Keywords: Google econometrics, forecasting, nowcasting, unemployment, unobserved components, diffuse initialization, Kalman filter, univariate treatment of time series, smoothing, multivariate models
    JEL: C22 C51 E32 E37
    Date: 2012
  15. By: Kajal Lahiri; Liu Yang
    Abstract: We propose serial correlation robust asymptotic confidence bands for the receiver operating characteristic (ROC) curves estimated by quasi-maximum likelihood in the binormal model. Our simulation experiments confirm that this new method performs fairly well in finite samples. The conventional procedure is found to be markedly undersized in terms of yielding empirical coverage probabilities lower than the nominal level, especially when the serial correlation is strong. We evaluate the three-quarter-ahead probability forecasts for real GDP declines from the Survey of Professional Forecasters, and find that one would draw a misleading conclusion about forecasting skill if serial correlation is ignored.
    Date: 2013
  16. By: Xuguang Sheng (American University); Jingyun Yang (Pennsylvania State University)
    Abstract: This paper proposes three new panel unit root tests based on Zaykin et al. (2002)’s truncated product method. The first one assumes constant correlation between p-values and the latter two use sieve bootstrap that allows for general forms of cross-section dependence in the panel units. Monte Carlo simulation shows that these tests have reasonably good size, are robust to varying degrees of cross-section dependence and are powerful in cases where there are some very large p-values. The proposed tests are applied to a panel of real GDP and inflation density forecasts and provide evidence that professional forecasters may not update their forecast precision in an optimal Bayesian way.
    Keywords: Density Forecast, Panel Unit Root, P-value, Sieve Bootstrap, Truncated Product Method
    JEL: C12 C33
    Date: 2013–04
  17. By: Beber, Alessandro; Brandt, Michael; Luisi, Maurizio
    Abstract: We propose a simple cross-sectional technique to extract daily latent factors from economic news releases available at different dates and frequencies. Our approach can effectively handle the large number of heterogeneous announcements that are relevant for tracking current economic conditions. We apply the technique to extract real-time measures of inflation, output, employment, and macroeconomic sentiment, as well as corresponding measures of disagreement among economists about these dimensions of the data. We find that our procedure provides more timely and accurate forecasts of the future evolution of the economy than other real-time forecasting approaches in the literature.
    Keywords: disagreement.; macroeconomic news; nowcasting
    JEL: G12
    Date: 2013–02
  18. By: Alan Woodfield (University of Canterbury); Stephen Hickson (University of Canterbury); Andrea Menclova (University of Canterbury)
    Abstract: Sentences for employers convicted of offences under NZ health and safety law have been subject to constraints from two main sources (i) legislation; and (ii) guideline judgment cases. Their effect is to effectively split sentencing into three distinct time periods, viz., the period following the introduction of the De Spa Guidelines to the implementation of the Sentencing Act 2002, the second following the joint implementation of the Sentencing Act and the Health and Safety in Employment Amendment Act to the Hanham & Philp Guideline judgment in December 2008, and the third is the post Hanham & Philp Guideline period. This article builds on previous work that analyses the various factors relevant to HSE sentencing, concentrating on the second and third periods. Among other results, this work shows that for period 3, although harm continues to play an important role in explaining sentences of reparation, its previous role in directly explaining levels of fines is replaced by various levels of employer culpability. The Hanham & Philp decisions incorporated harm in determining culpability and District Court judges appear to follow this judgment closely in this respect. The present article illustrates forecasted sentences for periods 2 and 3, and, for the forecasts of period 3 penalties using second period weights, finds that fines would have been frequently lower, often substantially so, than those that occurred, consistent with the Hanham & Philp Guidelines. Reparations, however, are largely unaffected.
    Keywords: Health & Safety Offences; Judicial Guidelines; Forecasting Fines
    JEL: K32
    Date: 2013–03–22
  19. By: Bell, William Paul; Wild, Phillip; Foster, John
    Abstract: This study investigates the transformative effect of unscheduled solar PV and wind generation on electricity demand. The motivations for the study are twofold, the poor medium term predictions of electricity demand in the Australian National Electricity Market and the continued rise in peak demand but reduction in overall demand. A number of factors contribute to these poor predictions, including the global financial crisis inducing a reduction in business activity, the Australian economy’s continued switch from industrial to service sector, the promotion of energy conservation, and particularly mild weather reducing the requirement for air conditioning. Additionally, there is growing unscheduled generation, which is meeting electricity demand. This growing source of generation necessitates the concepts of gross and net demand where gross demand is met by unscheduled and scheduled generation and net demand by scheduled generation. The methodology compares the difference between net and gross demand of the 50 nodes in the Australian National Electricity Market using half hourly data from 2007 to 2011. The unscheduled generation is calculated using the Australian Bureau of Meteorology half hourly solar intensity and wind speed data and the Australian Clean Energy Regulator’s database of small generation units’ renewable energy target certificates by postcode. The findings are that gross demand rather than net demand helps explain both the overall reduction in net demand and the continued increase in peak demand. The study has two main conclusions. Firstly, a requirement for policy to target the growth in peak demand via time of supply feed-in tariff for small generation units. Secondly, modellers of electricity demand consider both net and gross demand in their forecasts. The time of supply feed-in tariffs are intended to promote the adoption of storage technologies and demand side participation and management. Modellers considering both net and gross demand are required to model unscheduled generation. This requirement ensues that more comprehensive solar intensity data be provided by the Bureau of Meteorology and that the Australian National Electricity Market Operator provide data in GIS format of each demand node using the Australian Statistical Geography Standard developed by Australian Bureau of Statistics to enable easier integration of large quantities of geographic data from a number of sources. The applicability of these finding become more relevant to other countries as unscheduled generation becomes more wide spread. This study is instrumental to a range of further research. Other sources of unscheduled generations should be considered to form a more comprehensive concept of gross demand, for instance, solar hot water and small hydro. Replacing electrical hot water heaters with solar hot water reduces the overnight demand, which may provide a considerable transformative effect on net electricity demand. In addition, energy efficiency is meeting demand for electricity; incorporating energy efficiency would form an even more comprehensive concept of gross electricity demand and could help improve longer term electricity demand projections.
    Keywords: solar PV, wind generation, electricity demand, AEMO, electricity demand forecasting, renewable energy, transmission, climate change adaptation, Feed-in tariffs; non-scheduled generation; FiT; residential solar PV; Sustainable; DUOS; TUOS; smart meters
    JEL: O13 O3 O31 O33 Q01 Q2 Q28 Q31 Q4 R22 R38
    Date: 2013–04–03

This nep-for issue is ©2013 by Rob J Hyndman. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.