nep-for New Economics Papers
on Forecasting
Issue of 2012‒04‒23
ten papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting house price inflation: a model combination approach By Sarah Drought; Chris McDonald
  2. Forecasting House Prices for the Four Census Regions and the Aggregate US Economy: The Role of a Data-Rich Environment By Rangan Gupta
  3. Is there an optimal forecast combination? A stochastic dominance approach applied to the forecast combination puzzle. By Mehmet Pinar; Thanasis Stengos; M. Ege Yazgan
  4. "Forecasting Welfare Caseloads: The Case of the Japanese Public Assistance Program" By Masayoshi Hayashi
  5. Bayesian logistic betting strategy against probability forecasting By Masayuki Kumon; Jing Li; Akimichi Takemura; Kei Takeuchi
  6. Exchange rate expectations of chartists and fundamentalists By Dick, Christian D.; Menkhoff, Lukas
  7. Can Oil Prices Forecast Exchange Rates? By Domenico Ferraro; Kenneth S. Rogoff; Barbara Rossi
  8. National and Provincial Inflation in Canada: Experiences under Inflation Targeting By Graham M. Voss; M. Chaban
  9. The Predictive Space, or, If x predicts y, what does y tell us about x? By Donald Robertson; Stephen Wright
  10. Predicting Financial Crises: The (Statistical) Significance of the Signals Approach By Makram El-Shagi; Tobias Knedlik; Gregor von Schweinitz

  1. By: Sarah Drought; Chris McDonald (Reserve Bank of New Zealand)
    Abstract: In this paper we use a range of statistical models to forecast New Zealand house price in ation. We address the issue of model uncertainty by combining forecasts using weights based on out-of-sample forecast performance. We consider how the combined forecast for house prices performs relative to both the individual model forecasts and the Reserve Bank of New Zealand's house price forecasts. We find that the combination forecast is on par with the best of the models for most forecast horizons, and has produced lower root mean squared forecast errors than the Reserve Bank's forecasts.
    JEL: E17 E37
    Date: 2011–11
  2. By: Rangan Gupta (Department of Economics, University of Pretoria)
    Abstract: This paper considers the ability of large-scale (involving 145 fundamental variables) time-series models, estimated based on dynamic factor analysis and Bayesian shrinkage, to forecast real house price growth rates of the four US census regions and the aggregate US economy. Besides, the standard Minnesota prior, we also use additional priors that constrain the sum of coefficients of the VAR models. We compare one- to twenty four-months-ahead forecasts of the large-scale models over an out-of-sample horizon of 1995:1-2009:3, based on an insample of 1968:2-1994:12, relative to a random walk model and a small-scale VAR model comprising of just the five real house price growth rates. In addition to the forecast comparison exercise across large- and small-scale models, we also look at the ability of the “optimal” model (i.e., the model that produces the minimum average mean squared forecast error (MSFE)) for a specific region, in predicting ex ante real house prices (in levels) over the period of 2009:4 till 2012:2. Factor-based models (classical or Bayesian) performs the best for the North East, Mid- West, West census regions and the aggregate US economy, and equally as well to a small-scale VAR for the South region. The “optimal” factor models also tend to predict the downward trend in the data when we conduct an ex ante forecasting exercise. Our results highlight the importance of information content in large number of fundamentals in predicting house prices accurately.
    Keywords: House prices, Forecasting, Factor Augmented Models, Large-Scale, BVAR models
    JEL: C32 R31
    Date: 2012–04
  3. By: Mehmet Pinar (Fondazione Eni Enrico Mattei); Thanasis Stengos (University of Guelph.); M. Ege Yazgan (Istanbul Bilgi University)
    Abstract: The forecast combination puzzle refers to the finding that a simple average forecast combination outperforms more sophisticated weighting schemes and/or the best individual model. The paper derives optimal (worst) forecast combinations based on stochastic dominance (SD) analysis with differential forecast weights. For the optimal (worst) forecast combination, this index will minimize (maximize) forecasts errors by combining time-series model based forecasts at a given probability level. By weighting each forecast differently, we find the optimal (worst) forecast combination that does not rely on arbitrary weights. Using two exchange rate series on weekly data for the Japanese Yen/U.S. Dollar and U.S. Dollar/Great Britain Pound for the period from 1975 to 2010 we find that the simple average forecast combination is neither the worst nor the best forecast combination something that provides partial support for the forecast combination puzzle. In that context, the random walk model is the model that consistently contributes with considerably more than an equal weight to the worst forecast combination for all variables being forecasted and for all forecast horizons, whereas a flexible Neural Network autoregressive model and a self-exciting threshold autoregressive model always enter the best forecast combination with much greater than equal weights.
    Keywords: Nonparametric Stochastic Dominance, Mixed Integer Programming; Forecast combinations; Forecast combination
    JEL: C53 C61 C63
    Date: 2011
  4. By: Masayoshi Hayashi (Faculty of Economics, University of Tokyo)
    Abstract: Forecasting welfare caseloads has become more important than ever in Japan. One reason is the magnitude of the recent increase in its welfare caseloads. Given that most previous studies only concern US cases and have not exploited recent developments in the literature, this study employs several methods (exponential smoothing, ARIMA, LSTAR, VAR, and a set of forecast combinations) to forecast Japanese welfare caseloads and compare their performances. While a VAR model and a forecast-combination model tend to outperform the other methods in pseudo real-time forecasting, a simple average forecast-combination method appears to outperform the other methods in real-time forecast ing. In particular, the method predicts that PA caseload in Japan would surpass 1.7 million by the beginning of 2016, an approximately 20% increase from that at the beginning of 2011. </table>
    Date: 2012–04
  5. By: Masayuki Kumon; Jing Li; Akimichi Takemura; Kei Takeuchi
    Abstract: We propose a betting strategy based on Bayesian logistic regression modeling for the probability forecasting game in the framework of game-theoretic probability by Shafer and Vovk (2001). We prove some results concerning the strong law of large numbers in the probability forecasting game with side information based on our strategy. We also apply our strategy for assessing the quality of probability forecasting by the Japan Meteorological Agency. We find that our strategy beats the agency by exploiting its tendency of avoiding clear-cut forecasts.
    Date: 2012–04
  6. By: Dick, Christian D.; Menkhoff, Lukas
    Abstract: This paper provides novel evidence on exchange rate expectations of both chartists and fundamentalists separately. These groups indeed form expectations differently. Chartists change their expectations more often; however, all professionals' expectations vary considerably as they generally follow strong exchange rate trends. In line with non-linear exchange rate-modeling, professionals expect mean reversion only if exchange rates deviate much from PPP. Chartists survive in FX markets as they forecast equally accurately as fundamentalists. Unexpected from an efficient market viewpoint, chartists even outperform fundamentalists at short horizons. Overall, these findings clearly support the chartist-fundamentalist approach. --
    Keywords: exchange rate formation,expectation formation,heterogeneous agent models,forecasting performance
    JEL: F31 G15 D84
    Date: 2012
  7. By: Domenico Ferraro; Kenneth S. Rogoff; Barbara Rossi
    Abstract: This paper investigates whether oil prices have a reliable and stable out-of-sample relationship with the Canadian/U.S dollar nominal exchange rate. Despite state-of-the-art methodologies, we find little systematic relation between oil prices and the exchange rate at the monthly and quarterly frequencies. In contrast, the main contribution is to show the existence of a very short-term relationship at the daily frequency, which is rather robust and holds no matter whether we use contemporaneous (realized) or lagged oil prices in our regression. However, in the latter case the predictive ability is ephemeral, mostly appearing after instabilities have been appropriately taken into account
    JEL: C22 C53 F31 F37
    Date: 2012–04
  8. By: Graham M. Voss (Department of Economics, University of Victoria); M. Chaban
    Abstract: We examine the behaviour of national and provincial inflation in Canada under inflation targeting to determine the extent to which the inflation targeting regime adopted by the Bank of Canada in the 1990s has anchored inflation expectations. Inflation expectations are well anchored when there are no predictable departures of inflation from target at sufficiently distant horizons. To examine this condition, we consider the out of sample prediction of monthly inflation with specific focus on whether deviations from the 1-3% target band are consistently predictable. We find support for well anchored inflation expectations at the national level and some but not all provinces.
    Keywords: Inflation, monetary policy, inflation targeting, forecasting
    JEL: E31 E58
    Date: 2012–04–13
  9. By: Donald Robertson (University of Cambridge); Stephen Wright (Department of Economics, Mathematics & Statistics, Birkbeck)
    Abstract: A predictive regression for y(t) and a time series representation of the predictors, x(t), together imply a univariate reduced form for y(t). In this paper we work backwards, and ask: if we observe y(t), what do its univariate properties tell us about any x(t) in the "predictive space" consistent with those properties? We provide a mathematical characterisation of the predictive space and certain of its derived properties. We derive both a lower and an upper bound for the R^2 for any predictive regression for y(t). We also show that for some empirically relevant univariate properties of y(t), the entire predictive space can be very tightly constrained. We illustrate using Stock and Watson's (2007) univariate representation of inflation.
    Date: 2012–04
  10. By: Makram El-Shagi; Tobias Knedlik; Gregor von Schweinitz
    Abstract: The signals approach as an early warning system has been fairly successful in detecting crises, but it has so far failed to gain popularity in the scientific community because it does not distinguish between randomly achieved in-sample fit and true predictive power. To overcome this obstacle, we test the null hypothesis of no correlation between indicators and crisis probability in three applications of the signals approach to different crisis types. To that end, we propose bootstraps specifically tailored to the characteristics of the respective datasets. We find (1) that previous applications of the signals approach yield economically meaningful and statistically significant results and (2) that composite indicators aggregating information contained in individual indicators add value to the signals approach, even where most individual indicators are not statistically significant on their own.
    Keywords: early warning system, signals approach, bootstrap
    JEL: C15 E60 F01
    Date: 2012–04

This nep-for issue is ©2012 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.