## 2003): “Forecast uncertainties in macroeconometric modelling: an application to the UK economy

Venue: | Journal of the American Statistical Association |

Citations: | 44 - 14 self |

### BibTeX

@ARTICLE{Garratt_2003):“forecast,

author = {Anthony Garratt and Kevin Lee and M. Hashem Pesaran},

title = {2003): “Forecast uncertainties in macroeconometric modelling: an application to the UK economy},

journal = {Journal of the American Statistical Association},

year = {},

pages = {829--838}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper argues that probability forecasts convey information on the uncertainties that surround macro-economic forecasts in a straightforward manner which is preferable to other alternatives, including the use of confidence intervals. Probability forecasts obtained using a small benchmark macroeconometric model as well as a number of other alternatives are presented and evaluated using recursive forecasts generated over the period 1999q1-2001q1. Out of sample probability forecasts of inflation and output growth are also provided over the period 2001q2-2003q1, and their implications discussed in relation to the Bank of England’s inflation target and the need to avoid recessions, both as separate events and jointly. The robustness of the results to parameter and model uncertainties is also investigated by a pragmatic implementation of the Bayesian model averaging approach.

### Citations

2321 | Estimating the dimension of a model - Schwarz - 1978 |

357 |
The Bootstrap and Edgeworth Expansion
- Hall
- 1992
(Show Context)
Citation Context ...imates for output growth are computed from the forecasts of per capita output, assuming a population growth of 0.22% per annum. Recession is said to have occured when output growth (measured, quarter =-=[21]-=-on quarter, by 400× ln (GDP T +h/GDP T +h−1 ) becomes negative in two consecutive quarters. Also see the notes to Table 3. Figure 1: Predictive Distribution Functions for Inflation Using the Benchmar... |

236 |
Evaluating density forecasts with applications to financial risk managment
- X, Gunther, et al.
- 1998
(Show Context)
Citation Context ...e overidentifying restrictions imposed on the long-run coefficients, and the specification of the oil price equation. Given the limited time series data available, consideration of models with p =3or =-=[10]-=-more did not seem advisable. We also thought it would not be worthwhile to consider p =1on the grounds that the resultant equations would most likely suffer from residual serial correlation. Therefor... |

223 |
Model Selection and Inference: A Practical InformationTheoretic Approach
- Burnham
- 1998
(Show Context)
Citation Context ...tstrap predictive density defined by (10) that makes use of the normal approximation given by (11). For details of computations when zt follows linear vector error correcting models see the Appendix. =-=[6]-=-2.2 Estimation and Forecasting with Conditional Models The density function fi(.) can be decomposed in two ways. First, a sequential conditioning decomposition can be employed to write fi(.) as the p... |

199 |
Combining Forecasts: A Review and Annotated Bibliography
- Clemen
- 1989
(Show Context)
Citation Context ...ods and capital markets, solvency and portfolio balance conditions. The model comprises six domestic variables whose developments are widely regarded as essential to a basic understanding of the U.K. =-=[7]-=-economy; namely, output, inflation, the exchange rate, the domestic relative to the foreign price level, the nominal interest rate and real money balances. It also contains three foreign variables: f... |

175 | Remarks on a multivariate transformation - Rosenblatt - 1952 |

113 | Assessment and propagation of model uncertainty
- Draper
- 1995
(Show Context)
Citation Context ...t for the analysis of macro-economic policy in the UK. Inflation targets 3 For details of the Kolmogorov-Smirnov test and its critical values see, for example, Neave and Worthington (1992, pp.89-93). =-=[12]-=-have been set explicitly in the UK since October 1992, following the UK’s exit from the European Exchange Rate Mechanism (ERM). The Chancellor’s stated objective at the time was to achieve an average... |

84 | Forecast Evaluation and Combination
- Diebold, Lopez
- 1996
(Show Context)
Citation Context ...CDF) of the probability integral transforms, and U(x) =x, is the CDF of iid U[0, 1]. Large values of the Kolmogorov-Smirnov statistics, KSin, is indicative of significant departures of the sample CDF =-=[11]-=-from the hypothesized uniform distribution. 3 The test results are summarized in Table 2. For the over-identified benchmark specification, we obtained the value of 0.111 for the KolmogorovSmirnov sta... |

84 | Structural Analysis of Vector Error Correction Models with Exogenous I(1) Variables. Cambridge Working Papers in Economics 9706 - Pesaran, Shin, et al. - 1997 |

81 | A Simple Nonparametric Test of Predictive Performance - Pesaran, Timmerman - 1992 |

61 |
Present position and potential developments: Some personal views on statistical theory the prequential approach (with discussion
- Dawid
- 1984
(Show Context)
Citation Context ...vectors (each row of β), or a total of twenty-five restrictions on β. These represent only a subset of the 2 The computations were carried out using Pesaran and Pesaran’s (1997) Microfit 4.1. i=1 ) ′ =-=[9]-=-restrictions suggested by economic theory as characterized in (23), however. Estimation of the model subject to all the (exact- and over-identifying) restrictions given in (23) enables a test of the ... |

61 | 2001a. “Model Uncertainty in Cross-Country Growth Regressions
- Fernandez, Ley, et al.
(Show Context)
Citation Context ...ility forecasts for AT +h∩CT +h are, of course, considerably less than those for Pr (AT +h | ΩT ) alone. Figure 5 plots the values of the joint event probability over the forecast horizon alongside a =-=[15]-=-plot of the product of the single event probabilities; that is Pr (AT +h | ΩT ) × Pr ( ) BT +h | ΩT , h =1, 2,..24. This comparison provides an indication of the degree of dependence/independence of ... |

60 |
Economic and Statistical Measures of Forecast Accuracy
- Granger, Pesaran
- 2000
(Show Context)
Citation Context ...ndom walk) 0.123 0.642 2.701 0.136 0.605 2.094 Equal Weights 0.062 0.630 2.346 0.111 0.630 2.322 Akaike Weights 0.160 0.642 2.701 0.136 0.630 2.451 Schwarz Weights 0.111 0.605 1.873 0.099 0.617 2.109 =-=[20]-=-Table 3 Single Events Probability Estimates for Inflation Forecast Pr(∆p <1.5%) Pr(∆p <2.5%) Pr(∆p <3.5%) Pr(1.5% < ∆p <3.5%)) Horizon π ˜π π ˜π π ˜π π ˜π 2001q2 0.199 0.133 0.975 0.923 1.000 1.000 0... |

43 |
The Inflation Report Projections: Understanding the Fan Chart
- Britton, PG, et al.
- 1998
(Show Context)
Citation Context ... ) dZT +1,hdθi. (10) In practice, computations of πi ( a,h; ϕ(.), ̂ θiT ) and ˜πi (a,h; ϕ(.)) are typically carried out by stochastic simulations. For further details, see Section 4 and the Appendix. =-=[5]-=-In a Bayesian context, g (θi |ZT ,Mi) is given by (6). Alternatively, in the case where the asymptotic normal theory applies to ̂ θiT , it may be reasonable to compute the probability density functio... |

32 | 2000a, A Decision Theoretic Approach to Forecast Evaluation - Granger, Pesaran |

28 |
Asymptotic inference on cointegrating rank in partial systems
- Harbo, Johansen, et al.
- 1999
(Show Context)
Citation Context ...del and allowing for Parameter Uncertainty 1 0.9 0.8 0.7 1 Quarter ahead 1 Year ahead 2 Years ahead Probability 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.5 1 1.5 2 2.5 3 3.5 Threshold Values in Percent Per Annum =-=[22]-=-. Figure 3: Probability Estimates of Inflation Falling within the Target Range using the Benchmark Model 0.9 0.8 Future Uncertainty 0.7 0.6 Future & Parameter Uncertainty Probability 0.5 0.4 0.3 0.2 ... |

24 | Asymmetric density forecasts of inflation and the Bank of England’s fan chart - Wallis - 1999 |

19 |
Forecast Error Bounds by Stochastic Simulations
- Blake
- 1996
(Show Context)
Citation Context ...s Generally speaking model-based forecasts are subject to five different types of uncertainties: future, parameter, model, policy and measurement uncertainties. This paper focusses on the first three =-=[2]-=-and considers how to allow for them in the computation of probability forecasts using an error correcting vector autoregressive model of the UK economy. Policy and measurement uncertainties pose spec... |

16 | Distribution-free tests - Neave, Worthington - 1988 |

11 |
Estimating the Expected Predictive Accuracy of Econometric Models
- Fair
- 1980
(Show Context)
Citation Context ...3.5 per cent and avoiding a recession. Following the literature, we define a recession as the occurrence of two successive negative quarterly growth rates. See, for example, Harding and Pagan (2000). =-=[13]-=-4.1 Predictive Distribution Functions In the case of single events, probability forecasts are best represented by means of probability distribution functions. Figures 1 and 2 give the estimates of th... |

7 |
Predictive Likelihood: A
- Bjornstad
- 1990
(Show Context)
Citation Context ...ing coefficients converge to their asymptotic distribution at a faster rate than the other parameters in the model. However, the general results of this section are not affected by this complication. =-=[3]-=-denoted by Pr (ZT +1,h |ZT ). For this purpose, models and their parameters serve as intermediate inputs in the process of characterization and estimation of Pr (ZT +1,h |ZT ). The Bayesian approach ... |

4 |
Explaining forecast failure in macroeconomics
- Clements, Hendry
- 2002
(Show Context)
Citation Context ...0 0 0 1 0 0 0 0 −1 1 0 0 0 0 0 0 0 0 0 0 1 0 0 −1 0 0 0 −β42 0 −β43 0 1 0 0 0 0 1 −1 0 0 0 0 ⎞ ⎟ . (23) ⎠ Under the assumption that oil prices are “long-run forcing”, efficient estimation of the param=-=[8]-=-eters can be based on the following conditional error correction model: ∆yt = ay − αy [ β ′ ] zt−1 − b1(t − 1) p−1 ∑ + i=1 Γyi∆zt−i + ψ yo∆p o t + uyt, (24) where yt =(et,r ∗ t ,rt, ∆pt,yt,pt − p ∗ t... |

4 | On the Design of Inflation Targets”, in Targeting Inflation - Yates - 1995 |

3 |
A Structural Cointegrating Macroeconomic Model of the UK, monograph under preparation
- Garratt, Lee, et al.
- 2000
(Show Context)
Citation Context ...gular Choleski factor of ˆΣ. In our applications, for each r and s, wegenerateε (r,s) T +i as IIN(0, Im), although other parametric distributions such as the multi-variate Student t can also be used. =-=[18]-=-A.3.2 Non-Parametric Approaches The most obvious non-parametric approach to generating the simulated errors, v (r,s) T +h, whichwedenote‘Method { 1’, is simply to take h random draws with replacement... |

3 |
Dissecting the Cycle: A Methodological Investigation" mimeo
- Harding, Pagan
- 1999
(Show Context)
Citation Context ... 0.14 Future Uncertainty 0.12 Future & Parameter Uncertainty 0.1 Probability 0.08 0.06 0.04 0.02 0 2001q2 2001q4 2002q2 2002q4 2003q2 2003q4 2004q2 2004q4 2005q2 2005q4 2006q2 2006q4 Forecast Horizon =-=[23]-=-Figure 5: Probability Estimates of Meeting the Inflation Target without a Recession† 0.9 0.8 Event A Not B 0.7 Product A Not B 0.6 Probability 0.5 0.4 0.3 0.2 0.1 0 2001q2 2001q4 2002q2 2002q4 2003q2... |

3 | Optimal Monetary Policy with a Nonlinear Phillips Curve - Nobay, Pell |

2 | Predictive Fit for Natural Exponential Functions - Harris - 1989 |

1 |
Testing for Changes
- Barassi, Caporale, et al.
- 2001
(Show Context)
Citation Context ...ulty in measuring the uncertainties associated with forecasts in the large-scale macroeconometric models typically employed. Another explanation relates to the various types of uncertainty that are in=-=[1]-=-volved in forecasting. For example, probability forecasts typically provided in the literature deal with future uncertainty only, assuming that the model and its parameters are known with certainty. ... |

1 | Working with Microfit 4.0: An Interactive Introduction to Econometrics - Pesaran, Pesaran - 1997 |