Kerry Patterson

Data Revisions and the Expenditure Components of GDP

The constant price components of the expenditure estimate of GDP for the United Kingdom undergo an extensive process of revision as new information comes to light. This study is concerned with three central themes: the authors assess whether the preliminary vintages efficiently incorporate information available at the time of their compilation; whether there is any evidence for the nonstationarity of data revisions; and how to construct constant price time series when there are several changes of base. Copyright 1991 by Royal Economic Society.

Multivariate singular spectrum analysis for forecasting revisions to real-time data

Real-time data on national accounts statistics typically undergo an extensive revision process, leading to multiple vintages on the same generic variable. The time between the publication of the initial and final data is a lengthy one and raises the question of how to model and forecast the final vintage of data – an issue that dates from seminal articles by Mankiw et al. [51], Mankiw and Shapiro [52] and Nordhaus [57]. To solve this problem, we develop the non-parametric method of multivariate singular spectrum analysis (MSSA) for multi-vintage data. MSSA is much more flexible than the standard methods of modelling that involve at least one of the restrictive assumptions of linearity, normality and stationarity. The benefits are illustrated with data on the UK index of industrial production: neither the preliminary vintages nor the competing models are as accurate as the forecasts using MSSA.

DYNAMIC TIME SERIES MODELS WITH GROWTH EFFECTS CONSTRAINED TO ZERO

The contribution of this note to the analysis of the dynamic properties of estimated equations is as follows : ( i ) We derive the generating function for the growth coefficients (of which the dynamic multiplier is a special case) implicit in a lag distribution; and note the link between the error correction mechanisms of Salmon (1982) and growth coefficients. (ii) We suggest that the Wald principle is a sensible testing procedure given the hypothesis, that particular dynamic multipliers are zero, to be tested is (usually) non-linear. (iii) Demonstrate that the imposition of the constraint that a particular dynamic multiplier be zero is likely to induce substantial changes in the lag distribution not just of the variable being constrained, but also on the dynamic response of other unconstrained variables. (iv) We conclude that considerable caution needs to be exercised in the imposition of such constraints, and that this aspect of model analysis should not be applied as a matter of routine.

The Bank of England quarterly model of the UK economy

Abstract The Bank of England first acquired a macroeconomic model of the UK economy in early 1973, and used it for forecasting in June and July of that year. The initial model was obtained from the London Business School (LBS), but the last 14 years or so have, on the part of both the Bank and the LBS, led to developments which now make the models no closer to each other than to other large scale models of the UK. This article describes the structure and central properties of the current version of the Bank model, which has some 663 variables, 134 of which are modelled by behavioural equations, 153 by technical equations, 212 identities and 164 exogenous variables. In order to provide a transitional step between the kind of models with which most macroeconomists are familiar and the full scale version of the Bank model, one part of the article presents a very simple stylized version of the full model. This is a stepping stone to the full model which is described and analysed on a sectoral basis in Part 3, with a complete listing of equations in Part 4. Some of the simulation properties, and hence full model dynamic responses, are considered in Part 5.

A systems approach to the relationship between consumption and wealth

The single-equation approach to the determination of consumption, due to Hendry and von Ungern-Sternberg (HUS), is extended to a multi-equation system which links decisions on consumption and the components of wealth. The original HUS approach has proved to be a very durable one, being applicable to a number of countries and macroeconomic models. This approach is extended to allow a role for rates of return as well as income, and it is shown how this extension results in an error correction system. The empirical analysis used is based on Johansens maximum likelihood extension of the co-integration approach of Engle and Granger.

Bias Reduction through First-order Mean Correction, Bootstrapping and Recursive Mean Adjustment

Abstract Standard methods of estimation for autoregressive models are known to be biased in finite samples, which has implications for estimation, hypothesis testing, confidence interval construction and forecasting. Three methods of bias reduction are considered here: first-order bias correction, FOBC, where the total bias is approximated by the O(T−1) bias; bootstrapping; and recursive mean adjustment, RMA. In addition, we show how first-order bias correction is related to linear bias correction. The practically important case where the AR model includes an unknown linear trend is considered in detail. The fidelity of nominal to actual coverage of confidence intervals is also assessed. A simulation study covers the AR(1) model and a number of extensions based on the empirical AR(p) models fitted by Nelson & Plosser (1982). Overall, which method dominates depends on the criterion adopted: bootstrapping tends to be the best at reducing bias, recursive mean adjustment is best at reducing mean squared error, whilst FOBC does particularly well in maintaining the fidelity of confidence intervals.

Consumption: innovation persistence and the excess smoothness debate

Based on studies using US data, consumption is too smooth relative to the implications of the rational expectations permanent income hypothesis (REPIH) of consumption. We examine this and a number of related methodologicalissues for the UK using maximum likelihood estimation (MLE) of an autoregressive fractionally integrated moving average (ARFIMA) model of the income process. We conclude that an innovation to income has a persistent effect on consumption and that, in contrast to the findings from US studies, consumption is too volatile compared to the REPIH model. Further, MLE is substantially more efficient than the semi-parametric method of estimating ARFIMA models which has previously been used in this area.

The information content and gain of revisions to the components of GDP

Whilst it is well-known that data comprising the components of GDP are revised several times before a final figure is published, it is probably less well-known that the process of revision and refinement of the initial data is lengthy one. There can, for example, be six years between publication of the initial and final data. This study develops and applies information measures due to Theil to describe and analyse the revisions process. Our data set is a unique one made available to us by the CSO, comprising the constant price components of the expenditure measure of the GDP and the current price components of the income measure of GDP. We report on the size of revisions over time and the extent to which information gains or losses characterize later vintages of the data. We also consider the pattern of revisions to the growth rate of GDP and its components.

Aggregate consumption of non-durables and services, and the components of wealth: some evidence for the United Kingdom

The growth of spending on consumer goods in the United Kingdom in recent years has been a severe test of previously well-established consumption functions. This paper assesses the performance of a dominant ‘standard’ consumption function which relies on income and net liquid assets; it turns out that this has not been able to track the growth in consumer spending. However, developments of this consumption function which find a role for the components of personal sector wealth perform rather well, with satisfactory diagnostic statistics and an estimated equation standard error of around 0.4%.

Weighted symmetric tests for a unit root: response functions, power, test dependence and test conflict

The most frequently applied test statistics for a unit root are the Dickey–Fuller tests, which are built into many econometric packages along with MacKinnons empirical response functions. This article provides empirical response functions for some easy to compute alternative test statistics that are generally much more powerful than the Dickey–Fuller tests; specifically, these are the Dickey–Fuller tests and the weighted symmetric versions of the and tests. The empirical response functions presented here take into account adjustments for lag length in the maintained regression, and also extend the design of the simulation experiments compared to previous work. A second aspect of this study concerns the widespread practice in applied econometrics of using more than one test for the same feature without an assessment of the implications for the cumulative significance level and probability of test conflict. Tests for a unit root being are a leading example of this practice. Using the extended set of unit root tests considered here, the extent of test dependence is simulated and overall type one error calculated. Two empirical applications illustrate the key principles.