Academic journal article Atlantic Economic Journal

Output Expectations and Forecasting of UK Manufacturing

Academic journal article Atlantic Economic Journal

Output Expectations and Forecasting of UK Manufacturing

Article excerpt

Introduction: Output Expectations and Forecasting of UK Manufacturing

Expectations play a central role in macroeconomics (Evans and Honkapohja 2001) because, in an uncertain world, decision-making by the business community is based on their expectations concerning the future behaviour of key economic variables. As indicated by Branch (2004), the rational hypothesis has failed to explain the survey data on inflation expectations because of the knowledge required by economic actors on the structure of the economy. Even an updating rule-of-thumb, in the spirit of rational expectations, can require substantial resources and knowledge of the economy, which is on par with a skilled, experienced econometrician. It is this lack of ability and information for the formation of rational expectations that leads agents to adopt bounded rationality to explain the process of formation and diffusion of expectations (Sargent 1993); this is the approach applied in this paper to output expectations.

This article presents a model of output expectation formation applied to information derived from the Confederation of British Industry (CBI) Industrial Trends Survey. This data is explained in more detail in "The CBI Data" and "Output Expectations and Functional Form: A Logistic Model of Business Expectations" is concerned with relevant methodology and functional form. There are a range of approaches to modelling expectations of output in literature and three papers have been published, quite recently, applying these approaches specifically to CBI figures (Mitchell et al. 2002, 2005; Driver and Urga 2004). Mitchell et al. (2005) provide an overview of the alternative depictions of output expectation formation. Driver and Urga (2004) deal only with backward-looking models and, therefore, exclude forecasting. This paper maintains the monthly disaggregation of the government output statistics, which are used to explain the CBI Survey expectations. The articles by Mitchell et al. (2002, 2005) disaggregate by company respondent to the CBI Survey but, unfortunately, the ability to disaggregate in that way ceased in 1999, so there is no current forecasting application.

The CBI Data

The CBI Survey draws information from managers of at least 1,700 manufacturing companies, who are asked the following question, "Excluding seasonal variations, what has been the trend over the past 4 months and what are the expected trends for the next 4 months?" This is the company survey data on backward- and forward-looking expectations used in this article. According to Mitchell et al. (2005 p. 488), the majority of researchers in this field, "treat the difference between the 4-month period referred to in the survey and the quarterly frequency of our official data is being unimportant." The approach used in this study maintains the timing distinction between the quarterly survey and the monthly output figures, by comparing quarterly survey data with the monthly, non-seasonally adjusted government statistics. The survey respondents give an indication of output trends at the beginning of the month (m0) covering the previous 4 months (m0 to m-4), which are published each January, April, July and October (m+1). By contrast, official manufacturing output data, to the end of the month (m-1) is initially made available 2 months later (m+1), although subject to later correction (CBI 1983).

The CBI Survey started in 1958 on the basis of three questionnaires per year until 1972, when it became quarterly. We have converted the tri-annual figures from each year before 1972 into quarterly data by interpolation. The sources of the observations are the CBI and the Datastream database. The empirical analysis, in subsequent sections of this article, suggests that these 1959 to 1971 observations are reliable, with no evidence of heteroskedasticity in the regressions.

Respondents are asked to identify a trend over the preceding and coming 4 months without seasonal variation, but they may instinctively contrast their current response with that of a year earlier. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.