Turkish expectation surveys of the manufacturing industry: An investigation on prediction accuracy of production and sales data

ABSTRACTTURKISH EXPECTATION SURVEYSOF THE MANUFACTURING INDUSTRY:AN INVESTIGATION ON PREDICTION ACCURACYOF PRODUCTION AND SALES DATASelçuk, EbruIn Turkey, the so-called ?Expectation Surveys? includes the CEO?s subjectiveviews of manufacturing industry firms on quarterly basis. In terms of ?increase?,?remaining the same? and ?decrease? categories, evaluations of the present situationand predictions for the next period were asked by means of a survey, which wasformerly conducted by State Planning Organization and later on by State Institute ofStatistics.The primary aim of the study is to investigate the statistical accuracy ofpredictions with respect to the present situation evaluations for ?production? and?sales? data of ?Expectation Surveys?.Although the survey attempt started back in the early 1960?s, comparable datacould be gained for 47 quarters of 1992-2003 periods. Accuracy evaluations weremade on different grounds for different purposes, and each time by approaching tothe hypothesized problem through using an appropriate statistical approach.Long term correlations among ?expectation? and ?realization? data of the?manufacturing industries? aggregate had proven to be very significant. By ISIC-Rev. 3 subdivision of data for 23 quarters among 1998-2003, 22 2-digitmanufacturing series were also examined. Through applying a meta-correlation ratioupon ordinal scale correlations, a method had been devised to perform theevaluation. The outcome indicated significant prediction accuracy at 95% level andmore for all of the sub-sectors. On the other hand, as a contribution to some popularargument, to observe whether ?state? and ?private? sector respondents had thechances to predict significantly better, data were transformed into chi-squaredequivalents, in order to apply a sign test upon relative accuracies. Conclusively, nogroup seemed to be in a position of forwarding more accurate estimates of future.At 35 quarterly time points of 1992-2000, expectation aggregates could also becompared with quarterly GDP data both at current and constant prices, in order toinspect, whether the survey data could be an aid for a very short-run prediction ofGDP. The answer was not very promising; however, sales data seemed to be a moresuitable attribute for such a task.Those evaluations from different aspects leaded to the general conclusion thatthe ?Expectation Surveys? data might be an important indicator to refer for economicpolicy implications and short-run forecasts.Keywords: expectation surveys, production, sales, short-run prediction, correlationanalysis, sign test.