THE RELATIVE ACCURACY OF THE RANDOM WALK THEORY IN LONG-TERM MARKET FORECASTING.
Item
-
Title
-
THE RELATIVE ACCURACY OF THE RANDOM WALK THEORY IN LONG-TERM MARKET FORECASTING.
-
Identifier
-
AAI8302557
-
identifier
-
8302557
-
Creator
-
SCHNAARS, STEVEN PHILIP.
-
Contributor
-
Conrad Berenson
-
Date
-
1982
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Business Administration, Marketing
-
Abstract
-
Can long-term market growth be predicted? Do annual unit sales series follow a predictable pattern, as suggested by the product life cycle, or do they wander aimlessly? Will an econometric approach to long-range forecasting yield more accurate forecasts than time-series approaches? Do complex approaches to forecasting yield more accurate forecasts than simple methods? Does combining forecasts lead to greater accuracy? This research sought to answer these questions.;The relative accuracy of ten widely used market forecasting methods was examined on a "hold-out" sample for one through five year forecasts genrated from a large number of product sales series. In total, over 10,000 forecasts were generated and compared.;Not only were simple models found to provide the most accurate forecasts, a result consistent with recent research findings, but the most simple Random Walk model yielded the lowest mean absolute percentage error. This implies that popular marketing concepts such as the product life cycle should be reconsidered. A large percentage of the sales series tested in this research showed no inertia in their evolutionary process.;The predictability of various subgroups within the sample varied greatly by method. Product perishability, time horizon, data availability, judgmental assessments of stability, autocorrelation structure, and the number of runs were all found to affect the relative accuracy of the forecasting models tested. The product class/form dichotomy did not affect accuracy. These findings suggest that marketers should focus more intently on the forecasting situation.;Econometric models (single equation) were not found to yield superior forecasts even though the predictors chosen were deemed by experts to be crucial to the future of their industry.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
-
Program
-
Business
