Tourism Demand Forecasting - Time-Series Models
Methodology  -  Lake State Examples - Other Examples         
Back to Table of Contents                                                                                                   


Uysal, M. and J. L. Crompton. 1985. "An Overview of Approaches Used to Forecast Tourism Demand." Journal of Travel Research 23(Spring):7-15.

This paper presents a brief review of the tourism forecasting literature as of 1985. Three qualitative techniques are examined: simple survey techniques, Delphi models and judegement-aided models. Three quantitative techniques are also reviewed: Time-series, gravity and trip generation models and multivariate regression models.

Archer, B. H. 1980. "Forecasting Demand, Quantitative and Intuitive Techniques." International Journal of Tourism Management 1(1):5-12.

This paper reviews the art of demand forecasting in tourism as of 1980. Time-series analysis, the causal methods of multivariate regression and gravity and trip generation models, and the qualitative techniques of surveys, expert group processes and Delphi modeling are all discussed. The author concludes that integrated techniques, which combine quantitative methods with expert judgment may be the most accurate.

Lake States Examples:

Other Examples:

Song, H., S. F. Witt and T. C. Jensen. 2003. "Tourism forecasting: accuracy of alternative econometric models." International Journal of Forecasting 19(1):123-141.

The accuracy of econometric tourism demand models has been the subject of considerable academic investigation and literature over the past several decades. Despite the theoretical elegance of the models, in most cases they have poorer accuracy than more simple techniques. This paper claims to present the most comprehensive comparison to date of the performance of econometric forecasting models with a tourism context. The six econometric models evaluated are all special cases of a general autoregressive distributed lag specification. Two time-series models are also evaluated for baseline comparison. Each of the models are evaluated for inbound tourism to Denmark for one, two, three and four year predictions. The time-varying parameter (TVP) and long-run static cointegration regression model perform most consistently. The naive or no-change time-series model is not far behind in terms of accuracy, especially for predictions in years one and two.

Kulendran, N. and S. F. Witt. 2001. "Cointegration Versus Least Squares Regression." Annals of Tourism Research 28(2):291-311.

Least squares regression models that explain international tourism demand have been shown to generate less accurate forecasts than the naive "no change" model. This study investigates if the reason for such mediocre forecasting performance is the failure to adopt recent developments in econometric methods in the areas of cointegration, error correction models and diagnostic checking. The empirical results demonstrate that the forecasts produced using these methodological developments are more accurate than those generated by least squares regression, but that these newer econometric models still fail to outperform the "no change" model, as well as statistical time series model.

Tideswell, C., T. Mules and B. Faulkner. 2001. "An Integrative Approach to Tourism Forecasting: A Glance in the Rearview Mirror." Journal of Travel Research 40(November 2001):162-171.

This study evaluated the results of tourism forecasting exercise in South Australia for the period of 1996-1998. An integrated times-series and Delphi process was used to forecast tourism demand from both the international and domestic market. Three methods of generating time-series forecasts were used: Holt's exponential smoothing; a "naive" method using average annual rate of change from the past 11 years; and a linear trend using regression analysis. A group of 26 tourism industry practitioners responded to a Delphi survey and provided comments in a follow-up group meeting process. The final forecasting method used was based on the results of the Delphi process. Grouped forecasting trends (e.g. all international visitors) were quite accurate, but forecasts for segments of the market (e.g. Japanese visitors) were not as accurate. This was particularly true for small market segments. Over all, the "naive" approach to demand forecasting was most accurate, confirming the results of earlier studies.

Martin, C. A. and S. F. Witt. 1989. "Accuracy of Econometric Forecast in Tourism." Annals of Tourism Research 16:407-428.

This paper presents the results of a study which compared the accuracy of several econometric forecasting models. The models evaluated include a log-linear econometric model, two "naive" models, an exponential smoothing model, a trend curve analysis model, a Gompertz trend curve model and a step-wise autoregression model. Each model was used to predict the travel flows of international tourists 1 and 2 years in the future. A naive "no-change" model was found to be more accurate than econometrics in 70% of the cases.

Choy, D. J. L. 1984. "Forecasting Tourism Revisited." Tourism Management 5:171-176.

This article examines the accuracy and efficiency of forecasting techniques by applying time series regression to forecasting visitor arrivals. Past studies have shown that simpler time series techniques perform as well or better than complex forecasting models. Time series regression is compared to a naive times series model. An assessment of visitor forecasts developed at regional, destination and individual market levels suggests that time series regression performs well in producing annual forecasts of visitors. These forecasts can also serve as a baseline for evaluating the net returns from applying more complex techniques. Tourism managers should appreciate the usefulness of simpler formal methods in developing forecasts of tourism.

Back to Table of Contents