Forecasting Holiday Travel Propensity using Markov Chains

Purpose The purpose of this paper is to show how Markov chains can be used as an additional forecasting source. Design/methodology/approach Holiday Travel Propensity (HTP) can be seen as a Markov process, so that the transitions from taking or not taking a holiday travel in one year into taking or not taking a holiday in the subsequent year are modeled. As opposed to other papers which made use of the Markov chain approach, this paper models holiday travel propensity from a source market perspective and shows how to integrate influencing factors (age, health status) and linear regression analysis into the Markov chain approach. Findings Markov processes will reach an equilibrium (Steady state) if transition rates are stable. Although this is not perfectly the case in Germany during the last eleven years, an HTP equilibrium of slightly less than 80% can be estimated. In general, the transition rates from taking a holiday travel in one year into taking a holiday travel in the subsequent year grew by approximately 0.22 percentage points per year over the last decade, while transition rates from not taking a holiday travel in one year into taking a holiday travel in the subsequent year fell by approximately 0.75 percentage points per year. Analyzing transition rates in different age groups reveals that in older groups (60 years and older) equilibrium values are falling over time although actual HTP values are still on the rise. This might be an indicator that growth potential (in terms of participation) is lower for the older age groups compared to the younger groups. Furthermore, in the oldest group actual HTP values are slightly above the equilibrium, while in all other groups actual HTP values are below equilibrium. Same goes for people with poor health status. This might be an additional indicator that in those two groups (oldest, poor health) equilibrium has been reached, even if we take into consideration that these two indicators are correlated. Taking the same model one step further we can also derive equilibrium values for the group of intensive holiday travelers (taking two or more holidays per year). For this group, the equilibrium is at 19% of the German population. Research implications From a research perspective, the paper shows one of the main advantages of Markov chain transition analyses compared to other forecasting methods. This advantage is the integrated predictive value which makes it possible to derive steady state levels with very little prior knowledge. Time series analysis, in this paper in the form of a simple linear regression model, can however improve the assessment of the results. Practical implications From a more practical point of view, the paper shows that Markov chains are a relatively straightforward and efficient way of deriving predictive value from survey data. The paper also points out the limitations of the method, making it particularly suitable as an additional method requiring only very moderate analytical effort. Originality/value The paper shows the basic principle of Markov chain transition analysis and its applicability to real tourism data, so that access to the methodology for researchers and the industry can be facilitated. It shows the combination of Markov chains and time series regression models. For the German market, it gives estimates for the steady state in terms on holiday travel participation and shows ways to identify segments with market potential