A context-specific causal model for estimating the effect of extended length of overnight stay on traveller's total expenditure

Abstract

Tourism significantly affects the economies of many countries. Understanding the causal relationship between the length of overnight stay and traveller's expenditure is crucial for stakeholders to characterize spending profiles and to design marketing strategies. Causal mechanisms differ between personal and work-related travel because the decision-making processes have different drivers and constraints. We apply context-specific independence relations to model causal mechanisms in contexts specified by trip purpose and identify the causal effect of the length of stay on expenditure. Using the international visitor survey data on foreign travellers to Finland, we fit a hierarchical Bayesian model to estimate the posterior distribution of the counterfactual expenditure due to extending the length of stay by one night. We also perform a Bayesian sensitivity analysis of the estimated causal effect with respect to omitted variable bias.

Figures (7)

• Figure 1: DAG (left) and LDAG (right) for illustrating the same data generating process, where additional context specific information resulting from node $M$ makes the query $P(Y \,\vert\, \textrm{do}(X))$ to be identifiable. In the graph nodes indicate the observed (circles) and unobserved variables (squares) and the edges represent the causal directions. Dashed edges with notation indicate the vanishing causal directions in the specific context. A circle node with an inner circle points out to the variable to be intervened.
• Figure 2: LDAG representing traveller's total expenditure process: The circle nodes represent the observed variables in the data. The labels in the arrows, describe the causal paths that vanish when context is either personal trip ($M=0$), or work-related trip ($M=1$). The rectangle nodes $U_1$ and $U_2$ indicate unobserved confounders.
• Figure 3: An augmented LDAG, where traveller's income $I$ introduces new causal links including a confounding pathway between the length of stay $X$ and traveller's total expenditure $Y$.
• Figure 4: Counterfactual difference of increasing the length of stay by one night for personal trips and work-related trips. Labels on the x-axis indicate the original length of stay and the increment applied (in parenthesis).
• Figure 5: Prior distributions for correlations between income (I) and length of stay (X), and between income and total expenditure (Y) in both personal trip and work-related trip context.