IDENTIFYING AND MODELLING THE
MAIN MACRO DRIVERS OF TOURISM
DEMAND WITH MATLABIman Behzadian
Senior Data Scientist
Tourism and Events Queensland
12/06/2015 5:03 PM
2
Problem Definition
Why are econometric models important?
Developing a Framework
Methodology/Results
How did MATLAB help?
Make Insightful Decision Based on The Results
Contents
12/06/2015 5:03 PM
3
Modelling The Effects of the Macro Drivers of a
Source Market on Tourism Demand
Quantitative Analysis
What are the key drivers of tourism demand
What are their relationships?
What are the best proxies for them?
What are the relative importance of these drivers
Problem Definition
12/06/2015 5:03 PM
4
the need for an analysis of the adequacy of model specifications and the statistical deficiencies of
existing empirical tourism demand models is sensed in this area.
Why is it important to validate the model?
12/06/2015 5:03 PM
5
Choosing the right Technique depend on the nature of the problem.
Some of the Modelling techniques have higher explanatory power than others. Some are more accurate in terms of predicting the future.
Modelling Techniques
Parametric Modelling
•Curve Fitting
•Linear Regression Techniques
Machine Learning
•Supervised
•Unsupervised
Dynamic Modelling
•ARIMA
•GARCH
•VAR
•SDE
12/06/2015 5:03 PM
6
How Does Matlab Help?
Parametric Modelling
•Curve Fitting
•Linear Regression Techniques
Machine Learning
•Supervised
•Unsupervised
Dynamic Modelling
•ARIMA
•GARCH
•VAR
•SDE
• Curve Fitting Toolbox
• Statistics
and Machine Learning Toolbox
• Statistics and Machine Learning
Toolbox
• Neural Networks Toolbox
• DatafeedToolbox
• Database Toolbox
• Econometrics Toolbox
• Optimization toolbox
12/06/2015 5:03 PM
7
Selecting The Major Macro Drivers
Selecting Proper Proxies for Each
Preparing The Data Tuning VAR Model
Checking Impulse
Response Analysis and Back Testing
Results
Developing a Framework (VAR)
12/06/2015 5:03 PM
8
Methodology
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
Granger causality Test (Based on T-test) can be a good indicator
Tourism Demand
12/06/2015 5:03 PM
9
Unit Root Test for Stationarity
ARCH Effects Test for Residual Autocorrelation
Making The Data Stationary
Removing Seasonality
Adjusting the Shifts occurred due to This Process
Methodology
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
12/06/2015 5:03 PM
10
Methodology
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
Simple AutoCorrFunction for AR term of ARIMA or VAR model
Partial AutoCorrFunction for MA term of ARIMA or VARMA
Simple and Partial
AutoCorr Function on the squared version of signal for GARCH model
Seasonality
A djus tment
Needed
Suitable for VAR (n=2)
model
Diff
Needed
12/06/2015 5:03 PM
11
Results
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
The Econometric model is developed to estimate monthly relationship between total inbound tourism spending in Australia from each source market and the following explanatory variables.
Economic Activity and wealth as proxied by local stock exchange index;
Household Income proxied by Gross Domestic Product (GDP);
Exchange rate between Australia and the source market;
Consumer sentiment proxied by Consumer Confidence Index (CCI);
Consumer price index (CPI)
Vector Auto regression model is used.
12/06/2015 5:03 PM
12
Results
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
Tourism Demand Forecast, of a Specific Market- (Average) Relative Error 0.56%
Tourism
Dem
and
($M
)
12/06/2015 5:03 PM
13
Results
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
12/06/2015 5:03 PM
14
Results
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
Impulse Response Analysis Is a good indicator of the behaviour of the model (system)
12/06/2015 5:03 PM
15
Results
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
Changes i
n T
ouris
m D
em
and due t
o 1
-std
change in
driv
ers
($’000)
12/06/2015 5:03 PM
16
Results
Selecting Proxy
Preparing Data
Tuning
Model
Parameters
Checking the results
Indicator of How much of The dependent
variable is not explained by these
drivers
Changes i
n T
ouris
m D
em
and due t
o 1
-std
change in
driv
ers
($’000)
12/06/2015 5:03 PM
17
Wealth (proxied by local stock exchange index), Exchange rate and income (proxiedby GDP) respectively are the key drivers of Tourism Demand.
The negative effects of changes in exchange rate on Tourism Demand appears faster than the positive effects of changes in wealth proxy.
Sensible strategy is to invest behind the trend. In similar situation for 2 markets of equal size, we should therefore invest more in the market with highest growth in the key drivers.
These Macro drivers strongly influence and predict Tourism demand. Rational strategy is to increase our market share;
Interpreting The Results
12/06/2015 5:03 PM
18
The need for an analysis of the adequacy of model specifications and the
statistical deficiencies of existing empirical tourism demand models is sensed in this area.
To respond to this need, A framework for developing Econometric models
(using VAR) is suggested.
In this framework, to validate the model, We examine the impulse response
analysis of the structure of the model to see whether or not it matches the assumptions that we already have.
The results of our model using this platform, for tourism demand of a specific market was presented.
Conclusion