Hi,
I am estimating an international tourism demand model (inbound) taking individual origin countries into account incorporating both economic factors and noneconomic factors into consideration. There are some noneconomic variables such as climate and institution quality which are related to destination country. I wonder whether I can include those variable in a dynamic panel data model as the values of those variables will be the same for all countries (as those are related to destination country). The other issue is data for variable available for different time periods. For instance, for some variable 30 years and for some variables 16 years. Can I go for an unbalanced panel given this. Will it create additional problems?
Any advice is much appreciated.
Please refer to the article
"THE PULL FACTORS OF TOURISM DEMAND: A PANEL DATA ANALYSIS FOR LATIN AMERICAN AND CARIBBEAN COUNTRIES"
by
Smiljana Pivčević, Zvonimir Kuliš, Neven Šerić
at
http://www.fthm.uniri.hr/files/Kongresi/THI/Papers/2016/THI_April2016_319to333.pdf
You can use interpolation for some missing values but there should be the same time frame for the given variables
Good Luck
Dear, you can take a look at the following articles found in RG:
-Chasapopoulos, P., Butter, F.A.G. and Mihaylov, E. (2014). Demand for tourism in Greece: A panel data analysis using the gravity model. International Journal of Tourism Policy, Vol. 5, No. 3, pp. 173 - 191.
-Malaj, V. and Kapiki, S. T. (2016). Determinants of Tourism Flows to Greece: A Gravity Model Approach. TOURISMOS: An International Multidisciplinary Refereed Journal of Tourism (forthcoming).
I hope they are useful. Best wishes!
Hi dear you can use structural break to determine your dummy variables
regards
DEAR, YOU CAN USE THE LAG OF DEPENDENT VARIABLE AS A PROXY FOR NON ECONOMIC VARIABLES. AND SO THE MODEL WILL BE DYNAMIC.
AND ALSO YOU SHOULD USE UNBALANCED PANEL DATA BECAUSE OF NONIDENTICAL DATA.
How many destination countries? If only one then you cannot include any variable that is constant across origin countries because it will be multicolinear with the intercept. You might still include some weather indicators by taking the difference in (say) sunshine hours between destination and origin - the implied behavioural hypothesis is that tourists are more likely to travel for sunshine if they don't get it at home. Similarly for governance quality, you might construct (say) a 0/1 dummy to indicate whether the destination has poorer institutional quality than the origin,
(If several destination countries then consider making each origin+destination pair the cross-section observation index.)
Hi Vince,
Thank you for your advice. It is really helpful. I have only one destination country which is Sri Lanka and 20 origin countries. I am using Dynamic panel data model with GMM-Diff estimation technique. Institutional quality is measured using an index and it varies over time. Will it still create problems if I include those variables? Your advice is much appreciated.
Kind regards
Thushara
Thushara -
The index is "time-varying, x-section-constant" so you may have problems if any other variables have this behaviour. For example, you will not be able to use time-period fixed effects. You might try the admittedly crude approach of specifying your preferred model and noting (I) whether the estimates compute (ii) what their standard errors look like
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation