Hi there,
I am trying to estimate the effect of tourism on house prices in London with a panel of N=32 (London boroughs) and t=186 (monthly observations)
The dependent variable is the log mean house price in borough i at time t which exhibits significant nonstationarity so I expect that some dynamic technique will need to be used.
The key variable of interest is the number of tourist residences in each borough for each month but then I am also controlling for other factors that the literature assumes are determinants of house prices namely: population, number of residential units, crime rate, real incomes.
I had initially set out to use a fixed effects estimator (xtreg) to control for unobserved variables that differ across boroughs but not across time as well as binary variables for each period (i.time) to control for factors that are common across all boroughs but are time variant (interest rates, tax rates etc.)
What I'm wondering is if this is the best approach to deal with the question in hand, namely because I expect there to be some degree of simultaneous causality endogeneity between the house price and tourist residences (tourist demand is likely to concentrate in gentrifying or gentrified areas)
As this will likely be a dynamic model I am having issues with finding a suitable estimation technique as it seems that most of the packages are reliant on large N small t.
Is there any chance that anybody could point me in the right direction of some empirical strategies that might have similar conditions?
Many thanks
I am trying to estimate the effect of tourism on house prices in London with a panel of N=32 (London boroughs) and t=186 (monthly observations)
The dependent variable is the log mean house price in borough i at time t which exhibits significant nonstationarity so I expect that some dynamic technique will need to be used.
The key variable of interest is the number of tourist residences in each borough for each month but then I am also controlling for other factors that the literature assumes are determinants of house prices namely: population, number of residential units, crime rate, real incomes.
I had initially set out to use a fixed effects estimator (xtreg) to control for unobserved variables that differ across boroughs but not across time as well as binary variables for each period (i.time) to control for factors that are common across all boroughs but are time variant (interest rates, tax rates etc.)
What I'm wondering is if this is the best approach to deal with the question in hand, namely because I expect there to be some degree of simultaneous causality endogeneity between the house price and tourist residences (tourist demand is likely to concentrate in gentrifying or gentrified areas)
As this will likely be a dynamic model I am having issues with finding a suitable estimation technique as it seems that most of the packages are reliant on large N small t.
Is there any chance that anybody could point me in the right direction of some empirical strategies that might have similar conditions?
Many thanks
Comment