Ventana software support forum

Hi,

I am building a model of London's long-term housing developments, and in part of my model I need to capture the effect of changes in house prices on new construction. Based on my review of the literature, an increase in house prices does not necessary lead to an increase in new construction, but a fall in house prices can generally lead to a significant fall in new construction.

I was wondering if, off the top of your head, you could think of any way to model this sort of asymmetric causal effect in general where an increase in A does not affect B, but a decrease in A would reduce B.

Many thanks,
Kaveh.

I'd be interested to know what effects are omitted in the literature, because obviously there should be a supply response to higher prices, all else equal. Presumably the problem is land and legal constraints, or labor costs rising with prices, or something like that.

A simple formulation would be something like:

Long run price = SMOOTH(price, adaptation time)
Supply = f( price/long run price )

where f() could be a lookup or an analytic function with a steeper slope for input <1 than >1.

This captures the asymmetry, but it's not very explicit about physics or behavior. Something more elaborate that included things like available land, construction delays, labor supply, etc. would be preferable.

Hi
What is the purpose of the model? is it to study specifically the effect of price on housing or is it just necessary to know the future of housing in London. Because one can for that just build a time dependant lookup or an analytical time dependant function showing the evolution of housing, or one can look for the root causes of housing. The problem is the risk of replacing one uncertainty by many other uncertainties. For example, to explain the housing one must look for the explanations which is the first uncertainty (are the causes the right causes?), then find the correct influence of these causes on housing and if there are three causes, there is now four uncertainties and after that one must determine the behaviours of the three causes and that makes three other uncertainties and in total seven uncertainties. It may be preferable to work first with one uncertainty than jumping to multiple root causes, unless the purpose is to build an impressive model that will never deliver any utility.
Sorry to be very practical!
JJ