Potential doublecounting in a example model

Use this forum to post Vensim related questions.
Post Reply
WayneZhou2009
Senior Member
Posts: 105
Joined: Wed Oct 25, 2017 3:52 pm
Vensim version: PRO

Potential doublecounting in a example model

Post by WayneZhou2009 »

Hi,
This might be trivial to ask, but somehow I could not figure it out. It is about a sample in Vensim's UserGuide Chap11 folder, Houses.mdl.

The "planning" rate is defined to be MAX( 0, replacement houses + (gap in houses / time to respond to gap)), the "gap in houses" is defined to be "number of houses required - Houses Completed", and "replacement houses" is defined to be "demolishing".

However, since "demolishing" is already the outflow of "Houses Completed", why should it still be included in the "planning"? Wouldn't it cause sort of double-counting?

Thanks.
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: Potential doublecounting in a example model

Post by tomfid »

Good question.

Substituting the equations into one, you have:

Planning Rate = MAX( 0, demolishing + (number of houses required - Houses Completed) / time to respond to gap )

If you start with number of houses required = Houses Completed, the second term will be 0. Therefore the planning rate will just replace demolishing, and there will be no double-counting.

However, if a gap arises due to some other reason, like an increase in the number of houses required, the first term (replacing demolishing) would not adjust the housing stock to the desired level. So, you need the second term as well.

Similarly, if the planning process somehow fails to replace demolishing during some interval (perhaps the contractor ran off with the money instead of building the houses), a gap will develop, and the second term will correct for it. In a sense, this is double-counting, but you need it.

So, in short, the first and second term are correcting for different things: first, replacing the expected outflow; second, correcting for any accumulated mismatch of the stock to the desired level.

To be really strictly correct here, the first term should be expected demolishing, not actual demolishing, because actual demolishing isn't necessarily known instantly (fires are a surprise, for example). So, it might be better to set:
replacement houses = SMOOTH( demolishing, time to perceive losses )
LAUJJL
Senior Member
Posts: 1421
Joined: Fri May 23, 2003 10:09 am
Vensim version: DSS

Re: Potential doublecounting in a example model

Post by LAUJJL »

Hi

I had too noticed that.

But I considered that both solutions can be right.

If one considers only the gap, of course one must take away the replacement houses.

But one can too consider that taken into account the time to respond to gap, houses demolished will enter more quickly the planning and taken into account the time to plan to build and build, will be replaced after 9 months, one month more than the time to respond to the gap. In fact replacing the demolished houses generates a dynamic in 9 months, and the gap in houses generates a dynamic with longer delays. In fact there should exist two dynamics, one replacing immediately the houses demolished and another one trying to reach an objective of houses built.

To decide which is the best solution, one should explore better how the planning is really done.

Regards.

JJ
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: Potential doublecounting in a example model

Post by tomfid »

In a typical housing market, the term for replacement of demolition probably has low weight. In other words, the market doesn't notice the outflow directly; it only responds when the gap has grown big enough to affect prices. That, plus the inability to perceive the construction pipeline easily, makes the market unstable.
WayneZhou2009
Senior Member
Posts: 105
Joined: Wed Oct 25, 2017 3:52 pm
Vensim version: PRO

Re: Potential doublecounting in a example model

Post by WayneZhou2009 »

Hi JJ,

Thanks for your insights. If my understanding of what you said is correct, did you suggest to apply SMOOTH to demolition replacement, with around 1 year time to perceive (slightly more than 10 months), AND at the same time also apply SMOOTH to the gap in house (number of houses required - Houses Completed), with a substantially longer time to perceive?

Many thanks.
WayneZhou2009
Senior Member
Posts: 105
Joined: Wed Oct 25, 2017 3:52 pm
Vensim version: PRO

Re: Potential doublecounting in a example model

Post by WayneZhou2009 »

Hi Tom,

Thanks for your further comments. Did you mean to apply SMOOTH to demolition replacement, but NOT to the gap in house? But, in reality, as you said, the market only responds when the gap has grown big enough. It has to take enough time for the gap to grow and thus be perceived by the market. Hence, perhaps a SMOOTH should be used to the gap in house, too?

Many thanks.
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: Potential doublecounting in a example model

Post by tomfid »

I agree that there is also a delay between existence of a gap (desired-actual houses), but I think in that case a smooth is not the best way to represent it. It would be better to represent price formation explicitly. However, if that's not the focus of the model, a smooth, or nothing, might be OK.
LAUJJL
Senior Member
Posts: 1421
Joined: Fri May 23, 2003 10:09 am
Vensim version: DSS

Re: Potential doublecounting in a example model

Post by LAUJJL »

Hi Wayne

Not being a professional of that job, it is difficult to be categorical. The demolished houses having no delay, one can add a smoothing action, with what period?

Adding another smooth to the gap, seems a bit a lot, as there is already a time to respond to gap, which is already a smoothing action.

Anyhow, this is a learning model, that does not pretend to be used practically.

To be a bit closer to the reality, as suggests Tom, prices should be included as it is the primary driver of the behaviour of the three principal actors: house builders, house buyers and house habitants.

If you are interested by building a more realistic model, you can small step by small step, incorporate one after the others, additional factors. But this model has already a lot of things to be studied and is far from evident to understand intuitively.

I have for instance tried to suppress the double accounting by modifying the planning equation. But the model behaved strangely, until I understood that there is a bug in the house model that is not the double accounting. It is in the initialisation of the planned houses and houses in construction levels. These two levels are initialised with their input rates, something that does not make sense.
The input rates are happening during the month and cannot influence the state of their levels at the beginning of the month. Vensim accepts this formulation and calculates the rates before initialising the levels using the houses completed level that is correctly initialised. I have then corrected the initial state of both levels, putting them to 0, requiring a certain time to reach a steady state.

I have joined a model with three runs: two models with the double accounted formulation and another one with no double accounting, and another that starts with houses completed and houses required unequal.

One sees that having a double accounting or not, does not make a lot of difference, maybe in extreme conditions.

I have not studied deeply that model, that may still have some bugs or problems in it. But understanding it fully would require some time.

Regards.

JJ
HOUSES_3.vpm
(116.82 KiB) Downloaded 224 times
LAUJJL
Senior Member
Posts: 1421
Joined: Fri May 23, 2003 10:09 am
Vensim version: DSS

Re: Potential doublecounting in a example model

Post by LAUJJL »

It is possible too to just discard the replacement houses and keep the gap, to avoid double accounting. it genereates a slightly different behaviour, but it does not change it funtamentally.

See joined a corrected model with the three options.

Regards.

JJ
Attachments
HOUSES_4.vpm
(144.89 KiB) Downloaded 197 times
Post Reply