unit problem

[LAUJJL]

Hi
An enterprise having employees has a salary per employee per month constant that has unit equal to euro/(person*month).

Up to here everything is fine. If one wants to know the cost of all salaries per month one has to multiply the number of employees that has unit person with salary per employee per month that makes the cost of all salaries per month have a unit equal to euro/month. That makes sense.

But now imagine that there are four employees, Jim, Jack, Paul and Richard that have each a salary per employee per month equal to 2000,2200,2300 and 3000 Euros and one uses a subscripted constant for the salary of all the employees. Suppose that we give this constant the unit euro/(person*month) as the salary per employee per month. Now suppose that I want to know the cost of all these employees in a month.
The unit is obviously euro/month, but if I sum the salary per employee per month over the subscript sum(salary per employee per month[employees!]) in Vensim and probably in other packages too where employee is equal to Jim,Jack,Paul,Richard as a subscript the result has a unit equal to euro/(person*month) which is wrong. Where is the problem?

I think that the salary per employee per month is in fact the salary per month for all the subscripted employees and should have a unit equal to euro/month. Is it correct because it sounds weird to consider in one circumstance a constant with euro/(person*month) when the person is not defined and euro/month when the person is defined ?

One solution of course is to keep the unit euro/(person*month) for the subscripted constant and when making the sum to consider the formula: sum(salary per employee per month[employees!] * one person) where one person is equal to 1 with a unit equal to person. But I have tried this solution and it makes a model rather clumsy and not elegant.

I personally think that one can choose both solutions provided that one changes the units of all the other formulas accordingly. I prefer the euro/month solution because it is both logic: the salary per month of all the subscripted constants and easy to sum without adding the constant one person.

Any idea?
JJ

[tomfid]

Interesting puzzle. I slightly favor the sum(salary per employee per month[employees!] * one person) approach, but I can see the merits of doing it either way. I'll have to think about how this will play out in future versions of Vensim that are under development.

[LAUJJL]

Hi Tom

I favor the other solution.

The constant salary per employee per month unit (euro/(month*person)) means that if the constant is applied, the employee to which the salary is applied will earn a salary of euro/month or salary per employee per month * one person because the employee will earn the salary for one person.

The unit euro/(month*person) means Euros for one person during a month.

But in the case of a salary per employee per month[employee] where employee is Jim,Jack,Paul or Richard, one knows that they are one person and their salary has a unit Euro/month. Giving a unit
Euro/(month*person) does not make sense for the salary of Jim for instance.

If somebody asks you what is your salary you will say x dollars a month and not x dollars a month per person!

Besides it is much more convenient when used in a model.

I have a model with about 500 variables and I had the formula with euro/month first. When I changed to euro/(month*person) I had to change the unit of many other variable and to change nearly everywhere the salary to salary * one person, which is rather clumsy.

I do not think there is something to change in Vensim considering this problem. In fact both formulas describe something different.

Regards.
JJ

[tomfid]

I see your point.

I guess in an aggregate model, you'd still want $/person/mo, because there would not be the corresponding mapping of a person to a subscript element, so you'd still multiply by people to get the total expenditure flow.

In the (not immediate) future, it will be more apparent whether variables like salary are a property of an individual, department, etc.