Today in class, the professor said that the set of all consumption $c(S)$ is non-empty, compact and convex subset of $\mathbb{R}^T_+$. i.e. we know $\sum \limits_{t=1} ^T c_t = S$ where $c_t$ is consumption in period t and $S$ is total wealth. The set $c(S)$ is the set of all T-period consumption plan.

I could understand why $c(S)$ is compact, but I have no clue why it has to be convex and non-empty.

Following the same question, the professor also mentioned that $W(c)$ the summation of all the T-period utility functions denoted by $W(c)=\sum \limits_{t=1} ^T U(c_t)$ is continuous on $c(S)$. I do not understand why $W(c)$ has to be continuous on $c(S)$

I think this has something to with Berge's Maximum theorem, but I am unable to link it properly. Any help will be very appreciated.

  • $\begingroup$ How is $c(S)$ defined? Whether $W$ is continuous will depend on $U(.)$ is the sum of continuous functions itself a continuous function? Without more information about the context it is impossible to answer the question, so I think the best option is to consult the book/notes for the course or ask the professor. He is paid to teach after all. $\endgroup$ Dec 12, 2020 at 7:43
  • $\begingroup$ Hey @JesperHybel, the professor just mentioned C(S) be the set of all T–period consumption plans, which is not empty. With regards to the utilty functions, he mentioned that $U(.)$ is twice differentiable concave functions.. Also, I asked the Professor, he said it is for me to figure it out! So...not much help either $\endgroup$
    – Doubts
    Dec 12, 2020 at 8:30
  • $\begingroup$ Well I'm sorry to hear that. Still, you must have an optimization problem specifying some initial conditions as well as a transitionsmechanism. Without this there is nothing to suggest that $c(S)$ is even compact. Why could $c(S)=\{(1,1),(2,2)\}$ with $T=2$? Nothing in your question contradicts this. Short of someone guessing the details of the problem you are trying to grasp I am afraid you will not get an answer. So if you have anymore information pls. provide it. $\endgroup$ Dec 12, 2020 at 9:43

1 Answer 1


Assuming the following:

  • $ S > 0 $, $T > 0 $,
  • $ c(S) = \left\{c\in\Bbb{R}^T_{+} : \sum_{t=1}^T{c_t} = S \right\} $, and
  • $ U(\cdot) $ continuous.

Let's start with convexity. To demonstrate convexity, we need to show that for any two points $c^1$ and $c^2$ in $c(S)$, any linear combination of the two is also an element of $c(S)$.

Let $\lambda \in [0,1]$, $c^1 \in c(S)$, $c^2 \in c(S)$. Then, we have

$$ \sum_{t=1}^T{\lambda c^1_t + (1-\lambda) c^2_t} $$ $$ = \sum_{t=1}^T{\lambda c^1_t} + \sum_{t=1}^T{(1-\lambda) c^2_t} $$ $$ = \lambda \sum_{t=1}^T{c^1_t} + (1-\lambda) \sum_{t=1}^T{c^2_t} $$ $$ = \lambda S + (1-\lambda)S $$ $$ = S $$ $$ \Rightarrow \lambda c^1 + (1-\lambda) c^2 \in c(S) $$

Since any linear combination of any two points in c(S) is itself in c(S), we have c(S) convex.

For non-emptiness, we just need to show that a single point exists in c(S): $$ \sum_{t=1}^T{\frac{S}{T}} = S $$ $$ \Rightarrow \left\{ \frac{S}{T} \right\}^T \in c(S) $$

If $U(\cdot)$ is continuous on $c(S)$, then continuity of $W(c)$ follows from the triangle inequality (see e.g., here or here). If we don't know more about $U(\cdot)$, then there's not much we can say about $W(c)$. Berge's Theorem of the Maximum has to do with the continuity of an optimized function, not the sum of functions.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.