The number of workers in a fund equals the number of firms which participate in a pension scheme multiplied with number of workers per firm (one thousand for each firm). Currently there are $5$ firms. In every decade, a firm can go bankrupt ($x=0$), stay operating ($x=1$) or start new firms ($x=2,3,4,\ldots$). For each firm, the probability mass function of the number of firms in the next $10$-year period is given by
$p(x) = \frac{\Gamma (k+x)}{\Gamma(x+1) \Gamma(k)}p^k(1-p)^k$ for $x=0,1,2,\ldots$.

  1. Give an expression for the expected number of workers in $40$ years from now.


Let $Y_i$ denote the number of firms "created" by a firm after a decade. Then
$$\mu = \mathbb{E}(Y_i) = \frac{k(1-p)}{p}.$$ Let $W_n$ denote the number of workers after $n$ decades and $F_n$ denote the number of firms in $n$ decades. Then we have $W_n = 1000F_n$. Noting that $F_0 = 5$:
$$\mathbb{E}(F_4) = \mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\right]\\ = \mathbb{E}\left[\mathbb{E}\left[\sum_{i=1}^{F_3}Y_i | F_3\right]\right]\\ = \mu \mathbb{E}(F_3) \\ = ... (\text{applying iteratively})\\ = \mu^4\mathbb{E}(F_0)\\ =5\left(\frac{k(1-p)^4}{p}\right)^4$$

I don't understand in particular the second equality with them just bringing out another expectation and introducing a conditional... and then how the 3rd equality follows as well. could someone please explain?

  • $\begingroup$ Are you familiar with conditional expectations wrt a random variable and the general rule $\mathbb E[X]=\mathbb E[\mathbb E[X|Y]]$? $\endgroup$ – drhab Aug 15 '17 at 11:17
  • $\begingroup$ @drhab I haven't formally learnt it, the first time was encountering it in solutions to similar problems and I just accepted it. $\endgroup$ – Twenty-six colours Aug 15 '17 at 11:44
  • $\begingroup$ It is that rule that is applied here. Also see my answer. $\endgroup$ – drhab Aug 15 '17 at 11:54

Maybe this helps:

$$\mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\right]=\sum_{n=1}^{\infty}\mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\mid F_3=n\right]P(F_3=n)=\sum_{n=1}^{\infty}\mathbb{E}\left[\sum_{i=1}^{n} Y_i\right]P(F_3=n)=$$$$\sum_{n=1}^{\infty}n\mathbb{E}\left[Y_1\right]P(F_3=n)=\mathbb{E}\left[Y_1\right]\sum_{n=1}^{\infty}nP(F_3=n)=\mathbb{E}\left[Y_1\right]\mathbb E\left[F_3\right]=\mu \mathbb E\left[F_3\right]$$

It is evident that for every $n$: $$\mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\mid F_3=n\right]=\mathbb{E}\left[\sum_{i=1}^{n} Y_i\right]=n\mathbb EY_1=n\mu$$

This equality can also be expressed as: $$\mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\mid F_3\right]=F_3\mathbb EY_1=F_3\mu$$

In general we have the rule $\mathbb EX=\mathbb E[\mathbb E[X|Y]]$ which is applied on $X=\sum_{i=1}^{F_3} Y_i$ and $Y=F_3$ to achieve:$$\mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\right]=\mathbb E\left[\mathbb{E}\left[\sum_{i=1}^{F_3} Y_i\mid F_3\right]\right]=\mathbb E\left[F_3\mu\right]=\mu\mathbb E\left[F_3\right]$$


I'm confused by the notation of "Y_i", I think it should be number of firms created by a firm in the i-th decade. Then $F_i = Y_0 + ... + Y_i$ and $E(F_i) = E(E(F_i|F_{i-1})) = \mu E(F_{i-1})$. The equality should be $E(F_4)=E(\sum_{i=0}^4 Y_i) = E(E(\sum_{i=0}^4 Y_i) | F_3)= \mu E(F_3)$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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