This may be a silly question, but, for a random variable $X$ defined on reals, I am wondering if the existence of the finite fourth central moment $E[(X-E[X])^4]$ implies the existence of the finite third absolute moment $E[|X-E[X]|^3]$? Formally, does the following statement hold:

$$E[(X-E[X])^4]<\infty\Rightarrow E[|X-E[X]|^3]<\infty$$

The reason for my question is that, while trying to apply the Berry-Esseen Theorem to bound the total variation distance between the distribution of the (appropriately normalized) mean of $n$ i.i.d. random variables and the standard normal distribution, I am finding that the fourth central moment of the i.i.d. random variables the distribution of whose average is being approximated is much easier to compute than their third absolute moment...

  • $\begingroup$ This works because the measure of the space is finite. When integrating over spaces of infinite measure, this doesn't hold. $\endgroup$ – Michael Hardy Aug 10 '13 at 17:28
  • $\begingroup$ I am not that familiar with measure theory, but all probability measures are finite, correct? And thanks for fixing the typo in my question. $\endgroup$ – M.B.M. Aug 10 '13 at 17:41
  • $\begingroup$ Probability measures are finite. But in some other contexts than probability theory one integrates over spaces of infinite measure. $\endgroup$ – Michael Hardy Aug 10 '13 at 23:45


Proof: Apply the inequality $|x|^3\leqslant1+x^4$ to $x=X-E[X]$ and integrate.

  • $\begingroup$ This is a very pretty answer, thanks! $\endgroup$ – M.B.M. Aug 10 '13 at 16:25
  • $\begingroup$ Is there a way to bound $|x|^3$ by a polynomial involving $x^4$ and $x$? I just wonder how you come up with these inequalities. $\endgroup$ – Glassjawed Dec 7 '13 at 0:04
  • $\begingroup$ @Glassjawed $|x|^3\leqslant x^4+|x|$. Just draw the graphs of all these functions on, say, $(-5,5)$. $\endgroup$ – Did Dec 7 '13 at 7:12

Split the random variable into the following regions

  1. $ |X - E[X]| > 1 $. In this region, $|X-E[X]|^3 < |X-E[X]|^4$

  2. $|X - E[X]| \leq 1$. In this region, $|X-E[X]|^3 \leq 1 $.

Hence conclude that $E\left[|X-E[X]|^3\right] < E\left[|X-E[X]|^4\right] + 1$


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.