A previous post revealed an error in the textbook. Now I'm beginning to doubt another closely related claim:
$(x+1)^p \equiv_p x^p + 1$ for all $x \in \mathbb{Z}$ if and only if $p$ is prime.
I'm pretty confident in the argument in one direction. Using the Binomial Theorem we write $(x+1)^p$ as $$ \sum_{k=0}^p \binom{p}{k} x^k $$ Then any coefficient other than the first and last is of the form $$ p \frac{(p-1)!}{(p-k)!\,k!}\,. $$ Because the whole coefficient is an integer the fraction must either also be an integer or a multiple of $1/p$. The latter can't be since, as long as we're not dealing with an edge case, neither factorial on the bottom contains a $p$. Thus, the fraction must be an integer. And that leaves us with only the terms $x^p$ and $1$ as we'd hoped for.
In the other direction one can show following similar lines of reasoning that if $p$ isn't prime at least one intermediate coefficient isn't a multiple of $p$. However, as Ethan MacBrough pointed out in a similar case here that's not enough to conclude the argument. It may be that by some conspiracy the intermediate terms still cancel for all $x$.
That raises the question: did the textbook at least get this one right?
This statement is also false. A number for which this is true is called a Carmichael number. Non-prime Carmichael numbers are somewhat rare, but they exist. The smallest is 561.
To see the equivalence between the stated property and that of being a Carmichael number, suppose $n$ is a Carmichael number, so by definition we know for all $x$ we have $x^n\equiv x$. Thus in particular $(x+1)^n\equiv_n x+1\equiv_n x^n+1$.
Conversely, if $n$ satisfies the stated property then we can show that $n$ is a Carmichael number by induction. Suppose we've proven that $(x-1)^n\equiv_n = x-1$. By the assumed property, we have $x^n=((x-1)+1)^n\equiv_n (x-1)^n+1$, and by the induction hypothesis this is equivalent modulo $n$ to $x-1+1=x$. The base case where $x=0$ is trivial.