I'm studying the optimal choice of consumers with regards to taxation.

I read that for consumers, income tax is generally (for Cobb-Douglas preferences) preferred compared to ad valorem tax:

If the budget constraint is generally $p_1x_1+p_2x_2=m$, then an ad valorem tax $t$ will change it to $(p_1 + t)x_1+p_2x_2=m$.

If the income tax is the same amount, we get $p_1x_1+p_2x_2=m-R*$, where $R*=tx_1$ is the amount of taxes for both variants.

Now, the following figure demonstrates how an increased price of good 1 will lead to a preference of an income tax over ad valorem tax: h. varian

I wonder if this is the case for all indifference curves or if there are certain indifference curves where consumers might prefer an ad valorem tax. I doubt this, but I think the chapter in H. Varian lacks a mathematical (algebraic or analytical) explanation for this which I feel would help me understand.

  • $\begingroup$ I have a question. Shouldn't the equation for the budget constraint be this: (1 + τ ) p1x1 + p2x2 = m? The one that you have stated would be for quantity tax, even the picture you have attached says so. $\endgroup$ Oct 7, 2023 at 6:05

1 Answer 1


You question can be answered using a revealed preference argument.

Let $B = \{q \in \mathbb{R}^n_+| p' q \le m\}$ be some budget set of a consumer (i.e. $B$ gives all possible bundles that the consumer can choose).

Let $q^\ast$ be the optimal choice from $B$, i.e. the bundle that optimizes the utility. Then for any other bundle $q \in B$, it must be that $u(q^\ast) \ge u(q)$.

proof: The proof is simple, if, towards a contradiction, we would have that $q \in B$ and $u(q) > u(q^\ast)$, then this would contradict the assumption that $q^\ast$ was optimal in $B$.

Now, let's apply this to the two budget sets for your question.

The first budget set has a lump-sum tax equal to $R$. $$ B^1 = \{q \in \mathbb{R}^n_+ | p' q \le m - R\}. $$ The second budget set has an ad-valorem tax equal to $t$ for some good, say $i$. $$ B^2 = \{q \in \mathbb{R}^n_+| p' q + t q_i \le m\}. $$ Let $q^{1}$ be utility maximising, bundle in $B^1$ and let $q^2$ be the optimal, utility maximising, bundle in $B^2$.

In addition we assume that the revenue of the ad-valorem tax equals the revenue for the lump-sum tax: $$ t q_i^2 = R. $$ We would like to show that $u(q^1) \ge u(q^2)$, so the lump sum tax always gives at least as much utility as the ad-valorem tax.

Using the above revealed preference argument, it suffices to show that $q^2 \in B^1$, which means that $q^2$ was a feasible option when $q^1$ was chosen.

Indeed, we have: $$ q^2 \in B^2,\\ \iff p' q^2 + t q^2_i \le m,\\ \iff p'q^2 \le m - t q^2_i,\\ \iff p' q^2 \le m - R.\\ $$ So $q^2 \in B^1$.

Remark: notice that we did not impose any restrictions on the utility function. The only thing that we assumed was that the consumer was able (from any budget set) to pick an optimal bundle.

  • 1
    $\begingroup$ Awesome. I would have preferred writing $q^*$ (or using letters) though because it might look like q squared otherwise, but that's just a minor thing. You certainly can teach math to econs. $\endgroup$ May 2, 2021 at 12:27

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.