A person’s optimal choice depends on what they know. To consume more ice cream? Or to consume more alcohol? It depends on what we know about the expected utility across time. If a person thinks that alcohol has few calories, then it is understandable that they would choose to drink rather than eat. The person might be totally wrong, but they are acting optimally contingent on their knowledge about the world. (FWIW, 4oz of ethanol has 262 calories and 4oz of typical ice cream has 228 calories.)
The case is analogous for good government policy. The best policy is contingent on accessing the distribution of knowledge that’s inside of multiple people’s heads. It’s not sensible to assert that a policy is suboptimal if the optimal policy requires knowledge that neither a single individual nor all people together have. Even if the sum of all knowledge does exist, it may not be possible to access it.
Economists like to tell their undergraduate classes that it doesn’t matter who you tax. But that’s contingent on 1) identical compliance costs among buyers and sellers and 2) identical relevant information. If a tax comes as a surprise to the buyer or the seller, then it absolutely matters who is taxed.
When I was in 1st grade in North Carolina, my class went on a field trip to a Christmas tree farm. We learned a bunch about maintaining the farm and we got to choose a pumpkin to take home. At the end of our visit we took turns perusing the gift shop. My mother had generously given me a dollar to spend and I was eager to spend it (I rarely had money to spend). Unfortunately, even in the early mid-90s, most of the things in the shop cost more than $1. So, I settled on purchasing some beef jerky that cost 99 cents.
When I attempted to pay at the cash register, the cashier told me that my balance was $1.05. Given my budget, I was quite confident that she had made a mistake and I politely told her that, ‘no, the price definitely says 99 cents’. She told me that there were taxes and that the balance was $1.05. Being a bright 1st grader, I agreed that ‘yes, there are taxes’ – I had no reason to disbelieve her. ‘But, the price posted is clearly 99 cents’.
Can you picture it? A first grader, politely but firmly telling this 50-year-old woman that the receipt balance was wrong? In my mind, the existence of taxes was consistent with the evidence. Surely, the gift shop would need to pay those taxes at some point. I didn’t understand that the posted prices did *not* include all of the taxes already. The cashier ended up waiving my balance due and I am thankful for that. And I learned a lot on that day.
Had I known that I would have had to pay more than the sticker price, then I would not have made the purchase that I did. That is, if I had known better how the world works, then I would have behaved differently. There is some evidence that this might vary systematically in some markets (ungated).
The knowledge contingency of optimal policy is analogous to that of optimal consumer choice. To stick with the sales tax example, the modern ‘retail sales’ tax wasn’t introduced until 1947. This type of tax is calculated at each point of sale and added to the bill balance. An excise task, on the other hand, is a tax that is levied on the total sales of a good by a producer. The difference is in the de jur payer of the tax. A retail sales tax may require that 100 buyers pay a tax while the excise tax requires that 1 seller pay the tax.
Which is better? In the year 1900, cash registers could not automatically calculate the tax. Each transaction would need to include an additional time and effort intensive calculation. It was less costly in aggregate for a single firm to calculate the tax on their total sales rather than calculate the tax due from many individual transactions. Due to the technology at the time, the optimal policy was to tax firms rather than consumers. Now, with the low cost of quick computing power, it probably doesn’t make much of a difference. The transaction cost difference is minimal.
Let’s take this to the next level. In the USA, individual earners or households must file their own income tax calculations with the state or federal government. Given the diversity of tax understanding and computation costs, this is a very costly procedure (in excess of $100 billion for just income taxes). If this were 1990 and computing power was still in its adolescence, then I’d understand. But that’s not where we are. It’s totally possible for the government to incur the calculation cost for all tax payers. I suspect that the cost to do so would be well under $100 billion. The policy of government calculation is contingent on the cost of central vs distributed computing power. If the computing is cheaper for a single entity rather than many entities, then the technology contingent optimal policy is central calculation. If central calculation is too expensive, then the optimal policy is distributed calculation and the associated distributed costs.
If policy *is* optimal, which it isn’t, then what can we infer? It must mean that there are other costs that are born by the IRS rather than just computing costs. I suspect that, despite all of the compliance of firms and all of the W2 and 1099 forms that are submitted to the IRS, that the IRS doesn’t not have a handle on the numbers for each household. That is, if the IRS did economize on computation cost centrally, then they would probably neglect portions of your income that are substantial. If substantial enough, then the calculation savings would be outweighed by the costs of foregone tax revenue from misreported or unreported income.
The IRS doesn’t have access to adequate knowledge that would allow it to internalize the benefits of the change in policy. It’s better for IRS revenues to place the cost of compliance on the filer. If filers were scofflaws in general, conducting substantial and widespread tax evasion, then maybe the IRS would change. At that point, the unreported incomes aren’t reported even when individuals perform their own calculations.
That the IRS externalizes the cost of calculation and that sales taxes weren’t levied at the point of sale until recently are both due to the contingency of optimal policy on technology. We pay property tax on our homes, land and vehicles and not on TVs, computers, and appliances for similar reasons. Our current state of technology permits tax authorities to easily observe and evaluate the former and not the latter. Taxing the large items that one owns is an old type of tax because the technology that it demands is relatively rudimentary. The fact that we don’t observe states taxing many other types of property is not due to their understanding of tax avoidance or the inefficiency of taxing wealth. It is due principally to the fact that our current level of technology dictates that taxing all property would excessively costly to administer.
Economist Writing Every Day 