Thursday, May 24, 2018

A tax, not a ban, on high denomination banknotes

Ken Rogoff has famously called for a ban on high denomination banknotes in order to help combat tax evasion and hurt criminals. But rather than banning notes, why not implement a market-based approach such as a tax? Among other advantages, a tax leaves people with flexibility to determine the cheapest way to reduce their usage of the targeted commodity. This is how society is choosing to reduce green house gas emissions. So why not go the tax route for banknotes too?

My recent post for the Sound Money Project on pricing financial anonymity delves into this idea. The anonymity provided by banknotes is both a "good" and a "bad". People have a legitimate demand for financial alone time; a safe zone where neither their friends, family, government, nor any other third-party can watch what they are buying or selling. These days, cash is pretty much the only way to get this alone time.

But cash's lack of a paper trail can be abused when it used to evade taxes. The resulting gap in government finances forces the honest tax-paying majority to pay more than their fair share for government services. This state of affairs isn't just.

One way to fix this inequity is to raise the price of banknote usage high enough so that it includes the costs that tax evaders impose on everyone else. A tax on banknotes, call it a financial privacy tax, can do this. It internalizes the externality, or the harm done to others. 


Interestingly, financial privacy taxes already exist. For each banknote that it has issued, a central bank typically holds a risk free interest-yielding asset in its vault. In a free market, this interest would flow through to banknote holders, say by the implementation of note serial number lotteries. Rather than allowing the interest to flow through, however, the central bank withholds it. The amount it withholds constitutes the financial privacy tax.

In Canada, for instance, the overnight risk-free interest rate is currently 1.25%. The yield on banknotes being 0%, the Bank of Canada is withholding $1.25 in interest payments for each $100 bill held. So a note-using Canadian is effectively being taxed $1.25 year for each $100 worth of financial privacy he or she chooses to use. Anyone who wants to avoid the tax need only deposit the note into a bank account and earn 1.25% per year.  But once they do that, they will be giving up their privacy.


Modern taxes on banknotes aren't consciously designed as financial privacy taxes. By that I mean, it's not like central bankers have sat down at a conference table and thought long and hard about the costs and benefits of anonymity only to settle on the most appropriate level for the tax. Rather, the size of the tax has been arrived at by accident. Historically, central bankers have simply assumed that it was technologically impossible for banknotes to yield anything other than 0%. (Fully adjustable interest rates on notes, both positive and negative, are actually quite easy to implement, as I'll show). Which means by default, the privacy tax has always been at least as large as the foregone overnight interest rate.

The overnight rate is in turn a function of an entirely different thought process: monetary policy. Central bankers ratchet the overnight rate up or down in order to to hit their chosen inflation target. The problem with this setup is that two separate decisions have been jumbled together. The level at which the central bank sets its financial privacy tax has become the ill-conceived byproduct of its chosen macroeconomic policy.

Here's an example of this muddle. If the Bank of Canada decides to tighten monetary policy tomorrow by increasing its interest rate from 1.25% to 1.5%, it has simultaneously made an entirely separate decision to increase the privacy tax on banknotes by 0.25%. But whereas the monetary policy decision is guided by plenty of data and number crunching, the increase in the privacy tax is purely arbitrary—no thinking has gone into justifying an increase. It's a fait accompli.

Or think about it from another angle. Say that the Bank of Canada has determined that it is appropriate to increase the financial privacy tax by 0.25%. Using its current toolkit, the only way it can accomplish this is by increasing the overnight rate by 0.25%. But this tightening of monetary policy could potentially send the entire economy into a tailspin, all for the sake of satisfying an entirely different policy goal, that of setting the appropriate tax on privacy.

There's no reason that the two decisions can't be split up. The tool that would allow central bankers to do this is the ability to pay positive and negative interest rates on banknotes. I talked about note serial number lotteries as one way to pay positive interest here. Later on in this post I'll discuss a way to pay negative interest. To see how these tools could successfully split the monetary policy decision from the privacy tax decision, let's return to our previous example. If the Bank of Canada were to increase the overnight rate for monetary policy purposes from 1.25% to 1.5%, but it did not want to alter the financial privacy tax, it could simultaneously increase the interest rate on banknotes from 0% to 0.25%. The original 1.25% privacy tax stays intact. While the owner of a banknote is now forgoing the 1.5% overnight rate, he or she is also collecting 0.25% in interest.  

Conversely, these tools would allow the privacy tax to be increased or lowered without requiring a potentially damaging change in monetary policy. Using our example, to increase the privacy tax from 1.25% to 1.5% per year while keeping monetary policy constant, for instance, the Bank of Canada would move the interest rate on banknotes from 0% to -0.25% while keeping the overnight rate at 1.25%. So a banknote owner is now taxed 1.5% per year, of which 1.25% is due to the forgone overnight rate while the other bit is the 0.25% negative interest rate. This has been accomplished without any tightening or loosening of monetary policy.

So there you go, the monetary policy decision has been split from the privacy tax decision. The advantage of having the ability to split up these two thought processes is that it is now possible to think long and hard about what the proper privacy tax rate should be.


One question we might ask is if the current financial privacy tax on banknotes is sufficiently high. Remember, the problem we are trying to solve is that a small group of citizens are not paying their fair share of income taxes by taking advantage of the untraceability of banknotes. The 1.25%/year financial privacy tax on banknotes that is currently being imposed by the Bank of Canada may not be enough to recoup the damage that this untraceability is doing to everyone else. Maybe we need a 2% tax on banknotes, or 5%, or 10%.

Say we increased the Canadian financial privacy tax rate from 1.25% to 5%. (For now let's not be too concerned about how the tax gets levied. I'll get into that later). One unfortunate side effect is that licit users of banknotes—the unbanked and those who want financial alone time for reasons other than evading taxes and crime—would be caught up in a tax net that is intended for illicit users. This doesn't seem very fair. Might there be a finer sorting mechanism that allow us to tax crooks while letting non-crooks through?

In his controversial book on banning high-denomination notes, Ken Rogoff has proposed exactly this sort of fine sorting mechanism. Based on the assumption that criminal usage of bills is largely confined to high-denomination note, he proposes that only $100s, $50s, and maybe $20 bills be banned. We are interested in a tax in this post, of course, not a ban. But if Rogoff's assumption about criminal usage is right, then a graduated tax on banknotes might be a better option than a flat tax, with higher denominations facing a more aggressive levy than low denomination notes.

All central banks currently tax the full range of banknote denominations at the same rate. In Canada's case, the 1.25% tax rate that is currently applied to a C$1000 bill (yes, we have them in Canada, see top) comes out to the same amount incurred by one hundred $10 bills: $12.50 per year. But a $1000 note is far better for evading taxes because it contains more anonymity services per gram than a $10 note. After all, a bag full of tens is bulky and visible, an envelope with a few $1000 bills isn't.

Given the outsized anonymity provided by the $1000, perhaps we should keep the 1.25% tax rate on $10 bills but boost the tax rate on $1000s to (say) 12.5%. A tax evader who holds a $1000 bill would now incur a tax of $125 instead of just $12.50 while a regular Joe with just a few $10 bills would see no increase in banknote-related taxes. (Heck, it might even be a good idea to reduce the tax on small notes to zero.) By boosting the tax on high denomination banknotes, the Bank of Canada enjoys a larger revenue stream than before. Which means that at least some of the revenue gap due to tax evasion can now be plugged, thus fixing some of the damage inflicted on honest tax payers.


How would we go about increasing the financial privacy tax on high denomination notes?

Central banks currently have a policy of maintaining perpetually fixed exchange rates between various note denominations. Your $10 bill is always convertible into ten $1 bills, and your $100 into ten $10 bills. But this needn't be the case.

To implement the tax, central banks would begin to vary the exchange rate between banknotes. Let's take the U.S. as our example. Instead of redeeming the $100 bill at par, the Federal Reserve would slowly reduce the rate at which it redeems the $100 over time. This ratcheting down of the price of $100s would be passed off to the cash-using public in the form of a capital loss, this capital loss functioning as a tax. (For those with long memories, this is basically Miles Kimball's crawling peg idea, applied to the idea of financial privacy rather than evasion of the zero lower bound).

Let's work through an actual example. Say that the Fed wants to impose an extra 5% financial privacy tax on the $100 bill, but not on other bills. It sets December 31, 2018 as the last day that it will redeem a $100 bill for either: a) $100 worth of central bank deposits; or b) $100 worth of bills in $1s, $5s, $10s, $20s, and/or $50s. On the first day of the new policy—January 1, 2019—a $100 bill can be redeemed for a tiny bit less, say $99.93. Daily reductions continue so that by the end of 2019, the Fed will have scaled its redemption rate back by 5% to $95.

This means that if you deposit a $100 banknote at your bank on December 31, 2019, your bank in turn depositing said note at the Fed, the Fed will credit the bank with just $95 in deposit balances, not $100. In anticipation of this, your bank would have only credited you with $95 when you initially deposited the note. Voila, a financial privacy tax. Everyone holding a $100 note for any period of time will have incurred an 5% annualized tax. But if you hold twenty $5 bills, the tax is avoided.

Continuing with our example, by the end of 2020 the Fed's redemption rate will have declined by another 5% to $90.25. And by the end of 2021, the $100 would be worth $85.74, and on and on.

At some point things start to get a bit silly. By 2031, the market value of the $100 will have fallen below the $50 bill, and by the the first decade of the next century it will be worth less than the $1. To prevent this inversion, the Fed will at some point—say in 2026—demonetize the old issue of $100 bills and introduce a new $100 bill, resetting its market value at $100. The whole process of steady reductions starts anew.


This post has been a heavy one, so I'll just quickly summarize it before signing off.

The anonymity provided by banknote is often abused, but rather than banning notes why not tax the abusers? We wouldn't have to start from scratch. Banknotes yield 0% when overnight rates are positive, so society is already imposing a financial privacy tax of sorts on notes. Unfortunately, central banks set the privacy tax arbitrarily, as the unplanned by-product of monetary policy. New tools for increasing /decreasing the return on banknotes could facilitate a separation of the two decision-making processes. These tools could be used to set a higher tax on large denomination notes while leaving smaller notes untaxed, the true costs of anonymity being recognized for the first time.

P.S. If you're interested in this topic, David Birch has a good post on Austin Houldsworth's Crime Pays System or CPS. It's sort of tongue in cheek, but also quite relevant:
"During this talk, ‘Mr Rogers’ proposed the Crime Pays System, or CPS. Under this system, digital payments would be either “light” or “dark”. The default transaction type would be light, and free to the end users. All transaction histories would be uploaded to a public space (we were of course thinking about the bitcoin blockchain here), which would allow anybody anywhere to view the transaction details. This type of transaction is designed to promote an environment of social accountability.
The alternative transaction type would be dark. With this option, advanced cryptographic techniques would make the payment completely invisible, leaving no trace of the exchange, thus anonymising all transactions. A small levy in the region of 10-20% would be paid per transaction. The ‘Dark Exchange’ would therefore offer privacy for your finances at a reasonable price.
The revenue generated from the use of this system would be taken by the government to substitute for the loss of taxes in the dark economy."
Another worthwhile source is Josh Hendrickson's recent paper "Breaking the Curse of Cash" (written along with Jaevin Park). It's a pretty technical paper, but it explores a model in which coins and paper money circulate, but coins are a burden for illegal traders to use because they make noise, leading to detection.
"If illegal traders impose an externality on society,  the government can generate seigniorage from the illegal traders by setting low rate of return on paper money and providing transfers to legal traders by setting high rate of return on coins. Then the amount of illegal trade is reduced while the amount of legal trade increases. This is a standard solution to an externality problem."


  1. "The advantage of having the ability to split up these two thought processes is that it is now possible to think long and hard about what the proper privacy tax rate should be."
    If Scott Sumner is right, the central bank has at least three "normal" monetary policy levers: changing M0 (via OMOs), the overnight rate and IOER. By judicious application, they can already set the "banknote tax" (the overnight rate) independently of monetary policy. However, this is flat, giving the same rate on $1 and $100 notes.

    "But a $1000 note is far better for evading taxes because it contains more anonymity services per gram than a $10 note."
    I think it would be more damning to say that a $1000 note contains more anonymity service *per dollar* than a $10; hence more than 100 times as much anonymity per weight.

    "At some point things start to get a bit silly. [...] To prevent this inversion, the Fed will at some point—say in 2026—demonetize the old issue of $100 bills and introduce a new $100 bill"
    In my opinion, it would get silly at the moment a bill with "$100" printed on it would be legally worth not-$100. To introduce this scheme, the Fed would have to demonetize existing $100 notes and introduce a new "Franklin" with no fixed face value printed on it. You explain this scheme here:

    1. Hi Basil,

      "By judicious application, they can already set the "banknote tax" (the overnight rate) independently of monetary policy"

      I'm not so sure of that. The tax on banknotes (A) is equal to the overnight rate (B) minus the interest rate on banknotes (C), which is currently 0%. For A to be changed independently of monetary policy, only C can be changed. Any alteration to IOER or OMOs will cause B to rise or fall, and the resulting change in A won't be independent of monetary policy.

      "To introduce this scheme, the Fed would have to demonetize existing $100 notes and introduce a new "Franklin" with no fixed face value printed on it. "

      I didn't think about it, but you're right, that's one way to go about it. As you say, that way there is no divergence between the face value of the note and its market value. The drawback is that it requires an extra demonetization in order to get the blanks into circulation, and demonetizations are expensive.

    2. You are right, by the Fisher eq. this can't be done. Where I was confused is that I thought that liquidity returns must be equal between participants, whereas actually overall returns need to be equal. QE+IOER lowers the liquidity return *for banks* (to overnight-IOER), but not for others.