This weird trick will save Congress 1.7 billion dollars a year

In a 2011 letter to a school district, California’s then-Attorney General Kamala Harris decried “the practice of artificially inflating the interest rate” to obtain “additional bond proceeds over and above what the voters authorized.” If only she had known that this practice is subsidized by the federal tax code!

This costly and seemingly inadvertent subsidy is discussed in my forthcoming paper titled Tax distortions and bond issue pricing. The paper does not contain any policy proposals, but in this companion article I propose a couple of ways to eliminate the subsidy.

The paper starts from the observation that bonds with large issue prices are very common in the tax-exempt market, and quite rare elsewhere:


The histogram of issue prices of U.S. Treasury bonds looks like a candlestick: all bonds are issued with a tiny discount to face value (if face value is normalized to 100, a typical list of auction prices will be 99.9, 99.75, 99.82, …).1 Corporate bond prices are equally boring. Taxable municipal bonds? A little more scattered, but still solidly anchored around 100.

And then there are tax-exempt municipal bonds. Tax-exempt munis are routinely issued at large premiums, i.e., with prices like 110, 115, 120. I will not keep up the suspense: issuing premium bonds is a form of tax arbitrage. Compared to the counterfactual of issuing bonds at par, the issuance of premium bonds costs the U.S. Treasury an estimated 1.7 billion dollars a year in lost revenue.2

The one weird trick I would suggest would be to disallow capital losses for tax-exempt bonds issued at premium, unless the price drops below par. For instance, if a bond is issued at 120 and the next day it’s sold at 95, the capital loss is not 120 – 95 = 25, but rather 100 – 95 = 5. The rationale for this approach should become clear soon. If this measure were adopted retroactively (i.e., if it were to apply to losses on bonds already in circulation), substantial savings would likely begin to be felt within the 10-year budget window, especially if interest rates rise. If retroactive adoption is not an option, savings would take longer to materialize, but they could perhaps still help down the road.

A simpler and less weird alternative would be to ban premium tax-exempt bonds altogether. The U.S. Treasury and for-profit corporations almost never issue premium bonds, and it would be easy to argue that there is no good business reason to issue them.3 In fact, the U.S. Treasury has banned premium municipal bonds in the past–except that the ban only applied to Build America Bonds, which are taxable.4 I personally do not recall the reason, but hey, there are many good reasons to ban premium bonds. (Some of them are described at the end of the article). A drawback of this approach, however, is that it cannot be made retroactive because the bonds already in circulation cannot be un-issued.

Premium tax-exempt bonds are tax-efficient

You might be wondering: “How can the Treasury lose revenue? These bonds are already tax-exempt! There is no tax revenue coming from them.” Your perplexity is understandable. It is correct that the Treasury expects zero revenue, but it is not correct to assume that it has nothing to lose. From premium tax-exempt bonds, the U.S. Treasury expects negative revenue.5

To see why, let’s begin by examining the case in which Alice and Bob buy two similar bonds issued at par (i.e., they pay 100 to the issuer). After a short amount of time ε the bond prices drop to 90. If Alice and Bob were to simply sit on their bonds until maturity, they would never realize any gain or loss, and because the interest from the bonds is tax-exempt, they would never pay any taxes.

Instead of sitting on their bonds, however, Alice and Bob agree to a taxable swap: each will sell their bond at a fair market value to the other. As a result, their portfolios will be practically unchanged, but both will have realized a capital loss of 10. Assume this loss is deductible at the capital gains tax rate of 20%, so that it generates an immediate benefit of 2. Upon maturity, having paid 90 for the bond, Alice and Bob receive 100 (the bond’s face value) and thereby realize a gain of 10. This gain is labeled “market discount gain” and it is taxed at the ordinary rate of 40%, so that it generates a cost of 4. The swap and its consequences are represented in the diagram below.

It is not clear whether Alice and Bob gained or lost from the swap. Is it better to never pay any tax, or is it better to get a tax rebate of 2 now, and then pay tax of 4 upon the bond’s maturity? It depends on the investors’ discount rate and on the amount of time until maturity. However, roughly speaking, it would take a very long time to maturity or a very high discount rate for Alice and Bob to be happy about the swap.

Now consider the same exact vignette, except that the bonds are premium bonds. They are issued at 110. At time ε, their price drops to 100. Alice and Bob do the swap and get an immediate benefit of 2. Upon maturity, having paid 100 for the bonds, Alice and Bob receive 100 (the bonds’ face value) and… that’s it. There is no gain or loss. Alice and Bob receive an upfront benefit of 2 from the U.S. Treasury, and never pay any taxes since. Hence, the negative revenue. This tax arbitrage scheme works for all premium bonds, regardless of discount rate and bond maturity.

This scheme is very common, so much so that it features on FINRA’s list of “smart bond investment strategies” [emphasis mine]:

Bond Swapping involves selling one bond and simultaneously purchasing another similar bond with the proceeds from the sale. Why would you engage in this practice? You may wish to take advantage of current market conditions (e.g., a change in interest rates), or perhaps a change in your own personal financial situation has now made a bond with a different tax status appealing. Bond swapping can also cause you to receive certain tax benefits. In fact, tax swapping is the most common of bond swaps.

Now it should be clear what the weird trick does: disallowing losses above par directly eliminates this source of tax arbitrage and takes away the main tax incentive to issue premium bonds.

Alternative explanations

Although tax arbitrage would be enough of a motivation to issue premium bonds, there are other plausible explanations for their issuance. Hilariously, these alternative explanations also involve some sort of gimmick.

Cleveland Fed economists Whitaker and Ergungor argue that premium bonds are issued as a workaround to self-imposed debt limits. Debt limit rules appear to be unaffected by the GASB 34 requirement of accrual accounting, and therefore a premium bond issued at 120 contributes only 100 towards the debt limit. So far this gimmick has only been used by local issuers like the school district mentioned at the top of this article, but in principle it could be used by the federal government as well: Matt Levine proposed issuing premium bonds as a workaround to the United States debt ceiling problem. This factor may explain why relatively many taxable municipal bonds are issued at a premium (10% in 2015) compared to other taxable bonds (less than 1%). However, this factor does not depend on the bonds’ tax status, and therefore it cannot explain why, in the municipal market, tax-exempt bonds are overwhelmingly more likely to be issued at a premium (85%).

Similarly, Andy Kalotay argued that, for callable bonds, a premium issue price creates illusory savings from exercising the refunding option. Many issuers have self-imposed restrictions to the effect that a bond can be called only if the refunding achieves a certain savings threshold. However, refundings can serve as yet another way around borrowing limits. Illusory savings from the refunding of premium bonds fit the legal definition of savings and help get around the restrictions.6 This factor, too, may enhance state and local governments’ incentive to issue premium bonds. However, by definition it only applies to callable bonds. In reality, noncallable bonds are issued at a premium more frequently than callable bonds (94% versus 64% in 2015).

These additional “creative accounting” motives for the issuance of premium bonds are important, but they are not mutually exclusive with the tax arbitrage motive I highlight in my paper. Rather, they are complementary. The tax code contains a loophole that subsidizes the issuance of premium bonds, and this loophole rewards lack of fiscal discipline at the state and local level.

Is Rothification a budget gimmick? It depends on what you do with the money

Note: This post is a work in progress and comments are welcome. Everything that’s wrong in here is my own, and no one else is to blame.

In my paper with Stephen Zeldes I argue that a switch from Traditional tax-deferred to Roth retirement accounts would, over time, remove a $15 billion/year asset management fee subsidy. While we are comfortable with our conclusion (which was also independently reached by Daniel Hemel of U Chicago Law), it is also true that Roth accounts have a notorious history of being introduced or expanded whenever current budgetary proposals are in need of some cosmetic action. In this post I try to understand why Rothification, defined as the increased usage of Roth accounts, is often attacked as a budget “gimmick.”

As part of the recent tax reform debate, Congress has been rumored to have considered proposals ranging from unlikely doom scenarios (the forced conversion of Traditional accounts into Roths) to decidedly mild ones (capping annual contributions to Traditional accounts). These rumors, and the associated fear-mongering, ultimately forced President Trump to issue a strongly worded denial: “There will be NO change to your 401(k).”
Although Trump tweets need not signify a shift in policy, I think it would be socially beneficial to have an honest, long-form debate on the topic. But I digress.

Those who scorn Rothification as a gimmick start by pointing out that a switch from Traditional to Roth should not increase government ability to spend or cut taxes, and then argue that in fact it is being used to do just that. Accordingly, allow me to define a “gimmick” as a budgetary action that permits the current administration to spend more money without either raising new money or a special approval process.1 In what follows I argue that under this definition a switch to Roth is not a gimmick because, although it need not raise any new tax revenues, it does raise money by liquidating an existing asset. If this money were used to pay down Treasury bonds, no one would call it a gimmick. I will also argue that using a one-time asset liquidation to “pay for” a permanent tax cut is an unsustainable gimmick. This gimmick works only because of the rather transparent pretense that the tax cut is temporary.

Rothification is, or could be, revenue-neutral

The fiscal consequences of a switch from Traditional to Roth accounts are best understood by initially assuming that the tax system consists of a simple flat tax: the government will take a fraction τ (tau) of your income. For instance, if your income is $100,000 and τ is 20%, the government will take $20,000. Thus, for now we are ignoring the fact that tax brackets exist, as well as the fact that people may fall in different tax brackets between work life and retirement. Also assume that assets in a retirement account accumulate a net return r. Now consider an individual choosing to allocate one dollar of pretax salary to her retirement account.

  • If the account is Traditional, the government receives no immediate revenue and the individual contributes the whole dollar. Thirty years later, this dollar has grown to (1+r)30 dollars. When the money is paid out, the government nets τ·(1+r)30.
  • If the account is Roth, the government receives τ upfront. Assuming that the government can invest no more and no less successfully than the individual, the future value of this current tax revenue in 30 years is also τ·(1+r)30.

Thus, at a first pass, the future value (or equivalently the present value) of government revenue is left unchanged, making Rothification revenue-neutral.

Of course, real-world taxes are progressive, and many individuals have lower income in retirement than they do during their work life. Rothification would make these individuals unable to shift income from high-tax-bracket years to low-tax-bracket years, and ultimately raise their total tax burden. This is a very important matter, but not the subject of this article. There are many ways to counter this effect, leaving both the total tax revenue and the distribution of tax burden roughly unchanged. For instance, the government could accompany a switch to Roth with either a reduction in low-bracket tax rates, or an increase in bracket cutoffs. Or, Hemel suggests, one could simply equalize the ex-post tax burden with a one-time charge or credit at the time of withdrawal. Still, suppose that the government does not act to counter the effect of Rothification (e.g., it cuts high-bracket tax rates, or no tax cuts at all); the effect is to bring in real revenue via a real reduction in the progressivity of the tax code. Where is the gimmickry?

Rothification relaxes the budget constraint

The basic argument goes as follows. The U.S. Congress has some rules in place that make it more difficult to pass bills that increase the budget deficit.2 However, since budget bills usually cover a 10-year window, a policy that increases revenues inside the 10-year window at the expense of future revenues outside the 10-year window appears to “produce revenue” and it can then be used to “pay for” other deficit-inducing measures such as a tax cut.

This intuitively appealing explanation leaves something to be desired. Is Congress procedure that easy to game? And if it is, why bother with Rothification when Congress has at its disposal a myriad other forms of budget gimmickry that are less labor-intensive and much, much less politically controversial?

The thing is, Congress procedure is not that easy to game. (Surprising, I know). In particular, there are two U.S. Senate rules that merit discussion.3 The basic gimmick argument seems to rely on the “PAYGO rule,” which requires a more cumbersome approval procedure for measures that increase total deficit in a 1,-, 5-, or 10-year window. However, the “Byrd rule” (a set of restrictions championed by former Senator Robert Byrd) also requires a more cumbersome approval procedure for bills that increase the deficit during a fiscal year after the fiscal years covered by the bill. The usual budget bill covers 10 years, and therefore the two rules combined require that a bill create no new deficits, period.4 Thus, it is not obvious that shifting revenue from the far future into the near future would relax budget restrictions.

And yet, every single time the introduction or expansion of Roth accounts has been proposed in the past, opponents have pointed to some other measure whose cost was being illusorily paid for with more Roth. The 1989 Packwood-Roth proposal would pay for capital gains tax cuts; the actual introduction of Roth IRAs paid for some of the tax cuts in the 1997 Taxpayer Relief Act;5 most recently, the 2009 introduction of a Roth option for the federal employees’ Thrift Savings Plan paid for certain other federal employee benefits. Why does Roth work?

The magic of Rothification is that it shifts revenues from the far future into the near future without creating a budget deficit in the far future, and therefore it does not run afoul of the Byrd rule. The reason why there is no deficit in the far future is that Rothification also shifts revenue from the really far future into the far future. And from the really, really far future into the merely really far future. And so on, ad infinitum.6

Is this the best, subtlest Ponzi scheme ever conceived? Or, since the generation that will pay the eventual price of Rothification does not exist, is this actually not a gimmick at all?

A Rorschach test for fiscal hawks

The answer is subjective. When individuals contribute money to Traditional accounts, the government is voluntarily forgoing some current tax revenue in exchange for the “promise” that it will collect this revenue in the future. This forgone revenue accumulates as a deferred tax asset; the government has a claim on some 20-25% of the current value of all tax-deferred retirement accounts. If we include individual accounts (IRAs), employer-sponsored defined contribution plans (401k, 403b, TSP, etc), and employer-sponsored defined benefit pension plans, this deferred tax asset is worth about $4 trillion. While the government does include these future revenues in its cash-based financial statement projections, the asset does not appear on its balance sheet, nor is it ever explicitly talked about.

If you are a fiscal conservative, you may like the idea of the government squirrelling away money in its own virtual “retirement account,” putting pressure on current budgets and inducing some degree of fiscal responsibility. To discontinue this practice and enjoy the temporary benefits of the savings accumulated by past administrations is worse than a gimmick—it is a reckless, irrational act.

On the other hand, this practice itself may well be construed as an irrational, reckless self-imposed restriction on spending—a gimmick! The government is pretending to be poorer than it really is, as yearly increases in the value of the deferred tax asset do not show up as government surplus. This pretense can do real damage: vital spending programs are being forgone and taxes are too high. Moreover, ignoring the existence of a $4 trillion asset makes government indebtedness appear higher than it really is. In perfect, frictionless financial markets, only net indebtedness would matter; a government that expects more revenue later can simply borrow more now. But markets are not perfect or frictionless, and cash is king. If government debt buyers pay undue attention to headline debt figures, a higher apparent figure may be a threat to financial stability. As of December 2016, the U.S. public debt stood at $20 trillion or 108% of GDP. If we net out the $4 trillion of deferred tax asset, however, the figure drops to approximately 87%, safely below the Reinhart-Rogoff threshold of 90%.7

It depends on what you do with the money

There is one last missing piece of the puzzle.

Regardless of whether you are a hawk or a dove, I hope I convinced you that Rothification creates real, one-time fiscal slack by liquidating an existing asset that the government has accumulated over the years. The slack would be felt over a long transition period in which the deferred tax asset slowly evaporates until the last Traditional account is liquidated.

Whether you think the idea of liquidating government assets is good or evil depends on other factors, beside your degree of fiscal conservatism. (Not to mention that fiscal conservatism is highest when one’s party is in the opposition). Perhaps the most important factor is what is done with the money. If the government used the Rothification windfall to pay down debt, I suspect no one would call it a gimmick. Similarly, if the government were to use the money to build a vitally important new bridge, it would simply be substituting one asset for another. Moreover, the new asset (the bridge) is something more government-worthy than the old asset (stocks, bonds, and everything else that individuals hold in their retirement accounts).

Unfortunately, the typical motive for more Roth is the need to pay for temporary tax cuts. To avoid running afoul of the Byrd rule, these tax cuts must expire after 10 years, so that their budget impact disappears in the years beyond the budget bill. When the tax cuts are about to expire, some other politician—possibly from the other party—will face a “fiscal cliff,” and will have to make an unpopular decision between letting the tax cuts expire (“raising taxes”) and finding another way to pay for the unsustainable but now-permanent reduction in revenue (“slashing spending”). It is a politician’s dream—the gimmick of a lifetime.