Tuesday, November 13, 2018

Bad economics of the stamp duty discourse

This is an expanded version of an article first published at The Conversation

Most think-tanks and policy groups in Australia think that a policy to swap ‘stamp duties for land value taxes’, or what I will call SD4LVT, is going to provide some impressive economic and housing benefits.

But layers of bad economics hide beneath the new populism of SD4LVT.

It is extremely frustrating to me that leading minds in Australian policy have smashed their heads together and decided that the best reform they can think of is to replace one good tax on property with another good tax on property that is less politically popular. It's like subbing off the best player on the field for the best player on the team? Why not sub off the worst player instead?

The better minds in this game don’t even think SD4LVT will have any effect on reducing housing prices or making housing at all more affordable, yet still put it forward as the ‘holy grail’ of state tax reform.

Is this some kind of joke?

The bad economics behind SD4LVT arises because four key points are missing or overlooked.

1. Price effects

The economic incidence of stamp duty is on the seller. This means that if you remove stamp duty, all else equal, prices will rise by exactly the amount of the duty. Good analysts know this, even if they rely on poor analysis by others that assumes stamp duty “increases the price that property buyers pay”. If the average home price is $500,000, and stamp duty is 5%, prices will immediately rise to $525,000 when you remove the duty.

But if you want to replace the same tax revenue from stamp duty with revenue from land value taxes, you could wind up inflating prices and inadvertently creating an economic transfer to property owners.

Imagine an example economy with:
  • 20 houses 
  • Turnover of 1 house each year (a turnover rate of 5%) 
  • Average rent of $25,000 per year (5% gross yield) 
  • An average home price of $500,000 
  • Land price component of $250,000 per house (total land value of $5 million) 
  • Total stamp duty revenue of $25,000 per year. 
To replace the $25,000 stamp duty revenue with a land value tax requires taxing all twenty homes at $1,250 per year each.

Whether the market price of homes rises or falls depends on whether buyers think the upfront present value of $1,250 per year is lower or higher than the $25,000 upfront stamp duty they would have paid. At a 5% capitalisation rate in this example, $1,250 per year in perpetuity is equivalent to $25,000 and the taxes replace each other perfectly with no price effect.

But what if turnover is half that, say 2.5%, which in this example would be where one house is sold every two years? Here, the total tax revenue to be replaced is $12,500, which equals just $625 per house in land tax.

If we capitalise the perpetual cost of $625 per year at 5% we get a present value of only $12,500, which is half the stamp duty rate. A new buyer can now pay $512,500 for the house plus $625 per year in land tax and be equally as well off as paying $500,000 for the house plus $25,000 upfront in stamp duty. The net effect in this case is a land price increase from $250,000 to $262,500, or 5%.

If this situation happened nationally in Australia, that would be nearly a $200 billion economic transfer to landowners, which would completely offset the value of their new land tax obligations.

In short, replacing stamp duty revenue with land tax will increase prices if the capitalisation rate is higher than the turnover rate, and decrease prices if the opposite holds. Price effects are ambiguous for SD4LVT.

2. Asset churn

Taxing capital gains only happens when a transaction is made that realises gains. Thus, it is a transaction tax, just like stamp duty, though the size of the tax is related to historical changes to property values rather than current values.

Many in the housing discourse, such as the Grattan Institute, want to increase transaction taxes on housing by increasing capital gains taxes and expanding their scope to apply to owner-occupiers. At the same time, they want to decrease transaction taxes on housing by eliminating stamp duty because apparently transaction taxes stop people relocating.

Somehow abolishing a transaction tax in the form stamp duty is good because it reduces the cost of owner-occupiers relocating, but then putting back a large transaction tax in the form of a capital gains tax specifically on owner-occupiers is also good. Huh?

There are two problems with the ‘transaction costs impede household mobility and are bad’ view. First is the simplest. People who relocate for work generally don’t buy and sell houses to accommodate that move. 
Moving for work is not a common reason for buying a home, regardless of age, which might reflect a view by some households that work is a temporary reason for moving and therefore not sufficient to commit to homeownership.
Thus, the effect of stamp duty on reducing housing turnover is likely to fall mostly on investors, who make up nearly half of housing transactions and can easily time their decisions. Transaction taxes make it more expensive for people to quickly buy and sell and deters speculative buying that seeks only to capitalise on short bursts of capital growth before selling, which itself fuels the capital growth and accentuates the bust. Many investors fall in this category, and many homeowners also make their location and purchase decisions based on expectations of capital gains rather than housing need. Go on, ask new homebuyers about property prices. See if they bought where they did for housing need or capital growth?

The evidence on stamp duty deterring moving shows pretty clearly that moving for work is unaffected, as the below plot shows, with the main effect being from people making 'short moves' by relocating less than 10km. Indeed, the best evidence on turnover effects estimates only the size of the short-term shock as people adjust to the new price equilibrium.

Regardless, turnover as a whole seems relatively unaffected by stamp duties. The early 2000s boom also saw massive stamp duty increases yet turnover shot up dramatically. The role of stamp duty in these patterns of housing turnover is tiny compared to other factors.

Indeed, if you want people to move more frequently for a lower cost then implement a ‘moving house subsidy’, which could also apply to renters. It would also help renters forced to move unnecessarily due to investor turnover (yes, stamp duty also reduces undesirable moves for renters).

Second, taxing capital gains on owner-occupied housing will be a far bigger transaction tax than stamp duty for most homeowners. Stamp duties are typically around 5% of the property price. For a homeowner who bought a home prior to 2012 in Sydney or Melbourne, their home has appreciated in value about 50%, or to put it another way, 33% of the current price is capital gain.

If a capital gains tax applied to their sale, the tax rate as a proportion of the current value would be:

Capital gain (33%) x Marginal tax rate (45%) = 14%

This is triple the stamp duty for most properties. Even if the capital gains tax was discounted 50%, the nearly 80% of homeowners who bought prior to 2012 would still have much higher transaction costs from capital gains taxes than from stamp duties.

In effect, stamp duties are taxing some of the capital gains during a boom that would otherwise accrue to homeowners, particularly owner-occupiers who are exempt from the capital gains tax. This seems like a good thing to me.

So while stamp duties are often claimed to be economically costly because they deter households relocating, their main effect is actually reducing investor churn, which is a good thing for property market stability. Additionally, the costs of relocating are probably less than the costs from other property transaction taxes that opponents of stamp duty support.

3. Tax revenue stability 

Another concern in the discourse is that “stamp duty revenues are much more volatile than other taxes.” During a boom, revenues rise more than proportionally to prices since they depend on prices and turnover. Vice-versa in a bust.

If I was to think in the abstract about what sort of taxes are good macro-economically, I would say those that are pro-cyclical, meaning they automatically increase tax revenues during an economic boom, and decrease them in a bust. On this metric stamp duty is a terrific tax to help stabilise the economy, which is even more useful in Australia as our economic cycles are closely tied to the housing market.

The land value tax that many propose to replace stamp duty would have smaller automatic stabiliser characteristics, meaning the tax system as a whole would be less stabilising.

From what I can tell the call for more stable tax revenues is not driven by economic reasoning at all. At best it appears to rest on some kind of political preference for state governments not to be involved in macroeconomic objectives.

4. Wrong assumptions 

The metrics economic disaster caused by stamp duties are usually derived from economic analysis using computational general equilibrium (CGE) models of the macroeconomy. Sounds fancy.

You might have even seen a chart similar to the one below, with stamp duties presented as having massive flow-on economy-wide costs compared to other taxes. In this example, stamp duty is apparently forty times more economically costly per dollar of revenue than council rates, which are also levied on property values.

As you might have guessed, this is absolute nonsense. CGE models cannot account for transaction taxes because, wait for it, there are no transactions in these models. It’s true. Look it up.

But that hasn’t stopped many economists. Instead of using a better tool for the job, they simply invent a different thing that they pretend is stamp duty and put that into a CGE model.

Rather than a transaction tax that is incident on the seller, and therefore incident on land values, a key layer of bad economic analysis assumed this instead (p125).
conveyancing stamp duties are modelled as a tax on investment in residential and commercial structures
Against all the evidence they assume that stamp duty is not a tax whose economic incidence is on land. Instead, they assume that stamp duties raise the cost of housing to all buyers and renters. In addition, the model assumption requires that stamp duties also raise the cost of building new houses without affecting land prices, leading to reduced new housing construction in general. Garbage in, garbage out.

It’s hard to explain in plain English what is going on in these models. My best explanation is that the model says that stamp duty forces every new home to be built with an additional extension on the house, where the construction cost of the extension is the same as the stamp duty. But then as soon as it is built, the extension is immediately demolished, yet somehow the buyer is forced to pay for it. Meanwhile, all the other households in the economy are somehow forced to also pay more to occupy their existing houses, as if they were now renting non-existent home extensions. 

If that sounds crazy, that’s because it is. There is no actual economic evidence of broad economy-wide costs of stamp duties. Just garbage inventions that don’t look anything like stamp duties.

In sum

Stamp duties don’t reduce the cost of housing. They do reduce asset churn by investors and speculators, which is desirable for price stability, and the apparent economic costs from reducing household mobility are overblown. The ‘revenue instability’ of stamp duties is actually a huge positive for the economy as a whole, and the economic analysis that has underpinned the talk of the high economy-wide costs of stamp duty is simply made up.

All of this is true. Yet SD4LVT is the apparently the holy grail of Australian tax policy. Our best policy wonks want politicians to use up the precious political capital to swap one very good property taxes for another very good but electorally unpopular one, for no net economic gain.


  1. 1.
    "The economic incidence of stamp duty is on the seller."
    -- Can you justify this? The _legal_ incidence is, but I would have thought the economic incidence would be split between the seller and buyer (and any middlemen) in proportion to their price elasticities. If SD were abolished and the sell price jumped 5%, this would surely induce more to sell, causing a minor glut and price drop. I'd imagine the equilibrium price would go from $500k+$25k SD to ~$510k or $515k, and volumes traded would rise commensurately.

    Also, price effects are _supposed_ to be neutral-ish. It's a tax swap, not a tax hike. Specifically, it's a tax swap that removes a market distortion, the excess transactional cost of moving. This is also why I would oppose a moving subsidy: that's a distortion too.

    LVT is anathema to speculation. If we had, say, a 5% land tax, then land prices would have to increase by 5% pa before a speculator so much as broke even, even if his interest rate was zero; and, unlike with SD, this is true in both the long and short term. Under LVT with a reasonably high rate, the only way to make money from land is by productive use, viz by building something on it and charging rent. If churn eats that up, then that's its own incentive to not evict tenants too often.

    1. 1. If the economic incidence of costs of buying or using land were on the buyer, then you could remove those costs incrementally and be left with free land at the end, since there are no input costs to land.

      1.a. I disagree that the theoretical idea of a 'non-distortionary market equilbrium' has any merit. If you want people moving more, subsidise it. If you want them moving less, tax it. This whole 'I want the non-distortionary outcome' is a cop out. But this is a discussion for another day I would say.

      2. Not true. The land price would already take into account the present value of the future stream of 5%pa LVT obligations.

    2. 1.
      I think you're dividing by zero. The inputs of land are zero, but so are the outputs: you can't produce it.

      Even if there were no land taxes of any sort, land would still have a nonzero price because it's scarce. (At least, useful land is scarce. I assume you can buy tracts of desert for a song.) In a reasonably efficient market, the price should be about equal to the NPV of whatever revenue stream can be created with it.

      I personally don't care whether other people move more or less, so a government representing me should neither penalise nor encourage it, subject to asterisks such as externalities or equality of access. They should instead strive to distort in neither direction. Isn't this mainstream free market theory?

      Hypothetical: a LVT of 5% pa; no SD; a certain block of land is priced at $x, after the tax was introduced and the market reached equilibrium; and the value of land reliably increases by 4% pa.

      If I buy the land and hold for 1 year, then I owe $x up-front plus 5% in tax; and my assets are worth $x plus 4%. Even without paying any interest, I've made a loss. If I hold for another year, then I owe the taxman a further 5% of 1.04x, and my assets have increased by 4% of 1.04x, so I make another loss. The tax owed each year exceeds the appreciation 5:4, so I lose money unless I use the land productively.

    3. 1. "the price should be about equal to the NPV of whatever revenue stream can be created with it"

      Exactly. So add stamp duty to the buyer's NPV assessment, and the seller gets exactly that amount less.

      1.a. Yes. Mainstream free market theory does invoke a 'distortion-free' equilibrium. I personally don't think this theory makes any sense.

      2. Think about it this way. You are willing to pay $100k for land with zero income stream based on expectations it will rise by 4% per year.

      Now you have to pay 5% of land value as a tax each year. What happens?

      How much the price falls depends on how much of the $100k you would have to put aside in an account today to pay for future LVT, thus spending the same total upfront in PV terms.

      Let's make it simply and say that the interest rate is 5% and the LVT rate is also 5%.

      In this case, you can pay $50k for the land (the land value), and put $50k in an acount earning 5% interest, which is $100k total upfront. The interest earned on the account exactly equals the future LVT obligation. Paying the 5% LVT is not a loss - it was already accounted for in the price

      If you expect land prices to rise, then these expectations will be built into the upfront price in both situations.

      Does that make sense?

    4. 1.
      That logic surely implies that all sales/consumption taxes are borne entirely by the seller. If this were the case, then marginal transactions would become losses. My understanding is that in practice, the seller raises the price by less than the tax, such that both seller and buyer absorb some of the tax; that is the economic incidence is split between the two.

      If I put half my money into an account earning good interest, then I'm not profitably speculating on land; I'm profitably investing in whatever the account is used for, and also losing money on bad speculation. The two are independent. In that example, the account earns $2.5k a year; but the land costs $2.5k to make $2k, ie it loses $500. I should offload the land ASAP.

    5. 1. It implies that taxes are borne by wherever the economic rents sit upstream. E.g. if you apply a GST to a monopoly taxi company (who is already setting prices to profit-maximise) then the incidence is on the value of (return to) the monopoly licence.

      2. You can see it this way. But they are actually joined by the fact that willingness to pay for the land that is equal in both cases. Either you pay what you value the land upfront in one lump sum. Or you pay less upfront and more ongoing in the form of LVT. The PV of both expected cash-flows is will be equal. Perhaps this needs another blog post with some examples.

    6. 1.
      Hypothetical: a monopoly with marginal cost = 10 and demand(price) = 20-price. The profit-maximising price is 15. If a $2 tax is imposed, the profit-maximising price rises to 16 including tax. Both monopoly and consumer take a hit.

      I think the point here is that both parties have some economic rent -- as you'd expect, since both agree to do business -- so both take some incidence. But that applies in the housing market too.

      I'm not willing to pay for the land, not as a speculator. It loses money to tax faster than it gains it to appreciation, and will never turn a profit. I'd only buy it if I meant to use it or rent it out.

    7. 1. Not if the tax is applied as a % of profit, or as a % of revenue in the zero MC input case, like in land.

      2. Future blog posts coming.

    8. SD isn't just on land, though: it's also on improvements on the land, and those certainly aren't 0 MC. SD, at a minimum, should act like GST on the improvements; except that, unlike GST, inputs aren't claimed.

      Even for the land itself, the MC is the price you pay to buy it in the first place, which is nonzero because land is scarce and useful. In a hypothetical society with no taxes of any sort on land, a parcel of land should have a price of about the NPV of the best project that can only be done with that parcel, which is surely nonzero.

  2. I can’t resist jumping in with one abstract but interesting (to me) comment.


    1a. “I personally don't care whether other people move more or less, so a government representing me should neither penalise nor encourage it, subject to asterisks such as externalities or equality of access. They should instead strive to distort in neither direction. Isn't this mainstream free market theory?”

    That “government should aim for distortion-free i.e. economically efficient markets” is an ethical view. There are lots of others.

    It’s better labelled the “free market ethical judgement” than theory. There isn’t – because there cannot be – any theory that establishes this particular judgement as desirable, at least within economics.

    Zero-distortion markets do maximise something called “economic welfare” or “preference satisfaction” which economists identify (i.e. model and measure) as willingness-to-pay. So economic efficiency = maximising WTP.

    WTP obviously reflects pre-existing power (wealth). So economic efficiency as a social goal amounts to an intellectual curtain covering up preservation of status quo power inequities.

    Practical implications include concluding it is better for corn to fuel cars than feed starving people, better to ship toxic waste to countries where they are too poor to really object (google “Summers world bank memo”), etc. The efficiency of individual wage bargaining is another where talk of efficiency hides the important variable, power.

    In present context, “having the means to risk speculative investments and influence politics to gain economic rents” might be the kind of power inequity that non-distortionary land transfer markets play out and reinforce.

    Obviously the ethical rebuttal hasn’t stood in the way of many economists forging well-rewarded careers promoting economic efficiency as a social goal. See Productivity Commission, every treasury department, etc.

    (I know this was way off topic, but it's a bugbear of mine that many policy economists don't understand their hidden ethical assumptions.)

    1. To clarify: I wasn’t necessarily critiquing non-distortionary land transfer markets.

      As it happens I think free land transfer markets might be inherently inefficient/distorted and there is a market failure case for policy to correct them – but that’s way off topic and too involved for here.

  3. This comment has been removed by the author.